1

Aggregation Latency-Energy Tradeoff in Wireless Sensor Networks with Successive Interference Cancellation Hongxing Li∗ , Chuan Wu∗ , Dongxiao Yu∗ , Qiang-Sheng Hua† and Francis C.M. Lau∗ ∗ Department of Computer Science, The University of Hong Kong, Hong Kong † Institute for Theoretical Computer Science, Tsinghua University, China Email: {hxli, cwu, dxyu, qshua, fcmlau}@cs.hku.hk, [email protected]

Abstract—Minimizing latency and energy consumption is the prime objective of the design of data aggregation in battery-powered wireless networks. A tradeoff exists between the aggregation latency and the energy consumption, which has been widely studied under the protocol interference model. There has been however no investigation of the tradeoff under the physical interference model which is known to capture more accurately the characteristics of wireless interferences. When coupled with the technique of successive interference cancellation, by which a receiver may recover signals from multiple simultaneous senders, the model can lead to much reduced latency but increased energy usage. In this paper, we investigate the latency-energy tradeoff for data aggregation in wireless sensor networks under the physical interference model and using successive interference cancellation. We present theoretical lower bounds on both latency and energy as well as their tradeoff, and give an efficient approximation algorithm that can achieve the asymptotical optimum in both aggregation latency and latency-energy tradeoff. We show that our algorithm can significantly reduce the aggregation latency, for which the energy consumption is kept at its lowest possible level. Index Terms—Data aggregation, Latency-energy tradeoff, Wireless sensor network, Successive interference cancellation.

✦

1 I NTRODUCTION Wireless sensor networks have been extensively exploited for many environment monitoring applications in recent years. One of the core functions in these networks is data aggregation, which is to collect data from the wireless sensor nodes to deliver to a sink node. Typically, data aggregation is initiated by the sink using some SQL-like queries, such as “to find the highest temperature in the region”. Messages generated at individual sensors carrying temperature data, are first aggregated and processed at some relay sensors, e.g., to derive the local maximum temperature; the locally processed results are further aggregated, and so on, until the final result reaches the sink. Besides the max function, other functions such as min, sum, count, and average can all be effectively implemented using data aggregation. As the sensed data typically has a limited duration of validity, a fundamental requirement is that the total aggregation time, measured in time units and also referred to as the aggregation latency, must be minimized [1]–[3]. Additionally, the sensor nodes have to observe the hard constraint imposed by battery power and must strive for low energy consumption in each run of the data aggregation. Obviously, there exists some kind of tradeoff between aggregation latency and energy consumption (the latencyenergy tradeoff) in wireless sensor data aggregation [4]–[6]. There have been some efforts to derive latency-energy tradeoff theoretically [7] as well as practical algorithms [4]– [6], which are all based on the protocol interference model (or equivalently the pair-wise interference model). Under the protocol interference model, the transmission range and interference range of a node are simplified to two disks with radii rt and ri (ri ≥ rt ), respectively. A transmission is successful if and only if the receiver lies within the transmis-

sion range of the sender and outside the interference range of any other concurrent sender. There has been however no prior study that is based on the physical interference model (or the cumulative interference model) which has been shown to be able to more accurately characterize the wireless interferences than the protocol interference model [8]–[10]. Designs based on the physical interference model can lead to increased network capacity. Under the physical interference model, the cumulativePinterference from all α concurrent transmissions, e.g. the ej ∈Λi Pj /dji part in Eqn. (1), is taken into consideration at each receiver. A transmission along link ei is successful if the Signal-toInterference-plus-Noise-Ratio (SINR) at its receiver is above a certain threshold: α N0 +

P /d P i ii

ej ∈Λi

Pj /dα ji

≥ β.

(1)

Here Λi denotes the set of links that transmit simultaneously with ei . Pi and Pj denote the transmission powers at the transmitter of link ei and that of link ej , respectively. dii (dji ) is the distance between the transmitter of link ei (ej ) and the receiver of link ei . Fig. 1 explains these distances graphically. α is the path loss ratio which has a typical value of between 2 to 6. N0 is the ambient noise power. β is a positive constant as the SINR threshold for a successful transmission [3], [11]. ii i j ji jj

Fig. 1: An illustration of distances with two transmission links: ei and ej . With the physical interference model, a receiver can only successfully recover one signal from one sender in each

2

time slot, among possibly several or many simultaneous transmissions. In recent years, it has been shown that, by applying Multi-Packet Reception (MPR) techniques [12], it is possible to break this one-timeslot-one-sender barrier to let a receiver recover multiple individual signals from the mixed signal coming from multiple simultaneous senders. Successive Interference Cancellation (SIC), one subcategory of MPR, has been demonstrated practical by experimental study [13] implemented for the IEEE 802.15.4 [14] (ZigBee) physical layer (a common physical-layer standard for sensor networks and other wireless personal area networks). The idea of SIC is to repeatedly identify the strongest signal and then remove (cancel) it from the mixed one using channel estimation, signal regenerating and subtraction [15]. To make it work, an interference cancellation sequence needs to be identified, such that for the ith signal to be canceled, the following criterion is satisfied: N0 +

P

Pi /dα ii P ≥ β, α α ej ∈Λi −Γi Pj /dji + ek ∈Γi ,k≻i Pk /dki

(2)

where Γi is the set of concurrent transmission links connecting to the same receiver as ei ’s, and k ≻ i denotes that link ei is canceled before link ek . Extra energy is needed to recover the ith signal if it is not the last canceled one, in order to compensate for the cumulative interference from those links that are later canceled. Meanwhile, extra decoding delay, proportional to the number of canceled signals, is incurred for the entire signal cancellation process [13]. SIC techniques can potentially reduce the aggregation latency in wireless sensor networks significantly, because multiple transmissions can be scheduled in the same time slot while the saved scheduling latency overwhelms the incurred decoding delay (To be compared in Sec. 3.2). Inevitably, the cost is increased energy consumption. To the best of our knowledge, there is no previous study that has tried to characterize the latency-energy tradeoff under SIC. Such characterization is needed in order to accurately gauge the practical benefits of applying SIC in typical wireless applications. In this paper, we investigate the aggregation latencyenergy tradeoff in wireless sensor networks under the physical interference model with successive interference cancellation. Our contributions are as follows: ⊲ We prove a theoretical lower bound on the aggregation latency under the physical interference model with SIC: Ω(max{D, logX+1 n}), where D is the network diameter in terms of the number of hops (the maximum of the minimum number of hops between any pair of nodes, when the nodes are transmitting using PM and scheduled without mutual interference), n is the number of nodes, PM and X = ⌊log1+β N + 1⌋ with PM being the maximum 0β transmission power of any node. ⊲ We prove a theoretical lower bound, applicable to both the case with and that without SIC, on energy consumption und M )α , where der the physical interference model: N0 β (nmis α−1 n nmis is the size of the maximum independent set with PM (see Definition 1 in Section 4) of the given network and dM is the maximum transmission range with maximum power

PM and zero interference. ⊲ We prove a theoretical lower bound on the latencyenergy tradeoff under the physical interference model with SIC that, for any aggregation algorithm, the product of the energy consumption approximation ratio and the (α − 1)th power of the aggregation latency approximation ratio is lower bounded by Ω(∆α−1 ), where ∆ is the maximum node degree (maximum number of nodes within the transmission range dM of any node). ⊲ We propose EMA-SIC, an Energy-efficient Minimumlatency Aggregation algorithm under the physical interference model with SIC. As compared to existing work [3], [11] on minimum-latency data aggregation under the physical interference model, EMA-SIC can significantly lower the upper bound of the aggregation latency, to O(D), and at the same time achieves an energy consumption approximation ratio that is the lowest possible with respect to the latency-energy tradeoff lower bound. In other words, our proposed algorithm achieves the asymptotical optimum in both aggregation latency and latency-energy tradeoff. The remainder of the paper is organized as follows. We discuss related work in Sec. 2, and present the problem model in Sec. 3. We study the theoretical lower bounds for aggregation latency, energy consumption and their tradeoff in Sec. 4. The EMA-SIC algorithm and its analysis are presented in Sec. 5 and 6, respectively. The latency-energy efficiency of EMA-SIC is further studied via extensive simulations in Sec. 7. Finally, we conclude the paper in Sec. 8.

2 R ELATED WORK 2.1 Minimum-Latency Data Aggregation There is a large body of literature on data aggregation in wireless sensor networks [1]–[3], [11], [16]–[19]. Most of the work target at minimum aggregation latency, without much consideration of the energy consumption. The current best upper bound on aggregation latency is O(∆+R), which is based on the protocol interference model [1], [2], [16]– [18], where R is the network radius in hops and ∆ is the maximal node degree. The paper [1] is the first work that achieves the O(∆+R) aggregation latency upper bound. In [2], the minimumlatency data aggregation problem in a multihop wireless sensor network with the assumption that each node has a unit transmission range and an interference range of ρ ≥ 1 is studied. Xu et al. [17] propose an aggregation schedule based on a distributed algorithm, which achieves a guaranteed maximum aggregation latency of 16R + ∆ − 14; they also prove a lower bound of max{R, log2 n} on the aggregation latency for any interference model, where n is the network size. Different from the above work where connected dominating sets or maximal independent sets are employed, a novel approach of distributed aggregation with latency bound in O(∆ + R′ ) is introduced in [16] by clustering. Here, R′ is the inferior network radius satisfying that R′ ≤ R ≤ D ≤ 2R′ with D as the network diameter in hop-count. The MLAS problem is extended to the case with multiple sinks in [19] with latency bound of O(∆ + kR), where k is the number of sinks.

3

To the best of our knowledge, only two papers, [3] and [11], assume the physical interference model. A distributed aggregation scheduling algorithm with constant power assignment is proposed in [3], which achieves a latency upper bound of O(∆ + R). Li et al. [11] present a distributed algorithm with a latency bound of O(K), where K is the logarithm of the ratio between the length of the longest link and that of the shortest link, and a centralized solution with an aggregation latency of O(log3 n) which is the current best result among all proposed aggregation algorithms under the physical interference model. However, no limit on the power is assumed in [11]. 2.2 Latency-Energy Tradeoffs in Data Aggregation The existence of a tradeoff between energy consumption and aggregation latency in wireless sensor networks is widely recognized. There were some attempts targeting at efficient data aggregation algorithms with both low aggregation latency and low energy usage [4]–[7], [20], but all are based on the protocol interference model expect [20] which only considers primary interference without mutual interference from other concurrent transmissions. Yu et al. [4] explore the latency-energy tradeoff using techniques such as modulation scaling; algorithms are proposed to minimize the total energy consumption subject to a specified latency constraint. Arumugam et al. [5] propose a TDMA-based algorithm to effectively aggregate data in an energy-efficient way. In [6], the source node can specify its interest in minimizing energy consumption and/or source-to-sink delay, as input to the aggregation algorithm. The theoretical analysis in [7] demonstrates that there exists a latency-energy tradeoff in sensor data aggregation; an aggregation algorithm is designed, which achieves the asymptotical optimum for the tradeoff under the protocol interference model. To address the latency-energy tradeoff for in-network computation, of which data aggregation is a special case, an algorithm with order-optimal energy usage under given latency constraint is proposed in [20]. However, its order-optimality is derived over a network of uniformly random node distribution with simplified interference model as discussed previously, while not guaranteed over arbitrary network topologies. To the best of our knowledge, there has been no work addressing the tradeoff under the physical interference model, not to mention the case with arbitrary network topologies and/or SIC technique, which is targeted in this paper. 2.3 Successive Interference Cancellation The techniques of successive interference cancellation have been exploited in recent years. Weber et al. [15] have analyzed the transmission capacity of wireless ad-hoc networks using SIC, with both upper bound and lower bound in closed form. Simeone et al. [21] analyze the capacity of linear two-hop mesh networks with SIC; a decode-and-forward relaying mechanism is proposed by exploiting the possible relevant inter-cell channel gains and rate splitting with SIC. Wang et al. [22] present a polynomial-time heuristic algorithm to approximate the optimal network throughput in ad hoc networks with joint routing and scheduling using SIC.

Lv et al. [23] propose simultaneity graph to characterize the effect of SIC on link dependence due to interference, and present an independent set based greedy scheme to construct a maximal feasible schedule. Jiang et al. [24] advocate the use of joint SIC and interference avoidance and introduce a cross-layer optimization framework for the joint scheme. In [25], a SIC-based scheduling algorithm, with polynomialtime complexity, is proposed to find short schedules for networks with arbitrary distribution in the Euclidean plane. However, none of the above considers the decoding delay with SIC. Our paper addresses this issue.

3 T HE P ROBLEM M ODEL We consider a multi-hop wireless sensor network with n arbitrarily distributed sensor nodes v0 , v1 , . . . , vn−1 and a sink node vn . The directed graph G = (V, E) denotes the tree constructed for data aggregation from the sensor nodes to the sink, where V = {v0 , v1 , . . . , vn } is the set of all nodes, and E = {e0 , e1 , . . . , en−1 } is the set of transmission links in the tree with ei representing the link from sensor node vi to its parent. Without loss of generality, we assume that the minimum Euclidean distance between each pair of nodes is 1. We consider a time-slotted system. The transmission delay of one packet and the decoding delay to cancel one more signal with SIC are normalized to 1 time unit and τ time unit, respectively. The actual length of any time slot t is (1 + χt · τ ) time units, with χt as the maximum number of canceled signals at any scheduled receiver in this slot depending on whether the SIC technique is applied: χt = 0 if there is no SIC application, while χt > 0 otherwise. 3.1 The Data Aggregation Problem The data aggregation problem is to use the links in E to construct a suitable tree and to design a correct and collisionfree aggregation schedule S = {S0 , S1 , . . . , ST −1 }, where T is the total time slots for the schedule and St denotes the subset of links in E scheduled to transmit in time slot t, t = 0, . . . , T − 1. A correct aggregation schedule must satisfy the following conditions. ST −1First, any link should be scheduled exactly once, i.e., t=0 St = E and Si ∩ Sj = ∅ where i 6= j. Second, primary interference (that due to a node acting as a transmitter and a receiver in the same time slot) should be avoided; that is, T (St ) ∩ R(St ) = ∅, ∀t = 0, . . . , T − 1, where T (St ) and R(St ) denote the transmitter set and receiver set for the links in St , respectively. Third, a non-leaf node vi transmits to its parent only after all the links in the subtree rooted at vi have been scheduled, i.e., T (Si ) ∩ R(Sj ) = ∅ where i < j; in this way, vi can conduct local process to aggregate all data from its subtree (e.g. local maximum temperature of its subtree) and transmits the aggregated data only once to save energy. An aggregation schedule is collision-free if each scheduled transmission in time slot t, i.e., ∀ei ∈ St , can be correctly received by its receiver according to the interference model in Sec. 3.3, ∀t = 0, . . . , T − 1. Our objective is to minimize the aggregation latency, i.e., the overall time units of all slots, as well as the latency-energy tradeoff. Note that the aggregation latency already includes

4

α β N0 PM dM ∆ n

Path loss ratio, Sec. 1 SINR threshold, Sec. 1 Ambient noise power, Sec. 1 Max. transmission power, Sec. 1 Max. transmission range with PM , Sec. 1 Max. node degree, Sec. 1 Number of sensor nodes in network, Sec. 1

D nmis X χt τ h ncds

Network diameter in terms of hop count, Sec. 1 Size of maximum independent set, Sec. 1 Max. cancelable signals at one receiver, Sec. 1 Max. canceled signals at each receiver, slot t, Sec. 3 Decoding delay to cancel one more signal, Sec. 3 Side length of hexagons, Sec. 5 Size of connected dominating set, Appendix I

TABLE 1: Notation table. the end-to-end transmission delay, decoding delay with SIC 4 T HEORETICAL L OWER B OUNDS and the cumulative queueing delay, since it is defined as the We first investigate the theoretical lower bounds on the agtime-span between the time-point of first transmission and gregation latency, the energy consumption and the latencythat when the sink collects all data. energy tradeoff for the data aggregation problem, respec3.2 Decoding delay with SIC tively. With the SIC technique for ZigBee standard in [13], one packet typically has a length of 128 bytes, which are 4.1 Energy Consumption Bound modulated into 4096 physical-layer symbols. Symbols of We prove in Theorem 1 a lower bound, applicable to each signal are decoded and canceled sequentially, with the both the case with and that without SIC, on the overall requisite that three consecutive symbols should be buffered energy consumption for data aggregation under the physical for each canceled symbol. The decoding delay for each interference model. canceled signal is the time span for 3 symbols, which is Definition 1. (Maximum Independent Set with PM ) An 3 τ = 4096 of the transmission delay. If we have χt signals independent set with PM in a wireless sensor network G is 3χt to cancel out, the total decoding delay is 4096 time units a subset of nodes in the network graph, such that no node while the saved transmission delay is χt time units, i.e., in the set can successfully transmit to another node in the one time unit for each canceled signal. Thus, SIC has great set using the maximum power P with zero interference, M potential in reducing the aggregation latency in wireless and a maximum independent set with P is the largest M sensor networks. The above setting for decoding delay is such independent set in the graph, i.e., it has the maximum also applied in our simulation study in Sec. 7. number of nodes. 3.3 Interference and Energy Models We adopt the physical interference model with the appli- Theorem 1. (Energy Consumption Lower Bound) Supcation of successive interference cancellation. With SIC, pose the size of the maximum independent set with PM a receiver can recover multiple signals from simultaneous containing the sink in a wireless sensor network is nmis +1. transmitters from the mixed signal received, as long as an The overall energy consumption for data aggregation in with or interference cancellation sequence of the signals can be the network under the physical interference model, (nmis dM )α th without SIC, is lower-bounded by N β . 0 determined. The sequence is such that the i signal remains nα−1 strong enough, as judged by condition (2), after the previous We prove Theorem 1 by analyzing the energy consumpi − 1 signals have been removed (canceled) from the mixed tion when links are scheduled in a “TDMA” fashion, i.e., signal. If SIC is not applied at a receiver, the receiver can only one link is scheduled to transmit in each time slot, and recover at most one signal (from one sender) in each time the data aggregation tree is a minimum spanning tree of the slot, subject to condition (1). network, of which the weight of link ei is dα ii , where dii is In our study, we use the energy model that the power the Euclidean length of link ei . More details can be found attenuation along each transmission link of length r is in Appendix A. proportional to rα , α ≥ 2, i.e., the received power is P/rα if the sender uses transmission power P . 4.2 Aggregation Latency Bound Let Pi denote the transmission power used by node In their recent work, Xu et al. [17] give a latency lower vi , i = 0, . . . , n − 1, and the maximum transmission power bound of Ω(D + ∆) on sensor data aggregation under the at any sensor node be PM .1 We assume no isolated node, protocol interference model, and a latency lower bound of i.e. each node can transmit to at least one other node in Ω(max{R, log n}) under any interference model. We next 2 the network if the power level of PM is used. Let D be the prove an aggregation latency lower bound under the physical network diameter, which is in hops instead of the geometric interference model with SIC. distance, and is defined as the maximum of the minimum Theorem 2. (Aggregation Latency Lower Bound) The lanumber of hops between any pair of nodes when the nodes tency of data aggregation in a wireless sensor network under are transmitting using maximum power PM ; and dM be the the physical interference model with SIC, is lower-bounded maximum transmission range of a node when using PM PM by Ω(max{D, logX+1 n}), where X = ⌊log1+β N + 1⌋. 0β with zero interference. We first demonstrate that X is the maximum cancelable Important notations used in the paper are summarized in signals at one receiver, and then prove Theorem 2 by showtable 1 with descriptions and places of first appearances. ing that the aggregation latency lower bound is achieved 1. We consider homogeneous networks with identical maximum transusing maximum transmission power with D, n and X as mission power on each node. dominant factors. Detailed proof is in Appendix B.

5

4.3 Latency-Energy Tradeoff The lower bounds on aggregation latency and overall energy consumption, as just derived, may not be achievable concurrently: the lower bound on energy consumption given in Theorem 1 is achieved only when the aggregation tree is a minimum spanning tree of the network and exactly one transmission along the tree is scheduled in each time slot. In this case, the aggregation latency is n. On the other hand, to achieve the lower bound on aggregation latency of Ω(max{D, logX+1 n}) as given in Theorem 2, larger powers up to PM at the transmitters may need to be used. Consequently, the tradeoff between aggregation latency and energy consumption needs to be addressed in the design of any data aggregation algorithm, which is the main objective of this paper. Theorem 3 presents a theoretical lower bound on the combined performance of aggregation latency and energy consumption. The theorem is not to establish a definition for the latency-energy tradeoff, which may vary with different application concerns. However, the result in Theorem 3 can serve as a metric to examine whether an algorithm has achieved the best it can do in terms of both aggregation latency and energy consumption, with a tradeoff in between. A similar metric can be found in [7] under the protocol interference model. Theorem 3. (Latency-Energy Tradeoff Lower Bound) Let ρL and ρE denote the approximation ratios of the aggregation latency and energy consumption with regard to the lower bounds in Theorems 1 and 2, respectively, with any given data aggregation algorithm. The product of the energy consumption approximation ratio, i.e., ρE , and the (α− 1)th power of the aggregation latency approximation ratio, i.e., α−1 ρα−1 ) in wireless networks L , is lowered-bounded by Ω(∆ under the physical interference model with SIC. To prove the theorem, we show that there exists a sample network of n + 1 nodes with maximum node degree ∆ such α−1 that ρα−1 ) under the L ρE is lowered-bounded by Ω(∆ physical interference model with SIC. See more details in Appendix C. We will show in Sec. 6 that, our algorithm, to be proposed in Sec. 5, is asymptotically optimal in both aggregation latency and latency-energy tradeoff, with respect to the lower bounds in Theorems 1, 2 and 3.

5 EMA-SIC: E NERGY - EFFICIENT M INIMUM LATENCY DATA AGGREGATION ALGORITHM WITH S UCCESSIVE I NTERFERENCE C ANCEL LATION We now design our EMA-SIC algorithm, which can outperform any other existing algorithm by its reduced maximum aggregation latency and asymptotically optimal latencyenergy tradeoff. It uses the successive interference cancellation technique, and consists of two parts: tree construction (T) and link scheduling (S). 5.1 Tree Construction with SIC The aggregation tree construction in EMA-SIC comprises three steps, executed in a distributed fashion.

T1) Breadth-first search is launched by the sink vn to find a spanning tree of the network rooted at it, based on maximum transmission ranges of the nodes. Each node is assigned a level, indicating its hop-count to the sink. The sink node is initialized with level 0. T2) A connected dominating set (CDS) of the network (see Definition 2) is identified in the breath-first spanning (BFS) tree, by treating the sink as the first dominator and then finding other dominators using the most widely adopted algorithm, which is distributed, in [26]. This algorithm is executed in two phases to find the connected dominating set. In the first phase, a maximum independent set is constructed such that the distance between any pair of its complementary subsets is exactly two hops. Based on the constructed MIS, the second phase generates a connected dominating set by strategically selecting nodes to be added to or removed from the MIS. A tree rooted at the sink and connecting all other nodes in the connected dominating set can be built, such that each node in level l ≥ 1 connects to its parent node in level l − 1 of the BFS tree. Definition 2. (Connected Dominating Set with PM ) A dominating set with PM in a wireless sensor network G is a subset of nodes in the graph, such that every node outside the set can successfully transmit to at least one node in the set using the maximum power PM . The nodes in a dominating set are referred to as dominators, and those not in the set are dominatees. A connected dominating set with PM is a dominating set within which any node can transmit to at least another node using PM . Dominator Head dominatee

(a) Concurrent transmissions to the head dominatee.

Other Dominatee Link Constructed

(b) MST rooted at the dominator.

Fig. 2: The third step of tree construction in EMA-SIC: an example with one dominator. T3) This step consists of two phases. Consider the dominators in the connected dominating set derived in the previous step. In the first phase, i.e., step T3.a, each dominator finds a disk centered at itself with radius equal to the maximum transmission range dM , and use equalsized hexagons to cover the disk. An example is given in Fig. 2(a). The side lengthqof the hexagons √ is h = −1+

1+4/3(1+log

PM /(N0 β)

1+β min{d1 , d2 } with d1 = 2 1 1 PM 2α and d2 = 2 ( N0 β ) , which are carefully assigned to ensure the validness of our algorithm, i.e., K1 > 0 which is a parameter to be introduced shortly, proven in Appendix D. In each hexagon, the dominatee which is closest to the dominator is chosen as the head dominatee. We prove in Sec. 6 that, with our assignment of the hexagon side length, all dominatee nodes in a hexagon can concurrently transmit

6

to the head dominatee which can successfully recover these transmissions using successive interference cancellation. Note that, each dominator only needs the location information of the dominatees within the disk as in Fig. 2. Besides, the relative coordinate (x, y) of each dominatee with its dominator as the origin (0, 0), is adequate to decide which hexagon it resides in and whether it is the head dominatee, without the use of absolute coordinates in the global view. So the step T3.a can be executed in a fully distributed fashion on each dominator just with relative locations of its dominatees. In the next phase, i.e., step T3.b, for the much sparser topology consisting of only the head dominatees and the dominators, a local minimum spanning tree (MST) is built to connect the head dominatees to each dominator, as shown in Fig. 2(b). In constructing the MST, a link with length r has weight rα , which reflects the power attenuation along the link. Using the connected dominating set as the backbone, a data aggregation tree of the entire network is formed. In the above procedure, if a dominatee happens to reside in the overlapping area of hexagons or disks belonging to different dominators, it chooses to join the tree construction of the dominator geometrically closest to the sink. Fig. 12 in Appendix E illustrates the tree construction procedure with an example. 5.2 Link Scheduling with SIC The aggregation schedule consists of three steps: (S1) schedule the transmissions in individual hexagons from the non-head dominatees to the head dominatee; (S2) schedule the transmissions from head dominatees to their dominators along the local minimum spanning trees; (S3) schedule the aggregation transmissions of the dominators along the tree connecting them to the sink. For step S1, within each hexagon, all the non-head dominatees transmit concurrently to the head dominatee. In order for the head dominatee to recover all the transmissions correctly, the transmission power for the ith link in the cancellation sequence, which has a link length of ′ dii , is assigned to be (N0 + I)β(1 + β)X −i dα ii , where PM /(2h)α I = β(1+β)X ′ −1 − N0 is the upper bound of the cumulative interference from current transmissions in other hexagons and X ′ = 3h2 + 3h is the maximum number of non-head dominatees in any hexagon. The detailed derivation of I, X ′ and the power assignment can be found in the proof of Theorem 4 in Sec. 6 to show the correctness of EMA-SIC. To alleviate inter-hexagon interferences, a schedule of hexagons is designed following the rule that the head dominatees in any two hexagons, of which the transmissions are scheduled in the same time slot, should be separated by a distance of greater than K1 + 1 times the maximum 1 link length 2h in a hexagon, where K1 = (6X ′ ) α (1 + α 1 1 N0 (2h) − 1 1 ( √23 )α α−2 ) α ( β(1+β) ) α. X ′ −1 − PM To achieve collision-free link scheduling in steps S2 and S3, we apply the following two rules: (i) any two concurrent transmitters should be separated by a distance of at least 1 (K2 + 1)dM , with K2 = (6β(1 + ( √23 )α α−2 ) + 1)1/α , in order to bound the cumulative interference at each receiver; (ii) the transmission power for a link of length r is set

to N0 β(2 − 1/K2α )rα . This idea of separating concurrent transmitters by a predefined distance was also employed in [27] which however did not consider background noise in the interference model (we do here). Note that, similar approaches of covering the network with hexagons are also utilized in aggregation algorithms of [3] and [11], but with fundamental differences from ours in following ways: (i) in [3], the network is covered with equalsized grids just for link scheduling across different grids without contribution to tree construction while our application of hexagons both constructs the aggregation tree and schedules link transmissions within the same hexagon; (ii) in [11], hexagons with differentiated sizes are iteratively used to construct the tree but with unlimited power assignment while our paper shows practical concern with maximum transmission power and applies hexagons with unique sizes for only part of the tree construction; moreover, absolute locations of all nodes in the global view are required in [11] while our paper just needs a weaker condition of relative position of each dominatee to its dominator. The EMA-SIC algorithm is summarized in Algorithm 1.

6

OF EMA-SIC We next prove the correctness, as well as the latency and energy efficiencies of our algorithm.

6.1

A NALYSIS

Correctness

Theorem 4. (Correctness of EMA-SIC) EMA-SIC constructs a data aggregation tree and achieves a correct and collision-free aggregation schedule under the physical interference model. Proof Sketch: It is easy to see an aggregation tree rooted at the sink is correctly constructed from the tree construction algorithm in EMA-SIC. We prove that EMA-SIC achieves a correct and collision-free aggregation schedule (see Sec. 3.A) by presenting upper bounds for cumulative interferences at each receiver in each step of the link scheduling, and showing that each received signal in step S1 and in steps S2 and S3 of the link scheduling satisfies the SINR constraint in Eqn. (2) and (1) even with its corresponding upper bound of cumulative interferences, respectively. The detailed proofs can be found in Appendices F and G, respectively. ⊓ ⊔ 6.2 Energy and Latency Efficiencies We show that EMA-SIC outperforms any other algorithm as it can reduce the maximum aggregation latency to O(D), while maintaining an energy consumption approximation ratio that is the lowest possible—O(∆α−1 ). Theorem 5. (Latency Efficiency with EMA-SIC) The aggregation latency with EMA-SIC in any given network with network diameter D, is upper-bounded by O(D), and the aggregation latency approximation ratio (with respect to the lower bound in Theorem 2) is upper-bounded by O(1). We prove Theorem 5 by showing that the scheduling latencies for step S1 and S2 are bounded with constant values while step S3 has an upper-bounded latency in the order of O(D). See detailed proof in Appendix H.

7

Algorithm 1 EMA-SIC Algorithm Input: Node set V and the sink node vn . Output: Aggregation tree E and link schedule S. 1: Initialization: E, S ← ∅. 2: Step T1: Construct a BFS tree on V rooted at vn . 3: Step T2: Construct a CDS on the BFS tree which includes vn ; build a spanning tree of the dominators rooted at vn ; add tree links to E. 4: Step T3.a: Cover the network with hexagons and connect each non-head dominatee to its head dominatee in the hexagon; add links to E. 5: Step T3.b: For each dominator, construct a local MST of its head dominatees rooted at√ it; add tree links to E. q 6:

−1+

1+4/3(1+log

PM /(N0 β))

1+β ; d2 := d1 := 2 1 1 PM 2α ′ 2 ( ) ; h := min{d , d }; X := 3h + 3h; 1 2 2 N0 β α PM /(2h) I := β(1+β)X ′ −1 − N0 ;

K1

:=

1

1

1 1 (6X ′ ) α (1 + ( √23 )α α−2 ) α ( β(1+β) X ′ −1 −

N0 (2h)α − 1 ) α. PM

7:

8: 9:

10:

Step S1: Schedule transmissions in hexagons from nonhead dominatees to their head dominatees, such that any two concurrent receivers are separated by at least 2(K1 +1)h, and the transmission power is (N0 +I)β(1+ ′ th β)X −i dα link in the receiver’s cancellation ii for the i sequence with length dii ; add schedule to S. 1 K2 := (6β(1 + ( √23 )α α−2 ) + 1)1/α . Steps S2 & S3: Schedule the link transmissions in the aggregation tree containing only head dominatees and dominators, such that any two concurrent transmitters are separated by at least (K2 + 1)dM and transmission power is N0 β(2 − 1/K2α )rα for a link of length r; add the schedule to S. return E and S

Theorem 6. (Energy Efficiency with EMA-SIC) The energy consumption approximation ratio, that is, the upper bound of the overall energy consumption with using EMASIC to the lower bound in Theorem 1, is upper-bounded by O(∆α−1 ), in any given network with a maximum node degree of ∆. To prove Theorem 6, the upper bounds of energy consumption for step S1, S2 and S3 are analyzed and characterized with nmis , respectively. Detailed proof can be found in Appendix I. The following corollary shows that the energy consumption approximation ratio in Theorem 6 is indeed tight, for any algorithm achieving the aggregation latency upper bound in Theorem 5. Corollary 1. (Asymptotic optimum with EMA-SIC) The aggregation latency and the latency-energy tradeoff with EMA-SIC in any given network, are asymptotically optimal, or equivalently, with O(1) approximation ratios with respect to the lower bounds in Theorem 1, 2 and 3. This corollary can be easily proved by checking the approximation ratios of aggregation latency and energy consumption of EMA-SIC in Theorem 5 and 6, as well as

the latency-energy tradeoff lower bound in Theorem 3. Comparing our analytical results in Theorems 5 and 6 with those in [11] and [3], we can see that EMA-SIC reduces the upper bound of the maximum aggregation latency to O(D) (which is the current best result in literature), and at the same time achieves an approximation ratio of energy consumption that is the lowest possible (Corollary 1).

7 S IMULATION RESULTS We have presented the asymptotic performance of EMASIC in terms of aggregation latency and energy consumption together with their tradeoff by analyzing the respective upper bounds and approximation ratios in the worst cases. In this section, we further investigate the latency-energy efficiencies of EMA-SIC in average cases by comparison with two distributed aggregation algorithms under the physical interference model: Li et al.’s algorithm in [3] and the CellAS algorithm in [11]. We conduct our simulation in the Sinalgo [28] simulation framework, a packet-level wireless network simulator for testing and validating network algorithms. Using the similar setting as that in [11], we consider wireless sensor networks having 100 to 1000 nodes that are randomly distributed with Uniform, Poisson or Cluster distributions2 in a square field with side length from 100 to 200 meters.3 Fig. 14 in Appendix J gives an illustration of network topologies with 100 nodes under different distributions. The power of the background noise N0 is set to a constant 10−6 joule/timeunit. Since the path loss ratio α has a typical value between 2 and 6 and the SINR threshold β is generally assumed to be larger than 1 [3], [11], α is assigned 3, 4 and 5 in the various settings, while β is set to 2, 4, 6 and 10, 15, 20 for the low SINR and high SINR scenarios, respectively. The maximum transmission powers for Li et al.’s algorithm and the EMA-SIC algorithm are assigned values that would result in a transmission range of 40 meters and can maintain the network connectivity with high probability.4 The maximum transmission power in Cell-AS algorithm is infinite since no power limitation is assumed [11]. The decoding delay with SIC is calculated as in Sec. 3.2. Each datum is an average of 100 trials. Li et al.’s algorithm is only effective with uniform node distribution, network side length of 180 and 200, and (α, β) being the pairs (4, 2), (5, 2), (5, 4), and (5, 6), consistent with the report in [11]. We therefore compare the latency-energy performance of the three algorithms under those settings in Fig. 3–10. Complete simulation results with other node distributions, network side lengths and (α, β) value pairs can be found Appendix L. 7.1 Aggregation latency and energy consumption Fig. 3 and Fig. 45 show that EMA-SIC outperforms both the Cell-AS and Li et al.’s algorithms in aggregation latency in 2. Please refer to [11] for detailed explanation of each distribution. 3. With given number of nodes in the network, varying the network scale will change the node density. 4. Disconnected networks are meaningless in our problem since the sink cannot receive the data from all sensor nodes. 5. Results with network side length 180 under similar settings of Fig. 3– 7 can be found in Appendix K.

8

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) α = 4, β = 2

600 400 200

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) α = 5, β = 4

(d) α = 5, β = 6

Fig. 3: Aggregation latency (time units) comparison under selected network settings in a 200 × 200 m2 area.

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8

Energy consumption

Energy consumption

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) α = 4, β = 2

(b) α = 5, β = 2

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

9

Energy consumption

Energy consumption

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4

(d) α = 5, β = 6

Fig. 4: Energy consumption (joule) comparison under selected network settings in a 200 × 200 m2 area.

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

3 x 10 EMA−SIC Li et al.

1

Number of nodes

(b) α = 5, β = 2

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4

6

Energy consumption

Energy consumption

4

2

3

1

0 100 200 300 400 500 600 700 800 900 1000

19

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) α = 4, β = 2

0 100 200 300 400 500 600 700 800 900 1000

6

14

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

(a) α = 4, β = 2 6 x 10 EMA−SIC Li et al.

7.2 Latency-energy tradeoff Next, we adopt the metric for latency-energy tradeoff as “the product of energy consumption and the (α − 1)th power of aggregation latency” to examine the performance of these algorithms in Fig. 6 for selected settings. The metric proposed in Theorem 3 equals to this revised metric divided by “the product of the optimal energy consumption and the (α − 1)th power of the optimal aggregation latency”. Since the optimums of both aggregation latency and energy consumption of a given network should be the same for any aggregation algorithm, while the optimal aggregation latency of any given network is hard to find under the physical interference model (NP-hard), we adopt the revised metric here, which is equivalent to the previous metric multiplied by a constant factor.

8 x 10 EMA−SIC Li et al. 6

20

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) α = 5, β = 6

Fig. 5: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under selected network settings in a 200 × 200 m2 area.

Fig. 4 also shows that the energy consumption of CellAS may go down when the number of nodes reaches 700– 1000, which can be explained by the decrease of link lengths

(b) α = 5, β = 2 times energy

EMA−SIC Cell−AS Li et al.

α−1

800

Aggregation latency

Aggregation latency

1000

(b) α = 5, β = 2

Latency

Number of nodes

times energy

200

0 100 200 300 400 500 600 700 800 900 1000

α−1

400

Latency

600

times energy

200

EMA−SIC Cell−AS Li et al.

α−1

400

800

times energy

600

1000

α−1

EMA−SIC Cell−AS Li et al.

Latency

800

Aggregation latency

Aggregation latency

1000

between nodes (thus decreased transmission power per link) when the node density in the same square area increases. As stated previously, it is hard to tell the differences between the energy consumption curves of EMA-SIC and Li et al.’s algorithm in Fig. 4. Thus, we conduct a separate comparison of energy usage just between EMA-SIC and Li et al.’s algorithm, and show that EMA-SIC is superior to Li et al.’s algorithm in energy consumption in Fig. 5. The curves of Li et al.’s algorithm are straight lines in Fig. 5 as a result of its constant power assignment in [3]. Another observation with Fig. 3–5 is that: (1) the aggregation latency of each algorithm is lower in settings of larger α (which means more path loss of power, and thus lower interference from other nodes) and smaller β (corresponding to lower SINR requirement); (2) the energy consumption of each algorithm increases with the enhanced value of α (requiring higher transmission power to counteract the increased power loss and meet the SINR requirement) and β (higher SINR requirement).

Latency

all cases, while consuming similar levels of energy as Li et al.’s algorithm (their curves largely overlap in Fig. 4; they will be compared separately in Fig. 5), which are far lower than those of the Cell-AS algorithm. The high energy consumption of Cell-AS algorithm results from its assumption of unlimited transmission power in [11].

20

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) α = 5, β = 6

Fig. 6: Latency-energy tradeoff comparison under selected network settings in a 200 × 200 m2 area. It can be observed that, the Cell-AS algorithm has a significantly poorer latency-power tradeoff, compared with that of the other two algorithms. As the performance differences for Li et al.’s algorithm and the EMA-SIC algorithm are not distinguishable in Fig. 6, we present a separate comparison of latency-energy tradeoff between these two algorithms in Fig. 7. We can have that EMA-SIC algorithm achieves an evidently better latency-energy tradeoff, further confirming its superiority to other aggregation algorithms.

(d) α = 5, β = 6

Aggregation latency

Aggregation latency

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

(c) α = 5, β = 4

180

200

(b) α = 5, β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) α = 4, β = 2

200

1000

800

600

400 100

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

180

Energy consumption 180

2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

200

(d) α = 5, β = 6

Fig. 8: Impact of network scale on the aggregation latency (time units) under selected network settings with uniform distribution and 1000 nodes. With Fig. 8, we see that: i) Li et al.’s algorithm becomes effective when network scale reaches 180 × 180 square meters (detailed explanation in [11]); ii) Cell-AS algorithm has a relatively stable latency performance when network scale varies, which is due to its assumption of unlimited transmission power such that the network diameter and maximum node degree are fixed as 1 and n, respectively; and iii) the latency with EMA-SIC keeps decreasing when the size of network scales up (larger network diameter and smaller node degree), which can be understood that

2 1

120

140

160

Network scale

180

200

(b) α = 5, β = 2, 3 algorithms

4

3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

(c) α = 4, β = 2, 2 algorithms

3 x 10

6

2 EMA−SIC Li et al.

1

0 100

120

140

160

Network scale

180

200

(d) α = 5, β = 2, 2 algorithms

Fig. 9: Impact of network scale on the energy consumption (joule) under selected network settings with uniform distribution and 1000 nodes. 5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

times energy

14

times energy

α−1

3

1

0 100

120

140

160

Network scale

180

200

(a) α = 4, β = 2, 3 algorithms 13

4 x 10 EMA−SIC 3 Li et al. 2 1

0 100

120

19

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

140

160

Network scale

180

200

(c) α = 4, β = 2, 2 algorithms

120

140

160

Network scale

180

200

(b) α = 5, β = 2, 3 algorithms times energy

Fig. 7: A separate comparison of latency-energy between EMA-SIC and Li et al.’s algorithm under selected network settings in a 200 × 200 m2 area. 7.3 Impact of network scale We also examine the impact of network scale on the latency-energy efficiency of the three algorithms in various network settings. Note that, since the network topologies are generated randomly, it is not straightforward to adjust the network diameter or node degree for performance comparison. Thus, we indirectly change the network diameter and node degree by varying the network scale. With the same number of nodes in the network, increasing the network scale will result in enhanced average network diameter while decreased average node degree. We present results under selective settings, where Li et al.’s algorithm exerts its effectiveness, in Fig. 8–10 with uniformly distributed 1000 nodes. Detailed comparisons under other settings are included in Appendix L. 1000

4 x 10

0 100

Number of nodes

(c) α = 5, β = 4

160

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

140

Network scale

α−1

0 100 200 300 400 500 600 700 800 900 1000

4

Latency

1

18

6 x 10 EMA−SIC Li et al.

Energy consumption

times energy α−1

2

120

(a) α = 4, β = 2, 3 algorithms

times energy

18

Latency

times energy α−1

Latency

(b) α = 5, β = 2

4 x 10 EMA−SIC Li et al. 3

0 100

Number of nodes

(a) α = 4, β = 2

0.5

α−1

Number of nodes

1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latency

0.5

α−1

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

times energy

1

6

1

α−1

2

18

2 x 10 EMA−SIC Li et al. 1.5

Latency

13

3 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

9

3 x 10

2

18

EMA−SIC Li et al.

1

0 100

120

140

160

Network scale

180

200

(d) α = 5, β = 2, 2 algorithms

Fig. 10: Impact of network scale on the latency-energy tradeoff under selected network settings with uniform distribution and 1000 nodes. larger network scale leads to smaller node density and thus lower mutual interference among node pairs while more collision-free scheduling opportunities. Network diameter is the dominant factor for worst-case study to find theoretical bounds for aggregation latency as in Sec. 4 and 6. But, the impact of node degree overwhelms the network diameter on the average latency performance in random cases. Although Cell-AS performs better, in aggregation latency, than EMA-SIC does when the network scale is smaller than 160 × 160 square meters, we argue that the impractical assumption of unlimited power by Cell-AS algorithm will render it unapplicable in power-constraint cases (wireless sensor networks typically belong to these cases). In contrast, EMA-SIC fits well in all network settings, with any given maximum transmission power. Besides, EMA-SIC strictly outperforms Cell-AS in energy consumption and latencyenergy tradeoff as shown in Fig. 9 and 10. In Fig. 9 and 10, we present the simulation results for (α, β) ∈ {(4, 2), (5, 2)} while that for (α, β) ∈ {(5, 4), (5, 6)} have similar conclusions and are included in Appendix K due to space constraint. With Fig. 9.a and 9.b, we have that the Cell-AS algorithm consumes much higher energy, proportional to the network scale, than the others, as a consequence of its unlimited power assumption. In Fig. 9.c and 9.d, we conduct a separate comparison

10

on energy consumption between EMA-SIC and Li et al.’s algorithm. We can see that: i) Li et al.’s algorithm has the same energy consumption in various network scales as a result of its constant power assignment; and ii) EMASIC consumes significantly lower energy, also proportional to the network scale (which is a natural result because of the increased distances among node pairs in networks with larger scale). Fig. 10.a and 10.b present the latency-energy tradeoff by the three algorithms. Cell-AS algorithm has the poorest tradeoff performance, which scales up with the network side length, while Fig. 10.c and 10.d demonstrate that EMA-SIC algorithm also remarkably outperforms Li et al.’s algorithm in latency-energy tradeoff.

8 C ONCLUDING R EMARKS This paper investigates the latency-energy tradeoff of data aggregation in wireless sensor networks under the physical interference model and using the successive interference cancellation (SIC) technique. We derive the theoretical lower bounds on both aggregation latency and energy consumption as well as their tradeoff, and give an energy-efficient minimum-latency data aggregation algorithm (EMA-SIC) which can achieve the asymptotically optimal aggregation latency and latency-energy tradeoff. We show that the EMA-SIC algorithm has a constant approximation ratio of aggregation latency (with respect to the theoretical lower bound) while consuming the lowest possible amount of energy. We conduct simulation studies to further validate the superiority of EMA-SIC in terms of latency-energy performance over other work under the physical interference model. As our ongoing work, we plan to evaluate the latency-energy tradeoff in wireless sensor networks with dynamics, e.g. stochastic node sleep and wake up events, and heterogenous node capabilities, e.g. differentiated battery power. R EFERENCES [1] [2] [3]

[4] [5] [6] [7] [8] [9]

S. C. H. Huang, P. Wan, C. T. Vu, Y. Li, and F. Yao, “Nearly constant approximation for data aggregation scheduling in wireless sensor networks,” in Proc. of IEEE INFOCOM’07, 2007. P. J. Wan, S. C. H. Huang, L. X. Wang, Z. Y. Wan, and X. H. Jia, “Minimum-latency aggregation scheduling in multihop wireless networks,” in Proc. of ACM MOBIHOC’09, 2009. X. Y. Li, X. H. Xu, S. G. Wang, S. J. Tang, G. J. Dai, J. Z. Zhao, and Y. Qi, “Efficient data aggregation in multi-hop wireless sensor networks under physical interference model,” in Proc. of IEEE MASS’09, 2009. Y. Yu, B. Krishnamachari, and V. K. Prasanna, “Energy-latency tradeoffs for data gathering in wireless sensor networks,” in Proc. of IEEE INFOCOM’04, 2004. M. Arumugam and S. S. Kulkarni, “Tradeoff between energy and latency for convergecast,” in Proc. of IEEE INSS’05, 2005. H. M. Ammari and S. K. Das, “Trade-off between energy savings and source-to-sink delay in data dissemination for wireless sensor networks,” in Proc. of ACM MSWiM’05, 2005. X. Y. Li, Y. Wang, and Y. Wang, “Complexity of data collection, aggregation, and selection for wireless sensor networks,” IEEE Transactions on Computers, vol. 60, pp. 386 – 399, 2011. G. Brar, D. Blough, and P. Santi, “Computationally efficient scheduling with the physical interference model for throughput improvement in wireless mesh networks,” in Proc. of ACM MOBICOM’06, 2006. D. Chafekar, V. S. Kumar, M. V. Marathe, S. Parthasarathy, and A. Srinivasan, “Approximation algorithms for computing capacity of wireless networks with sinr constraints,” in Proc. of IEEE INFOCOM’08, 2008.

[10] T. Moscibroda, “The worst-case capacity of wireless sensor networks,” in Proc. of ACM/IEEE IPSN’07, 2007. [11] H. Li, Q.-S. Hua, C. Wu, and F. C. M. Lau, “Minimum-latency aggregation scheduling in wireless sensor networks under physical interference model,” in Proc. of ACM MSWiM’10, 2010. [12] J. Andrews, “Interference cancellation for cellular systems: A contemporary overview,” IEEE Wireless Communications, vol. 12, pp. 19–29, 2005. [13] D. Halperin, T. E. Anderson, and D. Wetherall, “Taking the sting out of carrier sense: Interference cancellation for wireless lans,” in Proc. of ACM MOBICOM’08, 2008. [14] IEEE Std. 802.15.4-2006: Wireless medium access control and physical layer specifications for low-rate wireless personal area networks, IEEE Std., 2003. [Online]. Available: http://www.ieee802.org [15] S. P. Weber, J. G. Andrews, X. Y. Yang, and G. de Veciana, “Transmission capacity of wireless ad hoc networks with successive interference cancellation,” IEEE Transactions on Information Theory, vol. 53, pp. 2799 – 2814, 2007. [16] Y. Li, L. Guo, and S. K. Prasad, “An energy-efficient distributed algorithm for minimum-latency aggregation scheduling in wireless sensor networks,” in Proc. of IEEE ICDCS’10, 2010. [17] X. Xu, X.-Y. Li, X. Mao, S. Tang, and S. Wang, “A delay-efficient algorithm for data aggregation in multihop wireless sensor networks,” IEEE Transactions on Parallel and Distributed Systems, vol. 22, pp. 163–175, 2011. [18] B. Yu, J. Li, and Y. Li, “Distributed data aggregation scheduling in wireless sensor networks,” in Proc. of IEEE INFOCOM’09, 2009. [19] B. Yu and J. Li, “Minimum-time aggregation scheduling in multi-sink sensor networks,” in Proc. of IEEE SECON’11, 2011. [20] P. Balister, B. Bollobas, A. Anandkumar, and A. Willsky, “Energylatency tradeoff for in-network function computation in random networks,” in Proc. of IEEE INFOCOM’11, 2011. [21] O. Simeone, O. Somekh, Y. Bar-Ness, H. V. Poor, and S. Shamai, “Capacity of linear two-hop mesh networks with rate splitting decodeand-forward relaying and cooperation,” in Proc. of IEEE Allerton’07, 2007. [22] X. Wang and J. Garcia-Luna-Aceves, “Embracing interference in ad hoc networks using joint routing and scheduling with multiple packet reception,” in Proc. of IEEE INFOCOM’08, 2008. [23] S. Lv, W. Zhuang, X. Wang, and X. Zhou, “Scheduling in wireless ad hoc networks with successive interference cancellation,” in Proc. of IEEE INFOCOM’11, 2011. [24] C. Jiang, Y. Shi, Y. T. Hou, W. Lou, S. Kompella, and S. F. Midkiff, “Squeezing the most out of interference: An optimization framework for joint interference exploitation and avoidance,” in Proc. of IEEE INFOCOM’12, 2012. [25] O. Goussevskaia and R. Wattenhofer, “Scheduling wireless links with successive interference cancellation,” in Proc. of IEEE ICCCN’12, 2012. [26] P. J. Wan, K. M. Alzoubi, and O. Frieder, “Distributed construction of connected dominating set in wireless ad hoc networks,” in Proc. of IEEE INFOCOM’02, 2002. [27] L. Fu, S. C. Liew, and J. Huang, “Effective carrier sensing in csma networks under cumulative interference,” in Proc. of IEEE INFOCOM’10, 2010. [28] Sinaglo: Simulator for network algorithms. [Online]. Available: http://www.disco.ethz.ch/projects/sinalgo/ [29] E. F. Beckenbach and R. Bellman, Inequalities. New York: SpringerVerlag, 1965. ¨ [30] H. Groemer, “Uber die einlagerung von kreisen in einen konvexen bereich,” Math. Z., vol. 73, pp. 285–294, 1960. [31] C. Ambuhl, “An optimal bound for the mst algorithm to compute energy efficient broadcast trees in wireless networks,” in Proc. of ICALP’05, 2005.

11

Hongxing Li received his B.Sc. and M.Eng. degrees in 2005 and 2008 from the Department of Computer Science and Technology, Nanjing University, China. He is currently a Ph.D. candidate in the Department of Computer Science, the University of Hong Kong, Hong Kong. His research interests include wireless networks and cloud computing.

Chuan Wu received her B.Eng. and M.Eng. degrees in 2000 and 2002 from Department of Computer Science and Technology, Tsinghua University, China, and her Ph.D. degree in 2008 from the Department of Electrical and Computer Engineering, University of Toronto, Canada. She is currently an assistant professor in the Department of Computer Science, the University of Hong Kong, Hong Kong. Her research interests include measurement, modeling, and optimization of large-scale peer-to-peer systems and online/mobile social networks. She is a member of IEEE and ACM.

Dongxiao Yu received his B.Sc degree at the School of Mathematics, Shandong University. Currently, he is a PhD candidate in the Department of Computer Science, the University of Hong Kong. His research interests include wireless networks, distributed computing and graph theory.

Qiang-Sheng Hua received his B.Eng. and M.Eng. degrees in 2001 and 2004 from the School of Information Science and Engineering, Central South University, China, and his Ph.D. degree in 2009 from the Department of Computer Science, The University of Hong Kong, China. He is currently an assistant professor in the Institute for Interdisciplinary Information Sciences, Tsinghua University, China. His research interests are on wireless networking and algorithms.

Francis C.M. Lau received his PhD in computer science from the University of Waterloo. He is currently a professor in computer science in The University of Hong Kong. His research interests include computer systems, networks, programming languages, and application of computing in arts.

12

A PPENDIX A P ROOF TO T HEOREM 1 We prove the theorem based on a lemma. Lemma 1. Consider the n positive real numbers, x1 , . . . , xn , we have that, n X i=1

xα i

≥

(

Pn

i=1 xi ) nα−1

α

, ∀α ≥ 1.

Proof: Consider the generalized mean of n positive real numbers, x1 , . . . , xn , with exponent p (where p is real number): Mp (x1 , . . . , xn ) = ( n1 · Pna non-zero p 1/p . We use the generalized mean inequality [29], i=1 xi ) Mp (x1 , . . . , xn ) ≤ Mq (x1 , . . . , xn ) if p < q, where the two means are equal if and only if x1 = x2 = · · · = xn . Since α ≥ 1, we have M1 (x1 , . . . , xn ) ≤ Mα (x1 , . . . , xn ) Pn n xi 1 X α 1/α ⇒ i=1 ≤( xi ) n n i=1 P n α X ( n i=1 xi ) ⇒ xα . ⊔ ⊓ i ≥ nα−1 i=1

Proof of Theorem 1: The minimum overall energy consumption for data aggregation is achieved when links are scheduled in a “TDMA” fashion, i.e., only one link is scheduled to transmit in each time slot (and there is no interference from concurrent transmissions) and the data aggregation tree is a minimum spanning tree of the network, of which the weight of link ei is dα ii , where dii is the Euclidean length of link ei . Note that, even if the application of SIC is an option, the optimal choice for minimum overall energy consumption is not to use SIC while scheduling the links as above. The rationale is that, extra energy is required to recover the signal that is not the first canceled one, in order to compensate for the cumulative interference from those links canceled before it. Thus, the transmission power Pi , which is also the energy consumption since the transmission delay is normalized to 1 time unit as in Sec. 3, to guarantee the success of transmission ei is such that Pi /dα ii ≥ β ⇒ Pi ≥ N0 βdα ii . N0

The minimum distance between any two nodes in the maximum independent set with PM should be larger than dM , the maximum transmission range between any two nodes. The overall length of edges in any tree connecting all the nodes in the maximum independent set with PM , Lmis , should be larger than nmis dM . The overall link length in the minimum spanning Pn−1 tree in the network is no smaller than Lmis , i.e., i=0 dii ≥ Lmis > nmis dM . By Lemma 1, the overall energy consumption is lower-bounded by

Lemma 2. Suppose a receiver can recover X simultaneous signals in one time slot using SIC. The transmission power used at the transmitter of the ith signal in the interference cancellation sequence is at least N0 β(1 + β)X−i dα ii , where dii is the link length of the ith signal. Proof: We prove this lemma by induction. The base case: Consider the last signal to recover in the interference cancellation sequence, i.e., i = X. The signal is affected by background noise, but not any interference from concurrent transmissions. Its SINR value satisfies PX /dα X−X α XX ≥ β ⇒ PX ≥ N0 βdα dXX . XX = N0 β(1 + β) N0

Therefore, the lower bound holds for the base case. Inductive step: Suppose the transmission power of the j th signal, Pj , j = i + 1, . . . , X, is at least N0 β(1 + β)X−j dα jj . th Then the transmission power of the i signal, Pi , satisfies N0 +

Pi /dα ii PX

j=i+1

⇒Pi ≥ (N0 +

⇒Pi ≥ (N0 +

Pj /dα jj

≥β

X X Pj )βdα ii α d jj j=i+1 X X

N0 β(1 + β)X−j )βdα ii

j=i+1

⇒Pi ≥ N0 β(1 + β

1 − (1 + β)X−i α )dii 1 − (1 + β)

= N0 β(1 + β)X−i dα ii .

⊓ ⊔

The lemma is proved.

Proof of Theorem 2: In a multi-hop wireless sensor network with diameter D, the maximum number of hops for the data from a sensor node to reach the sink is D, when all intermediate nodes transmit using PM . Therefore, at least D time slots are needed. Let T be the minimum number of time slots needed for data aggregation. We have T ≥ D. Suppose at most X simultaneous transmissions can be successfully recovered at a receiver under the physical interference model with SIC, when the maximum transmission power of each sender is PM . Consider the first signal in the interference cancellation sequence. Based on Lemma 2 and recall that the minimum distance between two nodes is 1 in our network model, we have PM ≥ N0 β(1 + β)X−1 1α ⇒ X = ⌊log1+β

PM + 1⌋. N0 β

Therefore, at most X transmissions can be concurrently scheduled among every group of X + 1 nodes. In a network with n nodes, each of which needs to transmit exactly once, X at most n X+1 nodes can be scheduled to transmit in the first n time slot, while X+1 nodes remain. Therefore, in time slot 1 X t (1 ≤ t ≤ T − 1), at most n( X+1 )t−1 X+1 nodes can be P n−1 n−1 1 n−1 t α α X X α scheduled for transmission while n( X+1 ) nodes remain. ( i=0 dii ) (nmis dM ) Pi ≥ N0 β dii ≥ N0 β > N0 β .⊓ ⊔ α−1 α−1 Finally at time slot T , only the sink node remains. We thus n n i=0 i=0 1 have n( X+1 )T ≤ 1, which gives T ≥ logX+1 n. Since the length of each time slot t is 1 + χt τ ≥ 1 A PPENDIX B time units, combining the two cases above, the aggregation P ROOF TO T HEOREM 2 latency is lower-bounded by Ω(max{D, logX+1 n}) time We prove the theorem based on Lemma 2. units. ⊓ ⊔

13

Ă A PPENDIX C P ROOF TO T HEOREM 3 Ă

2dM / n

n-1

2dM /

Ă Ă

2d / ĂĂ Ă m M

h = min{d1 , d2 } ≤ 2dM /

Ă

2dM /

0

(a) Network topology. n

n-1

ĂĂ

Ăm

Ă

0

(b) Minimum spanning tree. n-1

Ă

m

Ă

2

2

(1 + β)3h +3h−1

It is clear that N0 β(1 + β)3h

+3h−1

(2h)α < N0 β(

0

So we have (c) Any aggregation tree.

Fig. 11: A chain network with n + 1 nodes evenly distributed along a line. Proof: To prove the theorem, we only need to show that there exists some network of n + 1 nodes with maximum node degree ∆ such that ρα−1 L ρE is lowered-bounded by Ω(∆α−1 ) under the physical interference model with SIC. We consider the chain network in Fig. 11(a), where n + 1 M nodes are evenly distributed along a line segment [0, 2nd ∆ ], 2dM i.e., node vi is at position ∆ · (n − i), 0 ≤ i ≤ n. So, the network diameter D is 2n ∆ (the hop-count from v0 to vn ), and the maximum node degree is exactly ∆, e.g., the node degree of some intermediate node vm . We further let ∆ ≤ 2n/ log2 n in this network such that 2n ∆ ≥ log2 n ≥ PM logX+1 n, where X = ⌊log1+β N + 1⌋ ≥ 1 as defined 0β in Theorem 2. Consequently, the aggregation latency lower bound is D = 2n ∆ time units for this network. It is easy to see that the minimum energy data aggregation tree is simply the path v0 v1 . . . vn as in Fig. 11(b), with the minimum energy consumption being nN0 β( 2d∆M )α . For the general case where the data aggregation topology is a tree T as in Fig. 11(c) with any given aggregation algorithm, we assume there are k edges along the unique path from v0 to vn in the tree. Then the data aggregation along Pk tree T takes at least k time slots or equivalently lengths t=1 (1+χt τ ) ≥ k time units. Denote the Euclidean Pk of the k edges in T by ri , i = 1, . . . , k. Clearly, i=1 ri ≥ 2ndM along this path is ∆ . Then the energy consumption P α α Pk ( k M) α i=1 ri ) at least ≥ N0 β (2nd i=1 N0 βri ≥ N0 β kα−1 kα−1 ∆α , where the first inequality is based on Lemma 1. Then, for any given aggregation algorithm, we have that N0 β(2ndM )α /(kα−1 ∆α ) k )α−1 2n/∆ nN0 β(2dM /∆)α α−1 = (∆/2) .

A PPENDIX D T HE VALIDITY

⊔ ⊓

OF HEXAGON SIDE LENGTH

With Eqn. (3) in Appendix B, we can have that K1 must be larger than zero. Otherwise, we will have the problem of “division by zero” in Eqn. (3) if K1 = 0, or negative value for cumulative interference I, which is unreasonable, if K1 < 0 and α is an odd number. We next prove that, with our hexagon side length h, K1 is always larger than zero. According to the definition of h, we have that

p PM /(N0 β),

p PM /(N0 β))2 < PM .

N0 (2h)α 1 − > 0 ⇒ K1 > 0. 2 PM β(1 + β)3h +3h−1

In conclusion, the value of h validate the assignment of K1 .

A PPENDIX E A N EXAMPLE

OF AGGREGATION TREE CON -

STRUCTION

Fig. 12 illustrates the tree construction procedure with an example.

A PPENDIX F C ORRECTNESS S TEP S1

OF

L INK S CHEDULING

IN

We make use of the following theorem from [30] and Lemma 6 to complete our proof. Theorem 7. (Groemer Inequality [30]) Suppose that C is a compact convex set and U is a set of points with mutual distances at least one. Then |U ∩ C| ≤

area(C) peri(C) √ + + 1, 2 3/2

where area(C) and peri(C) are the area and the perimeter of C, respectively. Lemma 6. In a hexagon of side length d, there are at most 3d2 + 3d + 1 nodes if the minimum distance between any two nodes is 1. Proof: Since hexagon is a compact convex set, on the basis of Theorem 9, we can have the maximum number of nodes in the hexagon as follows, |U ∩ C| ≤

ρα−1 ρE ≥ ( L

d1 , d2 .

Then we have the following result (

2

n

(

√ 6 · 1/2 · 3/2 · d2 6·d √ + + 1 = 3d2 + 3d + 1.⊓ ⊔ 2 3/2

We have that each pair of concurrent receivers in step ′ 1 S1 are separated by 2(K1 + 1)h and K 1 = (6X ) α (1 + α 1 1 N (2h) 2 1 1 0 ( √3 )α α−2 ) α ( β(1+β)3h2 +3h−1 − PM )− α with X ′ = 3h2 + 3h. Since the maximum number of signals that can be canceled at one receiver at a time is the maximum number of nodes in one hexagon minus 1, we know, according to Lemma 6, that at most 3h2 + 3h = X ′ signals can be canceled. Using the same approach with [27], i.e., assuming maximum power assignment at all concurrent transmitters when each pair of concurrent receivers are separated by a constant distance, the cumulative interference at any receiver with successive interference cancellation can be upperbounded as,

14

(a) Network topology with the dark node as the sink and the dashed circle as its maximum transmission range.

(b) BFS tree after step T1.

(c) CDS after step T2 with all dark nodes as the dominators.

(d) Step T3.a with dark nodes as dominators, (e) Step T3.b with dark nodes as dominators, grey nodes as head dominatees and white nodes grey nodes as head dominatees and white nodes as non-head dominatees. as non-head dominatees.

(f) Aggregation tree after step T3.

Fig. 12: An example of aggregation tree construction.

X ′ (6(

N0 β(2 − 1/K2α )rα . According to the conclusion in [27], the cumulative interference I ′ at any receiver is

∞ 1 α X 2 ) + 6j( √ )α )PM /(2h)α K1 3jK 1 j=2

= 6X ′ ( = 6X ′ (

∞ X 1 α 2 ) (1 + ( √ )α )PM /(2h)α K1 3j j=2

∞ X 1 α 2 1 ) (1 + ( √ )α )PM /(2h)α K1 3 j=2 j α−1

1 α 2 1 ) (1 + ( √ )α )PM /(2h)α K1 3 α−2 PM /(2h)α = − N0 = I. β(1 + β)X ′ −1

(3)

′

The power assignment is Pi = (N0 + I)β(1 + β)X −i dα ii for the ith canceled link with length dii . Then we prove that the link can be correctly scheduled with successive interference cancellation with a valid SINR value, Pi /dα ii N0 + I + =

N0 + I +

′

−i

′ (N0 +I)β((1+β)X −i −1) β

= β.

We can conclude that each link transmission in step S1 is successful under the physical interference model with successive interference cancellation.

A PPENDIX G C ORRECTNESS OF L INK S CHEDULING S TEPS S2 AND S3

We can conclude that each link transmission in steps S2 and S3 is successful under the physical interference model.

A PPENDIX H P ROOF TO T HEOREM 5 The proof of Theorem 5 is based on the following lemma. Lemma 3. A disk with radius d can be covered by at most 2(d+h)(2d−h) + 1 hexagons with side length h. 3h2

′ (N0 +I)β(1+β)X −j dα ji j=i+1 dα ji

(N0 + I)β(1 + β)X

So the SINR value for any scheduled link with length of r should be P/r α N0 β(2 − 1/K2α ) ≥ = β. ′ N0 + I N0 + N0 (1 − 1/K2α )

= 6X ′ (

PX ′

1 α 2 1 PM ) (1 + ( √ )α ) K2 3 α − 2 dα M 1 α 2 1 = 6N0 β( ) (1 + ( √ )α ) K2 3 α−2 = N0 (1 − 1/K2α ).

I ′ ≤ 6(

IN

We know that, in steps S2 and S3 of link scheduling, any two concurrent transmitters are separated by at least (K2 +1)dM , 1 where K2 = (6β(1+( √23 )α α−2 )+1)1/α . For any scheduled link with length r, we have the power assignment P =

Proof: As illustrated by Fig. 13, we can divide the disk into 6 equal-sized non-overlapping cones, labeled A to F . It is obvious that the maximum number of hexagons to cover the disk is at most 6 times the number to cover each cone. In each cone, there are at most 61 of a hexagon in the range of 12 h to the center of the disk, 16 + 1 hexagons in the range of 2h, 16 + 1 + 2 hexagons in the range of 72 h, and soP on. We can prove by induction that there are at most 1/6 + ji=0 i hexagons in the range of 1+3j 2 h in each cone. Therefore, in a range of d, corresponding to j = ⌊ 2d−h 3h ⌋, 2d−h

( 2d−h +1)

there are at most 1/6+ 3h 23h hexagons in one cone. We can then derive that there are at most 2(d+h)(2d−h) +1 3h2 hexagons in the entire disk with radius d. ⊓ ⊔

15

A PPENDIX I P ROOF TO T HEOREM 6 We first present two lemmas and then the proof of Theorem 6. Lemma 4. Let Vi be the set of nodes in the disk of radius dM centered at a dominator vi , including the dominator. Let e1 , e2 , . . . , e|Vi |−1 be the links in the minimum spanning tree of Vi rooted at the dominator, derived using link weights proportional to rα (α ≥ 2) for a link with length r. Then |Vi |−1

Fig. 13: An illustration of the proof of lemma 3. Proof of Theorem 5: In step S1 of the link scheduling, each pair of concurrent receivers (head dominatees) are separated by at least a distance of 2(K1 + 1)h. So at least one head dominatee, in a disk of radius 2(K1 + 1)h, can successfully receive all the transmissions from the other dominatees in its hexagon in each time slot. In addition, there are at 2 most 16 3 K1 + 12K1 + 7 head dominatees in a disk of radius 2(K1 + 1)h according to Lemma 3. Therefore, the 2 aggregation latency in this step is at most 16 3 K1 + 12K1 + 7 16 2 time slots with at most (1 + Xτ )( 3 K1 + 12K1 + 7) time units, which is a constant value. When the first step finishes, there are at most 2(dM +h)(2dM −h) + 1 remaining nodes in a disk of radius 3h2 dM , which have not transmitted. In step S2 of the link scheduling in EMA-SIC, transmissions from head dominatees to their dominators along the respective minimum spanning trees are scheduled. Since each pair of concurrent transmitters are separated by a distance of at least (K2 + 1)dM , one head dominatee can be scheduled for transmission in a disk of radius (K2 + 1)dM M −h) + 1 head in each time slot. As at most 2(dM +h)(2d 3h2 dominatees reside in a disk of radius dM , there are at M −h) most (K2 + 1)2 ( 2(dM +h)(2d + 1) head dominatees to 3h2 be scheduled in a disk of radius (K2 + 1)dM . Therefore, the aggregation latency in this step is at most (K2 + M −h) 1)2 ( 2(dM +h)(2d + 1) time units (SIC is not applied 3h2 in Step S2), which is also a constant. In step S3 of the link scheduling, transmissions from the dominators are scheduled along the tree connecting them to the sink. Since each dominator has a bounded degree in the connected dominant set [26], which is denoted by c here, and at least one dominator can transmit in a disk of radius (K2 + 1)dM in one time slot, each dominator needs to wait at most (K2 + 1)2 c time slots before it can transmit. In addition, the depth of the tree rooted at the sink spanning the connected dominating set is O(D) [26]. Therefore, the aggregation latency in this step is at most O((K2 + 1)2 cD) = O(D) time units (SIC is not utilized in Step S3). In summary, the overall aggregation latency of EMASIC is upper-bounded by O(D) time units. The aggregation latency approximation ratio, i.e., the ratio of this upper bound to the lower bound in Theorem 2, is upper-bounded by O(D/ max{D, logX+1 n}) = O(1). ⊔ ⊓

X

k=1

|ek |α ≤ 6dα M.

Proof: Theorem 3 in [31] gives the following bound for the sum of squares of the link lengths in the minimum spanning tree which is constructed among the nodes in a unit disk centered at a dominator: |Vi |−1

X

k=1

|ek |2 ≤ 6.

If we normalize the length of each link, |ek |, to become |ek | dM , our case will be equivalent to the case in [31]. Therefore, we have |Vi |−1

X |ek | 2 ( ) ≤ 6. dM k=1

As α ≥ 2 and |ek | ≤ dM , we derive |Vi |−1

|Vi |−1 i |−1 X |ek | α |VX X |ek | 2 ( ) ≤ ( ) ≤6⇒ |ek |α ≤ 6dα ⊔ M .⊓ d d M M k=1 k=1 k=1

Lemma 5. Suppose the minimum node degree of the nodes in the connected dominating set is δ and the sizes of the connected dominating set and maximum independent set with PM are ncds and nmis , respectively. We have ncds ≤ (1 +

1 )nmis . δ−1

Proof: According to the connected dominating set construction algorithm in [26], the dominators can be divided into two disjoint sets: VI and VC , where VI is a set of independent nodes including the sink, VC is the set of nodes that are parents of the nodes in VI (expect the sink) in the breadth-first spanning tree, and |VC | ≤ |VI | − 1 [26]. We call the nodes in VI independent nodes and those in VC connector nodes. We next give a tighter relation between |VC | and |VI |, which leads to this lemma. Since the minimum node degree is δ in the connected dominating set, we know that one connector node at level l + 1 can be the parent node of at least δ − 1 independent nodes in level l + 2 and the child node of one independent node in level l. Suppose the number of the independent nodes in level l + 2 is |VI (l + 2)|, then the number of the 1 |VI (l + connector nodes in level l + 1 is |VC (l + 1)| ≤ δ−1 2)|. Suppose the total number of levels in the connected dominating set is L. For completeness, we have |VI (0)| = 1 and |VI (1)| = |VC (0)| = |VC (L − 1)| = 0, since the sink, which is in VI , is the only node at level 0 and there is no independent node in level 1 or connector node in level 0 or L − 1. Then, we derive

16

(a) Uniform

(b) Poisson

(c) Cluster

Fig. 14: Sample topologies of 100 nodes with different distributions. connected dominating set is ncds . Let Vi be the set of head dominatee nodes of any given dominator vi , and L−1 L−2 L−1 X X [ 1 |VI (l + 1)| |VC | = | VC (l)| = |VC (l)| ≤ e1 , e2 , . . . , e|Vi | be the links in the minimum spanning tree δ−1 l=2 l=0 l=1 connecting Vi to vi . Since the energy consumption on each L−1 link ek is N0 β(2 − 1/K2α)|ek |α , we derive the upper bound 1 X = ( |VI (l)| − 1) on overall energy consumption to transmit data from all the δ − 1 l=0 head dominatees to the dominator as follows (the inequality 1 (|VI | − 1). = is based on Lemma 4) δ−1 |Vi | |Vi | Since ncds ≤ |VI | + |VC | and |VI | ≤ nmis , we have X X α α α α

ncds ≤ |VI | + |VC | ≤ (1 +

1 1 )|VI | ≤ (1 + )nmis .⊓ ⊔ δ−1 δ−1

Proof of Theorem 6: In step S1 of the link scheduling with EMA-SIC, the length of the ith link in the interference cancellation sequence at each receiver (head dominator) satisfies dii ≤ 2h, and the transmission power assigned to ′ the ith link to cancel satisfies (N0 + I)β(1 + β)X −i dα ii ≤ ′ (N0 +I)β(1+β)X −i (2h)α ≤ PM . The overall energy consumption to schedule the concurrent transmissions within one hexagon is ′

X X

(N0 + I)β(1 + β)X

′

−i α dii

i=1

≤(N0 + I)β(2h)

′

α

X X

(1 + β)X

′

−i

i=1

′ 1+β (N0 + I)β(1 + β)X −1 (2h)α − (N0 + I)(2h)α β 1+β PM . < β

=

Therefore, the overall energy consumption to aggregate data within one hexagon, from all the non-head dominatees to the head dominatees, is upper-bounded by 1+β β PM . Consider a maximum independent set with PM of the given network with size nmis . Since a maximum independent set with PM is also a dominating set of the network, the disks with radius dM centered at each node in the maximum independent set with PM can cover the network. From Lemma 3, we know that there are at most M −h) C = 2(dM +h)(2d + 1 hexagons of side length h in 3h2 each disk with radius dM . Therefore, there are at most Cnmis hexagons in the entire network. The overall energy consumption for transmissions in all the hexagons in step S1 is then at most Cnmis 1+β β PM . We next analyze the energy consumption in step S2 and step S3 of the link scheduling in EMA-SIC. Suppose the size of the maximum independent set with PM in the network is nmis and the size of the constructed

k=1

N0 β(2 − 1/K2 )|ek | = N0 β(2 − 1/K2 ) ≤ 6N0 β(2 −

|ek |

k=1 1/K2α )dα M.

1 )nmis Lemma 5 shows that there are at most (1 + δ−1 dominators in the connected dominating set. So energy con1 sumption of at most (1+ δ−1 )nmis ×6N0 β(2−1/K2α)dα M is needed to aggregate data from all dominatees in the network to the connected dominating set, and energy consumption 1 of at most (1 + δ−1 )nmis × N0 β(2 − 1/K2α )dα M is needed to aggregate data from all the dominators to the sink. Therefore, the overall energy consumption with EMA-SIC in steps S2 and S3 of the link scheduling is at most 1 7N0 β(1 + δ−1 )nmis (2 − 1/K2α )dα M. Dividing the sum of the upper bounds on energy consumption in all three steps by the lower bound in Theorem 1, we have an energy consumption approximation ratio of at most

Cnmis 1+β PM + 7N0 β(1 + β

1 )nmis (2 δ−1 (nmis dM )α N0 β nα−1

=



− 1/K2α )dα M

 C(1 + β) 1 n α−1 + 7(1 + )(2 − 1/K2α ) ( ) , β δ−1 nmis

where N0 βdα M = PM is used. Since ∆ ≥ n/nmis while C, β, δ, K2 and α are constants, we further derive that the energy consumption approximation ratio is upper-bounded by 

 C(1 + β) 1 + 7(1 + )(2 − 1/K2α ) ∆α−1 = O(∆α−1 ).⊓ ⊔ β δ−1

A PPENDIX J Fig. 14 in Appendix J gives an illustration of network topologies with 100 nodes under different distributions.

A PPENDIX K A DDITIONAL SIMULATION

RESULTS

This section includes additional simulation results, under cases where Li et al.’s algorithm exerts its effectiveness.

17

Number of nodes

Number of nodes

(a) α = 4, β = 2

(b) α = 5, β = 2

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

9

Energy consumption

Energy consumption

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4

(d) α = 5, β = 6

Fig. 16: Energy consumption (joule) comparison under selected network settings in a 180 × 180 m2 area. 4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6

Energy consumption

Energy consumption

4

3 x 10 EMA−SIC Li et al.

(b) α = 5, β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4

Energy consumption

Energy consumption

2

6

8 x 10 EMA−SIC Li et al. 6

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) α = 5, β = 6

Fig. 17: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under selected network settings in a 180 × 180 m2 area.

times energy

times energy α−1

times energy

α−1

Latency

times energy α−1

Latency

times energy

α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) α = 4, β = 2

(b) α = 5, β = 2

18

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

18

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) α = 5, β = 4

(d) α = 5, β = 6

Fig. 19: A separate comparison of latency-energy between EMA-SIC and Li et al.’s algorithm under selected network settings in a 180 × 180 m2 area. 8

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

180

200

6 x 10

4

2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

180

200

(b) α = 5, β = 6, 3 algorithms

6

(c) α = 5, β = 4, 2 algorithms

4

1 0.5

Number of nodes

1

(a) α = 4, β = 2

4

0 100 200 300 400 500 600 700 800 900 1000

(a) α = 5, β = 4, 3 algorithms

Number of nodes

6

6 x 10 EMA−SIC Li et al.

1

2

0 100 200 300 400 500 600 700 800 900 1000

1.5

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

18

2.5 x 10 EMA−SIC 2 Li et al.

α−1

1

13

4 x 10 EMA−SIC Li et al. 3

Latency

Energy consumption

Energy consumption

8

3 x 10 EMA−SIC Cell−AS Li et al. 2

Fig. 18: Latency-energy tradeoff comparison under selected network settings in a 180 × 180 m2 area.

Energy consumption

Fig. 15: Aggregation latency (time units) comparison under selected network settings in a 180 × 180 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) α = 5, β = 6

Energy consumption

Number of nodes

(d) α = 5, β = 6

times energy

0 100 200 300 400 500 600 700 800 900 1000

(c) α = 5, β = 4

0.5

1

(c) α = 5, β = 4

200

6

20

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

400

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

600

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

5

α−1

200

EMA−SIC Cell−AS Li et al.

Latency

400

800

Number of nodes

19

α−1

Number of nodes

(b) α = 5, β = 2 Aggregation latency

Aggregation latency

600

1

(b) α = 5, β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

0 100 200 300 400 500 600 700 800 900 1000

(a) α = 4, β = 2 800

2

(a) α = 4, β = 2

200

1000

19

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

400

Number of nodes

EMA−SIC Cell−AS Li et al.

1

600

0 100 200 300 400 500 600 700 800 900 1000

1000

3

0 100 200 300 400 500 600 700 800 900 1000

times energy

200

EMA−SIC Cell−AS Li et al.

α−1

400

800

Latency

600

1000

Energy consumption

EMA−SIC Cell−AS Li et al.

2

Energy consumption

800

Aggregation latency

Aggregation latency

1000

14

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

Complete simulation results, with various distribution patterns, network scales, node numbers and (α, β) combinations, have 1080 figures and are in Appendix L.

8 x 10

6

6 4 2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) α = 5, β = 6, 2 algorithms

Fig. 20: Impact of network scale on the energy consumption (joule) under selected network settings with uniform distribution and 1000 nodes.

0 100

120

140

160

Network scale

180

200

(c) α = 5, β = 4, 2 algorithms

4 2

0 100

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

120

140

160

Network scale

180

200

800

600 400 200

200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800

200

Number of nodes

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(d) β = 4 1000

Aggregation latency

Aggregation latency

800

600 400 200

1000

800

800

400 200

Number of nodes

(f) β = 6

Fig. 22: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 100 × 100 m2 area.

EMA−SIC Cell−AS Li et al.

600 400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(f) β = 6

Fig. 23: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 100 × 100 m2 area. 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

0 100 200 300 400 500 600 700 800 900 1000

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

600

200

1000

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

400

EMA−SIC Cell−AS Li et al.

600

1000

400

EMA−SIC Cell−AS Li et al.

(c) β = 3

600

(c) β = 3 EMA−SIC Cell−AS Li et al.

200

Number of nodes

Number of nodes

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

1000

400

Number of nodes

Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

600

(a) β = 1 1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

600

(a) β = 1

(d) α = 5, β = 6, 2 algorithms

Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

This section includes complete simulation results, with various distribution patterns, network scales, node numbers and (α, β) combinations. EMA−SIC Cell−AS Li et al.

800

18

8 x 10 EMA−SIC 6 Li et al.

Fig. 21: Impact of network scale on the latency-energy tradeoff under selected network settings with uniform distribution and 1000 nodes. A PPENDIX L C OMPLETE SIMULATION RESULTS

1000

0 100 200 300 400 500 600 700 800 900 1000

200

Aggregation latency

2

Network scale

180

Aggregation latency

4

160

200

Aggregation latency

EMA−SIC Li et al.

140

400

1000

Aggregation latency

18

times energy

6 x 10

120

EMA−SIC Cell−AS Li et al.

600

(b) α = 5, β = 6, 3 algorithms

α−1

Latency

α−1

times energy

(a) α = 5, β = 4, 3 algorithms

0 100

800

(d) β = 4 1000

Aggregation latency

200

Aggregation latency

Network scale

180

Aggregation latency

160

Aggregation latency

140

2

Aggregation latency

120

1000

Aggregation latency

times energy α−1

0 100

Latency

1

20

6 x 10 EMA−SIC Cell−AS Li et al. 4

Aggregation latency

20

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latency

Latency

α−1

times energy

18

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 24: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 100 × 100 m2 area.

19

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 26: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 120 × 120 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 27: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 120 × 120 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 25: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 120 × 120 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 28: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 140 × 140 m2 area.

20

0 100 200 300 400 500 600 700 800 900 1000

400 200 0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 30: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 140 × 140 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 31: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 160 × 160 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 29: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 140 × 140 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

600

1000

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 32: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 160 × 160 m2 area.

21

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 34: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 180 × 180 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 35: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 180 × 180 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 33: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 160 × 160 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 36: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 180 × 180 m2 area.

22

200 0 100 200 300 400 500 600 700 800 900 1000

600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

(e) β = 5

600 400 200

1000

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

400 200

400 200

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 38: Aggregation latency (time units) comparison under α = 4 and uniform distribution in a 200 × 200 m2 area.

EMA−SIC Cell−AS Li et al.

600 400 200

600 400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 800

EMA−SIC Cell−AS Li et al.

Fig. 39: Aggregation latency (time units) comparison under α = 5 and uniform distribution in a 200 × 200 m2 area.

Number of nodes

1000

800

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al.

600

(c) β = 3 EMA−SIC Cell−AS Li et al.

600

(e) β = 5

200

800

EMA−SIC Cell−AS Li et al.

(f) β = 6

400

1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

200

Number of nodes

600

(a) β = 1 800

400

1000

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

Number of nodes

Fig. 37: Aggregation latency (time units) comparison under α = 3 and uniform distribution in a 200 × 200 m2 area.

800

EMA−SIC Cell−AS Li et al.

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

200

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 1000

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

400

(b) β = 2 Aggregation latency

400

800

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1

Aggregation latency

600

1000

800

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

Number of nodes

Number of nodes

(a) β = 1

600

1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 40: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 100 × 100 m2 area.

23

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 42: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 100 × 100 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 43: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 120 × 120 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 41: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 100 × 100 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 44: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 120 × 120 m2 area.

24

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 46: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 140 × 140 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 47: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 140 × 140 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 45: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 120 × 120 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 48: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 140 × 140 m2 area.

25

0 100 200 300 400 500 600 700 800 900 1000

400 200 0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 50: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 160 × 160 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 51: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 160 × 160 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 49: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 160 × 160 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

600

1000

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 52: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 180 × 180 m2 area.

26

800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

Aggregation latency

Aggregation latency

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(d) β = 4 1000

Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 53: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 180 × 180 m2 area.

27

200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

200

Number of nodes

1000

0 100 200 300 400 500 600 700 800 900 1000

800

200

Number of nodes

(f) β = 6

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

800

400 200

200

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

600 400 200

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 56: Aggregation latency (time units) comparison under α = 4 and poisson distribution in a 200 × 200 m2 area.

800

EMA−SIC Cell−AS Li et al.

600 400 200

800 600 400 200

600 400 200

Aggregation latency

800

800

EMA−SIC Cell−AS Li et al.

600 400 200

600 400 200

Number of nodes

(b) β = 2 1000 800 600 400 200

Number of nodes

(c) β = 3

(f) β = 6 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Fig. 55: Aggregation latency (time units) comparison under α = 3 and poisson distribution in a 200 × 200 m2 area.

800

(a) β = 1

(d) β = 4 EMA−SIC Cell−AS Li et al.

1000

Number of nodes

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

200

(d) β = 4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

Aggregation latency

400

Number of nodes

(c) β = 3

1000

600

200

200

(c) β = 3 800

400

400

1000

Number of nodes

EMA−SIC Cell−AS Li et al.

600

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

400

800

(b) β = 2 Aggregation latency

Aggregation latency

600

1000

0 100 200 300 400 500 600 700 800 900 1000

1000 800

Number of nodes

(b) β = 2

600

(a) β = 1 EMA−SIC Cell−AS Li et al.

200

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

600

1000 400

Fig. 54: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 180 × 180 m2 area. EMA−SIC Cell−AS Li et al.

400

600

(e) β = 5

1000

600

(a) β = 1

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000 400

(d) β = 4 Aggregation latency

Aggregation latency

800

200

800

600

(c) β = 3 1000

400

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

600

1000

Aggregation latency

800

1000

Aggregation latency

Aggregation latency

1000

(b) β = 2

EMA−SIC Cell−AS Li et al.

Aggregation latency

Number of nodes

(a) β = 1

800

Aggregation latency

1000 400

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

(d) β = 4 1000

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 57: Aggregation latency (time units) comparison under α = 5 and poisson distribution in a 200 × 200 m2 area.

28

200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

200 0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

600 400 200

200

Number of nodes

1000

0 100 200 300 400 500 600 700 800 900 1000

800

200

Number of nodes

(f) β = 6

600 400 200

1000

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

EMA−SIC Cell−AS Li et al.

800

EMA−SIC Cell−AS Li et al.

600 400 200

(d) β = 4 1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 60: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 100 × 100 m2 area.

800

EMA−SIC Cell−AS Li et al.

600 400 200

1000 800

400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

400 200

Number of nodes

(b) β = 2 1000 800 600 400 200

Number of nodes

(c) β = 3

Number of nodes

(f) β = 6

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

600

(a) β = 1

0 100 200 300 400 500 600 700 800 900 1000

Fig. 59: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 100 × 100 m2 area.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

600

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

Aggregation latency

600

200

EMA−SIC Cell−AS Li et al.

200

(c) β = 3 800

400

800

400

1000

Number of nodes

EMA−SIC Cell−AS Li et al.

600

1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 Aggregation latency

Aggregation latency

800

Number of nodes

(b) β = 2

600

(a) β = 1 EMA−SIC Cell−AS Li et al.

200

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

1000 400

Fig. 58: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 100 × 100 m2 area.

800

400

600

(e) β = 5

1000

600

(a) β = 1

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000 400

(d) β = 4 Aggregation latency

Aggregation latency

800

800

600

(c) β = 3 1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

400

200

Aggregation latency

600

800

400

Aggregation latency

EMA−SIC Cell−AS Li et al.

600

1000

(d) β = 4 1000

Aggregation latency

800

1000

Aggregation latency

Aggregation latency

1000

(b) β = 2

EMA−SIC Cell−AS Li et al.

Aggregation latency

Number of nodes

(a) β = 1

800

Aggregation latency

1000 400

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 61: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 120 × 120 m2 area.

29

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 63: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 120 × 120 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 64: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 140 × 140 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 62: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 120 × 120 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 65: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 140 × 140 m2 area.

30

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 67: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 160 × 160 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 68: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 160 × 160 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 66: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 140 × 140 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 69: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 160 × 160 m2 area.

31

0 100 200 300 400 500 600 700 800 900 1000

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

600 400 200

1000

Aggregation latency

Aggregation latency

800

0 100 200 300 400 500 600 700 800 900 1000

800

600 400 200

400 200

1000 800 600 400 200

400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Aggregation latency

Aggregation latency

600

600 400 200

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 71: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 180 × 180 m2 area.

400 200

Aggregation latency

Number of nodes

(d) β = 4 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 72: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 180 × 180 m2 area.

1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 1000 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(e) β = 5

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

800

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

1000

200

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 800

400

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

600

(f) β = 6

Fig. 70: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 180 × 180 m2 area.

EMA−SIC Cell−AS Li et al.

(c) β = 3

Number of nodes

(e) β = 5

200

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

800

(d) β = 4 Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

400

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 800

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

EMA−SIC Cell−AS Li et al.

600

(a) β = 1 Aggregation latency

EMA−SIC Cell−AS Li et al.

800

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

800

200

(b) β = 2 Aggregation latency

Aggregation latency

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

600

1000

Aggregation latency

Number of nodes

EMA−SIC Cell−AS Li et al.

Aggregation latency

200

Aggregation latency

400

800

(d) β = 4 1000

Aggregation latency

0 100 200 300 400 500 600 700 800 900 1000

600

1000

Aggregation latency

200

EMA−SIC Cell−AS Li et al.

Aggregation latency

400

800

Aggregation latency

600

1000

Aggregation latency

EMA−SIC Cell−AS Li et al.

Aggregation latency

800

Aggregation latency

Aggregation latency

1000

800

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 73: Aggregation latency (time units) comparison under α = 3 and cluster distribution in a 200 × 200 m2 area.

32

0 100 200 300 400 500 600 700 800 900 1000

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

800

800

EMA−SIC Cell−AS Li et al.

600 400 200

EMA−SIC Cell−AS Li et al.

800

EMA−SIC Cell−AS Li et al.

400

Number of nodes

(f) β = 6

600 400 200

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(a) β = 1 800

EMA−SIC Cell−AS Li et al.

600 400 200

(b) β = 2 1000

Aggregation latency

Aggregation latency

1000

0 100 200 300 400 500 600 700 800 900 1000

800 600 400 200

Number of nodes

(c) β = 3 EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(d) β = 4 1000

Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000

EMA−SIC Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

800

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Aggregation latency

Aggregation latency

EMA−SIC Cell−AS Li et al.

3 2 1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

600 400 200 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 75: Aggregation latency (time units) comparison under α = 5 and cluster distribution in a 200 × 200 m2 area.

(d) β = 4

4

0 100 200 300 400 500 600 700 800 900 1000

200

Fig. 74: Aggregation latency (time units) comparison under α = 4 and cluster distribution in a 200 × 200 m2 area.

800

0.5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

600

(e) β = 5

1000

1

Number of nodes

1000

Number of nodes

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

400

(d) β = 4 Aggregation latency

Aggregation latency

1000

0 100 200 300 400 500 600 700 800 900 1000

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2000

(a) β = 1

600

(c) β = 3

4000

Number of nodes

(b) β = 2 1000

EMA−SIC Cell−AS Li et al.

6000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Aggregation latency

Aggregation latency

800

EMA−SIC Cell−AS Li et al.

500

200

(a) β = 1 1000

1000

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1500

8000

Energy consumption

200

600

EMA−SIC Cell−AS Li et al.

2000

(e) β = 5

5

Energy consumption

400

800

EMA−SIC Cell−AS Li et al.

Energy consumption

600

1000

Energy consumption

800

EMA−SIC Cell−AS Li et al.

Aggregation latency

Aggregation latency

1000

Energy consumption

2500

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 76: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 100 × 100 m2 area.

33

3 2 1

(a) β = 1

1.5 1 0.5

4

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

(d) β = 4 5

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

(f) β = 6

Fig. 77: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 100 × 100 m2 area.

(e) β = 5

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 78: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 100 × 100 m2 area.

Number of nodes

(b) β = 2 6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 7

6

8

4

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

(d) β = 4 Energy consumption

4

2

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

4

6

0 100 200 300 400 500 600 700 800 900 1000

8

6

(a) β = 1 Energy consumption

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3 8 x 10 EMA−SIC Cell−AS 6 Li et al.

0.5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

Energy consumption

2

Energy consumption

4

1

(b) β = 2

7

6

1.5

Number of nodes

8

10 x 10 EMA−SIC 8 Cell−AS Li et al.

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

5

5

Energy consumption

4

Energy consumption

6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(f) β = 6

Fig. 79: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 120 × 120 m2 area.

7

6

10 x 10 EMA−SIC 8 Cell−AS Li et al.

1

Number of nodes

Number of nodes

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0.5

Energy consumption

5

1

Energy consumption

2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

4

6

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

5

Energy consumption

Energy consumption

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Number of nodes

6

Energy consumption

Number of nodes

(b) β = 2 Energy consumption

1

Energy consumption

Energy consumption

Energy consumption

2

(c) β = 3

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5000

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

EMA−SIC Cell−AS Li et al.

10000

Number of nodes

6

0 100 200 300 400 500 600 700 800 900 1000

15000

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

5

5

1000

Number of nodes

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

EMA−SIC Cell−AS Li et al.

2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

3000

Energy consumption

Energy consumption

Energy consumption

5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

5

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 80: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 120 × 120 m2 area.

34

2

(c) β = 3

1

1.5 1 0.5

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

1

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(f) β = 6

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 82: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 140 × 140 m2 area.

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

(b) β = 2

8

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

5

Energy consumption

Energy consumption

1

2

(d) β = 4

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

Number of nodes

Number of nodes

Energy consumption

2

1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

3

6 x 10 EMA−SIC Cell−AS Li et al. 4

(b) β = 2

4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

1.5

0 100 200 300 400 500 600 700 800 900 1000

2

Fig. 83: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 140 × 140 m2 area.

(d) β = 4

9

Energy consumption

2000

3

7

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

Energy consumption

4000

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(e) β = 5

4

EMA−SIC Cell−AS Li et al.

(d) β = 4

Number of nodes

(f) β = 6

Fig. 81: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 120 × 120 m2 area.

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

7

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

2

7

Energy consumption

Energy consumption

Energy consumption

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

4

(c) β = 3

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

6

Number of nodes

(d) β = 4

8

6000

1

Number of nodes

Number of nodes

8000

2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

3

6

Energy consumption

0.5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

1

Energy consumption

1.5

Energy consumption

Energy consumption

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(b) β = 2

6

8

6 x 10 EMA−SIC Cell−AS Li et al. 4

Number of nodes

(a) β = 1

(b) β = 2

8

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

9

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

2

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

4

6 x 10 EMA−SIC Cell−AS Li et al. 4

Energy consumption

6

Energy consumption

1

5

7

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

7

3 x 10 EMA−SIC Cell−AS Li et al. 2

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 84: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 140 × 140 m2 area.

35

(b) β = 2

1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

4 2

7

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

5

Energy consumption

Energy consumption

5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(b) β = 2

6

8 x 10 EMA−SIC Cell−AS 6 Li et al.

4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 85: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 160 × 160 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Number of nodes

(c) β = 3

(d) β = 4

7

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

7

Energy consumption

(c) β = 3

1

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 86: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 160 × 160 m2 area.

8

7

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2

Energy consumption

Number of nodes

2

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

8

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2

9

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

9

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

9

Energy consumption

2

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

4

2

5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

Energy consumption

4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

4

6

Energy consumption

Number of nodes

Number of nodes

(a) β = 1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

1

Energy consumption

2000

5

Energy consumption

4000

2

Energy consumption

6000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

4

8000

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 87: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 160 × 160 m2 area.

36

Number of nodes

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

5

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 88: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 180 × 180 m2 area.

1

(c) β = 3

(d) β = 4

7

Energy consumption

4

2

Number of nodes

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

Energy consumption

8 x 10 EMA−SIC Cell−AS 6 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Number of nodes

5

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 89: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 180 × 180 m2 area.

8

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

1

(b) β = 2 7

6

Energy consumption

2

2

(a) β = 1

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 9

9

Energy consumption

4

Energy consumption

Energy consumption

6

2

Number of nodes

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

0.5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5

Energy consumption

Number of nodes

(a) β = 1

1

Energy consumption

1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

10

Energy consumption

5000

6

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

3

Energy consumption

10000

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

4

15000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 90: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 180 × 180 m2 area.

37

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0.5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 91: Energy consumption (joule) comparison under α = 3 and uniform distribution in a 200 × 200 m2 area.

3 2 1

Number of nodes

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

(d) β = 4

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2

1

5

8

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 92: Energy consumption (joule) comparison under α = 4 and uniform distribution in a 200 × 200 m2 area.

8

Energy consumption

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

9

Energy consumption

6

(b) β = 2 7

7

6

Energy consumption

10 x 10 EMA−SIC 8 Cell−AS Li et al.

2

(a) β = 1 Energy consumption

2

4

Number of nodes

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

9

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4 10

10

Energy consumption

5

Energy consumption

Energy consumption

15 x 10 EMA−SIC Cell−AS Li et al. 10

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

4

Energy consumption

0.5

Energy consumption

(a) β = 1

Energy consumption

Number of nodes

1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

6

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

0.5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

1

Energy consumption

Energy consumption

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 93: Energy consumption (joule) comparison under α = 5 and uniform distribution in a 200 × 200 m2 area.

38

Number of nodes

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

2

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

6

1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 94: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 100 × 100 m2 area.

Energy consumption

5

4

5

Energy consumption

Energy consumption

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

6

6

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

6

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

2

1

(a) β = 1

6 4 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 95: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 100 × 100 m2 area.

7

6

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

5000

2

Number of nodes

Number of nodes

6 4 2

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

7

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

10000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Cell−AS Li et al. 4

6 4 2

8

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

8

Energy consumption

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

4

Energy consumption

(a) β = 1

5

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2000

0 100 200 300 400 500 600 700 800 900 1000

15000

5

4

4000

Energy consumption

500

EMA−SIC Cell−AS Li et al.

6000

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

8

Energy consumption

1000

8000

Energy consumption

EMA−SIC Cell−AS Li et al.

1500

Energy consumption

Energy consumption

2000

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 96: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 100 × 100 m2 area.

39

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

4 2

Number of nodes

(a) β = 1

(b) β = 2 8

8

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(d) β = 4

(c) β = 3

(d) β = 4

5

0.5

1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 97: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 120 × 120 m2 area.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

9

Energy consumption

1

3 x 10 EMA−SIC Cell−AS Li et al. 2

Energy consumption

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

5

Number of nodes

(e) β = 5

(f) β = 6

Fig. 99: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 120 × 120 m2 area.

5

2

4000 3000 2000 1000

(a) β = 1

1.5 1 0.5

Energy consumption

6

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(b) β = 2

2 1

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 98: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 120 × 120 m2 area.

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

7

Energy consumption

5

3

(d) β = 4

6

15 x 10 EMA−SIC Cell−AS Li et al. 10

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Number of nodes

Number of nodes

Number of nodes

4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

(b) β = 2 6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

1 0.5

Number of nodes

Number of nodes

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

Energy consumption

0.5

4

4

5000

Energy consumption

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4 5

5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1.5

Energy consumption

5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

Energy consumption

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

Energy consumption

4

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

Energy consumption

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

6

4

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

0.5

(b) β = 2

4

1

1

Number of nodes

(a) β = 1 3 x 10 EMA−SIC Cell−AS Li et al. 2

1.5

Energy consumption

1000

5000

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

2000

10000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

3000

7

7

EMA−SIC Cell−AS Li et al.

Energy consumption

4000

15000

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

5000

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 100: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 140 × 140 m2 area.

40

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 1

Number of nodes

2 1

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

1

4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

Energy consumption

6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 8 x 10 EMA−SIC Cell−AS 6 Li et al.

(b) β = 2

4 2

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 102: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 140 × 140 m2 area.

7

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

7

Energy consumption

1

Energy consumption

1.5

(f) β = 6

Fig. 103: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 160 × 160 m2 area.

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Number of nodes

(e) β = 5

(d) β = 4

9

2

6

6

0 100 200 300 400 500 600 700 800 900 1000

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

2

1

(b) β = 2

8

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

5

1

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4 5

0 100 200 300 400 500 600 700 800 900 1000

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

7

3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(f) β = 6

Fig. 101: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 140 × 140 m2 area.

Number of nodes

Energy consumption

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

Number of nodes

1 0.5

5

Energy consumption

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

(c) β = 3

Energy consumption

0.5

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

Energy consumption

1

Energy consumption

Energy consumption

1.5

(b) β = 2 5

0 100 200 300 400 500 600 700 800 900 1000

7

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

7

Energy consumption

(a) β = 1 Energy consumption

4

(c) β = 3

1

Number of nodes

Number of nodes

Energy consumption

6

Number of nodes

2

4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

6

Energy consumption

Energy consumption

1

2000

(b) β = 2

6

2

4000

Number of nodes

(a) β = 1 4 x 10 EMA−SIC Cell−AS 3 Li et al.

6000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

7

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

1

5

8000

Energy consumption

2

4

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 104: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 160 × 160 m2 area.

41

8

1

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

4000 2000

2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

3

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 106: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 180 × 180 m2 area.

(b) β = 2 9

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

9

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Number of nodes

8

(d) β = 4

5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

(a) β = 1

Number of nodes

Number of nodes

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

2

1

(f) β = 6

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

4

Energy consumption

Energy consumption

6

Energy consumption

5

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

Number of nodes

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

(b) β = 2

4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

7

Number of nodes

Number of nodes

5

Fig. 107: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 180 × 180 m2 area.

Energy consumption

6000

Energy consumption

Energy consumption

8000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

(e) β = 5

4

EMA−SIC Cell−AS Li et al.

(d) β = 4 7

Number of nodes

(f) β = 6

Fig. 105: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 160 × 160 m2 area.

10000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

7

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

2

(c) β = 3

9

Energy consumption

Energy consumption

2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

9

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

2

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

7

Energy consumption

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

2

1.5

Energy consumption

4

(b) β = 2

6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

1

(a) β = 1

9

Energy consumption

Energy consumption

8

2

Number of nodes

Number of nodes

Energy consumption

(a) β = 1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

9

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al.

5

Energy consumption

2

6

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

5

Energy consumption

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 108: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 180 × 180 m2 area.

42

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

(b) β = 2

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 109: Energy consumption (joule) comparison under α = 3 and poisson distribution in a 200 × 200 m2 area.

2 1

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

7

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

6 4 2

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 110: Energy consumption (joule) comparison under α = 4 and poisson distribution in a 200 × 200 m2 area.

8

Energy consumption

6

5

5

Energy consumption

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 7

6

Energy consumption

1

(a) β = 1 Energy consumption

2

2

Number of nodes

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 9

9

Energy consumption

5

Energy consumption

Energy consumption

15 x 10 EMA−SIC Cell−AS Li et al. 10

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

4

Energy consumption

5

Energy consumption

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

6

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4 10

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

5000

3

Energy consumption

10000

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

4

15000

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 111: Energy consumption (joule) comparison under α = 5 and poisson distribution in a 200 × 200 m2 area.

43

1500

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

Number of nodes

(c) β = 3

(d) β = 4 4

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

Number of nodes

(e) β = 5

(f) β = 6

Fig. 112: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 100 × 100 m2 area.

(b) β = 2 Energy consumption

Energy consumption

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 114: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 100 × 100 m2 area.

1000

EMA−SIC Cell−AS Li et al.

800 600 400 200

4000

EMA−SIC Cell−AS Li et al.

3000 2000 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

6

Energy consumption

Energy consumption

1

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

5

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

2

(c) β = 3

Number of nodes

2

(b) β = 2 4

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

6

Number of nodes

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4 8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2

5

0.5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

1

(c) β = 3

4

Number of nodes

1.5

1

6

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

2

Number of nodes

Number of nodes

Energy consumption

1

10 x 10 EMA−SIC 8 Cell−AS Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5

8000

EMA−SIC Cell−AS Li et al.

6000 4000 2000

1

Energy consumption

2

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

5

7

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

Energy consumption

4

1

(b) β = 2 7

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

Number of nodes

6

0 100 200 300 400 500 600 700 800 900 1000

2

(a) β = 1

10000

5000

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

0.5

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 113: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 100 × 100 m2 area.

(d) β = 4

4

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

4

Energy consumption

2000

1

6

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Cell−AS Li et al.

Energy consumption

4000

15000

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

6000

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

Energy consumption

500

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

6

1000

Energy consumption

200

EMA−SIC Cell−AS Li et al.

2000

Energy consumption

400

2500

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

600

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 115: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 120 × 120 m2 area.

44

0.5

0 100 200 300 400 500 600 700 800 900 1000

1000

500

EMA−SIC Cell−AS Li et al.

4000 3000 2000 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

5000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

Energy consumption

1

Energy consumption

2

1500

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

4

6 x 10 EMA−SIC Cell−AS Li et al. 4

Number of nodes

(a) β = 1

(b) β = 2

(b) β = 2

6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

1

Energy consumption

Energy consumption

2

15000

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

10000

5000

(c) β = 3

1

Number of nodes

Number of nodes

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

EMA−SIC Cell−AS Li et al.

Energy consumption

4

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

(d) β = 4

(d) β = 4

4

0.5

3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

(f) β = 6

Fig. 116: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 120 × 120 m2 area.

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 118: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 140 × 140 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

6

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 117: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 120 × 120 m2 area.

Energy consumption

0.5

4

(d) β = 4 Energy consumption

1

(b) β = 2 6

0 100 200 300 400 500 600 700 800 900 1000

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Number of nodes

8

Number of nodes

5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3

1

(a) β = 1

7

Energy consumption

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

7

3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

5

Energy consumption

5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

6

Energy consumption

1

Energy consumption

2

6

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

15 x 10 EMA−SIC Cell−AS Li et al. 10

4

6

Energy consumption

2

Number of nodes

Number of nodes

Energy consumption

4

4

Energy consumption

1

Energy consumption

Energy consumption

1.5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

6

6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 119: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 140 × 140 m2 area.

45

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

2 1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4 2

1500 1000 500

EMA−SIC Cell−AS Li et al.

6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

Energy consumption

(f) β = 6

7

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(b) β = 2

(a) β = 1

(b) β = 2

4

0.5

2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 121: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 160 × 160 m2 area.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

5

Energy consumption

5

5

(d) β = 4

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

8

7

Energy consumption

1

6 x 10 EMA−SIC Cell−AS Li et al. 4

(d) β = 4

8

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

8

Energy consumption

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

4

Energy consumption

Number of nodes

Fig. 122: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 160 × 160 m2 area.

Number of nodes

(a) β = 1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

Energy consumption

2000

5

7

8000

Energy consumption

Energy consumption

EMA−SIC Cell−AS Li et al.

2

(f) β = 6

Fig. 120: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 140 × 140 m2 area.

2500

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

Number of nodes

Number of nodes

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

6

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

Energy consumption

3

(d) β = 4

6

8

8 x 10 EMA−SIC Cell−AS 6 Li et al.

1

(c) β = 3

(d) β = 4

8

2

Number of nodes

Number of nodes

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5 x 10 EMA−SIC 4 Cell−AS Li et al.

5

Energy consumption

0.5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

1

(b) β = 2 6

5

Energy consumption

2

Energy consumption

Energy consumption

4

(a) β = 1

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2

Number of nodes

Number of nodes

(b) β = 2

7

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

6

0.5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10 x 10 EMA−SIC 8 Cell−AS Li et al.

1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

1

5

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

Energy consumption

4

7

Energy consumption

Energy consumption

6

8 x 10 EMA−SIC Cell−AS 6 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 123: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 160 × 160 m2 area.

46

10000

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

2

Number of nodes

(d) β = 4

1 0.5

Number of nodes

(f) β = 6

Fig. 124: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 180 × 180 m2 area.

Energy consumption

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

1

(d) β = 4 7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 125: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 180 × 180 m2 area.

7

Energy consumption

Number of nodes

2

(c) β = 3 6

1

3

Number of nodes

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

8

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Cell−AS Li et al. 2

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4 9

8

Energy consumption

0.5

Energy consumption

1

(b) β = 2 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

(a) β = 1 Energy consumption

4

(c) β = 3 Energy consumption

Number of nodes

6

0 100 200 300 400 500 600 700 800 900 1000

5

2 0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

Energy consumption

1

6

0 100 200 300 400 500 600 700 800 900 1000

4

Energy consumption

Energy consumption

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

5000

3 x 10 EMA−SIC Cell−AS Li et al. 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1000

EMA−SIC Cell−AS Li et al.

Energy consumption

2000

15000

Energy consumption

EMA−SIC Cell−AS Li et al.

3000

Energy consumption

Energy consumption

4000

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 126: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 180 × 180 m2 area.

47

4

(b) β = 2

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

6

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

5

Energy consumption

Energy consumption

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

(b) β = 2

6

3 x 10 EMA−SIC Cell−AS Li et al. 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 7

7

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 127: Energy consumption (joule) comparison under α = 3 and cluster distribution in a 200 × 200 m2 area.

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5

Energy consumption

Number of nodes

Number of nodes

(a) β = 1

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 128: Energy consumption (joule) comparison under α = 4 and cluster distribution in a 200 × 200 m2 area.

7

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1 0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

8

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

1

0.5

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 9

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

2

1

0 100 200 300 400 500 600 700 800 900 1000

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Number of nodes

Number of nodes

(a) β = 1

6 x 10 EMA−SIC Cell−AS Li et al. 4

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

0.5

Energy consumption

1000

5

Energy consumption

2000

1

Energy consumption

3000

Energy consumption

4000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

5000

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 129: Energy consumption (joule) comparison under α = 5 and cluster distribution in a 200 × 200 m2 area.

48

500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

15 x 10 EMA−SIC Li et al. 10

5

Number of nodes

(e) β = 5

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 130: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 100 × 100 m2 area.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

1

(c) β = 3 3 x 10 EMA−SIC Li et al. 2

1

5

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 131: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 100 × 100 m2 area.

6

Energy consumption

2000

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

3 x 10 EMA−SIC Li et al. 2

1

6

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

6

Energy consumption

4000

2 x 10 EMA−SIC Li et al. 1.5

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 7

7

Energy consumption

6000

Energy consumption

Energy consumption

EMA−SIC Li et al.

(b) β = 2 5

0 100 200 300 400 500 600 700 800 900 1000

4

8000

2

Number of nodes

4

1000

4

(a) β = 1

3000 2000

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

500

0 100 200 300 400 500 600 700 800 900 1000 EMA−SIC Li et al.

Energy consumption

1000

1

Energy consumption

1500

4000

2

4

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

2000

(b) β = 2

4 x 10 EMA−SIC Li et al. 3

Energy consumption

4

1000

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 132: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 100 × 100 m2 area.

49

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1500 1000 500

EMA−SIC Li et al.

4000

2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

4 2

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

1

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 x 10 EMA−SIC Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

400

200

1500

EMA−SIC Li et al.

1000

500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

3000

EMA−SIC Li et al.

2000

1000

8000

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 134: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 120 × 120 m2 area.

EMA−SIC Li et al.

6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

5

Energy consumption

Energy consumption

1

EMA−SIC Li et al.

(d) β = 4

5

2

600

Number of nodes

(c) β = 3

3

1

Fig. 135: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 120 × 120 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

2

5

Energy consumption

Energy consumption

5

0.5

(b) β = 2

4

10

1

3 x 10 EMA−SIC Li et al.

(e) β = 5

Number of nodes

(a) β = 1 15 x 10 EMA−SIC Li et al.

(d) β = 4 7

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

Energy consumption

2

Number of nodes

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

Fig. 133: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 120 × 120 m2 area.

5

7

Energy consumption

1.5

10

(c) β = 3

2.5 x 10 EMA−SIC 2 Li et al.

Energy consumption

5000

15 x 10 EMA−SIC Li et al.

Number of nodes

4

4

Energy consumption

10000

6

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4 Energy consumption

Energy consumption

EMA−SIC Li et al.

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6

4

15000

2

Number of nodes

Energy consumption

2000

6000

4

(a) β = 1

Energy consumption

EMA−SIC Li et al.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 Energy consumption

Energy consumption

2500

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

1

Energy consumption

500

2

6

Energy consumption

1000

3 x 10 EMA−SIC Li et al.

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

6

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 136: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 140 × 140 m2 area.

50

Energy consumption

Energy consumption

5

1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Li et al.

3000 2000 1000

10000

6000 4000 2000 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

(d) β = 4

4

4 2

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

Fig. 137: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 140 × 140 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

4

2

2 1

4

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

5

0 100 200 300 400 500 600 700 800 900 1000

10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1 0.5

Number of nodes

(d) β = 4

5

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 138: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 140 × 140 m2 area.

Energy consumption

1

Energy consumption

7

1.5

(c) β = 3

(d) β = 4 7

2 x 10 EMA−SIC Li et al. 1.5

2.5 x 10 EMA−SIC 2 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

5

Energy consumption

10

2

Number of nodes

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

Energy consumption

2

15 x 10 EMA−SIC Li et al.

Energy consumption

4

4

(a) β = 1

6

15 x 10 EMA−SIC Li et al.

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4

8 x 10 EMA−SIC Li et al. 6

Energy consumption

Energy consumption

4 x 10 EMA−SIC Li et al. 3

(b) β = 2

6

Energy consumption

(f) β = 6

Fig. 139: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 160 × 160 m2 area.

Number of nodes

(a) β = 1

1

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

Number of nodes

Number of nodes

4

6 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

6

Energy consumption

Energy consumption

6

1

0.5

Number of nodes

(e) β = 5

2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

1.5

Energy consumption

1

Energy consumption

2

6

2.5 x 10 EMA−SIC 2 Li et al.

Energy consumption

3

10 x 10 EMA−SIC 8 Li et al.

Energy consumption

Energy consumption

5 x 10 EMA−SIC 4 Li et al.

5

Energy consumption

4

5

EMA−SIC Li et al.

8000

Number of nodes

Number of nodes

(c) β = 3

Number of nodes

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1 4000

5

2.5 x 10 EMA−SIC 2 Li et al.

500

Number of nodes

(b) β = 2

4

EMA−SIC Li et al.

1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

10

200

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Li et al.

400

1500

Energy consumption

2

EMA−SIC Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

600

Energy consumption

1

8 x 10 EMA−SIC Li et al. 6

Energy consumption

2

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 140: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 160 × 160 m2 area.

51

0 100 200 300 400 500 600 700 800 900 1000

4

2

Number of nodes

(b) β = 2

5

5

(c) β = 3

0.5

4 x 10 EMA−SIC Li et al. 3 2 1

8 x 10 EMA−SIC Li et al. 6 4 2

(e) β = 5

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(f) β = 6

Fig. 141: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 160 × 160 m2 area.

15 x 10 EMA−SIC Li et al.

Number of nodes

Number of nodes

Number of nodes

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

5

Energy consumption

1

Energy consumption

7

(f) β = 6

Fig. 143: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 180 × 180 m2 area.

200

EMA−SIC Li et al.

1500 1000 500

0 100 200 300 400 500 600 700 800 900 1000

2

1

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

6

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

Energy consumption

400

Energy consumption

Energy consumption

EMA−SIC Li et al.

2000

Energy consumption

6

600

1

(c) β = 3

(d) β = 4 7

2 x 10 EMA−SIC Li et al. 1.5

2

Number of nodes

Number of nodes

Number of nodes

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

10

Energy consumption

10

Energy consumption

2

Energy consumption

4

2

5

15 x 10 EMA−SIC Li et al.

6

15 x 10 EMA−SIC Li et al.

4

(a) β = 1 4

8 x 10 EMA−SIC Li et al. 6

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

6

Energy consumption

1

Number of nodes

(a) β = 1

Energy consumption

2

4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 x 10 EMA−SIC Li et al. 3

Energy consumption

1

6 x 10 EMA−SIC Li et al.

Energy consumption

2

4

6

Energy consumption

Energy consumption

6

3 x 10 EMA−SIC Li et al.

Number of nodes

(a) β = 1

(b) β = 2

(b) β = 2

3000 2000 1000

10000

EMA−SIC Li et al.

8000 6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

Number of nodes

(c) β = 3

(d) β = 4

4

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 142: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 180 × 180 m2 area.

7

7

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1.5

5 x 10 EMA−SIC 4 Li et al.

Energy consumption

2.5 x 10 EMA−SIC 2 Li et al.

Energy consumption

4

Energy consumption

4

15 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

6

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

4000

Energy consumption

6

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 144: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 180 × 180 m2 area.

52

500 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

10000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(b) β = 2 5

15 x 10 EMA−SIC Li et al. 10

5

4

Energy consumption

Energy consumption

4

2

Number of nodes

4

5000

4

(a) β = 1

5 x 10 EMA−SIC 4 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

3

8 x 10 EMA−SIC Li et al. 6

Energy consumption

1000

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

5

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 145: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 200 × 200 m2 area.

4 2

5

15 x 10 EMA−SIC Li et al.

Energy consumption

2000

0 100 200 300 400 500 600 700 800 900 1000

10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 146: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 200 × 200 m2 area.

6

3 x 10 EMA−SIC Li et al. 2

1

6

Energy consumption

3000

1

EMA−SIC Li et al.

Energy consumption

4000

15000

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

5000

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al.

Energy consumption

Number of nodes

2

4

10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 7

7

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Li et al. 3

Energy consumption

1000

Energy consumption

4

Energy consumption

200

EMA−SIC Li et al.

1500

Energy consumption

400

2000

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 147: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 200 × 200 m2 area.

53

500

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

500

3000

1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

15 x 10 EMA−SIC Li et al. 10

5

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 148: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 100 × 100 m2 area.

Energy consumption

Number of nodes

2 x 10 EMA−SIC Li et al. 1.5

(c) β = 3 3 x 10 EMA−SIC Li et al. 2

1

5

0 100 200 300 400 500 600 700 800 900 1000

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 149: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 100 × 100 m2 area.

6

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 5

Number of nodes

3 x 10 EMA−SIC Li et al. 2

1

6

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

6

Energy consumption

5000

2 x 10 EMA−SIC Li et al. 1.5

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 7

7

Energy consumption

10000

Energy consumption

Energy consumption

EMA−SIC Li et al.

2

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

4

15000

4

(a) β = 1 4

2000

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

1000

4000

0 100 200 300 400 500 600 700 800 900 1000 EMA−SIC Li et al.

Energy consumption

1500

5000

1

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

2000

(b) β = 2

2

4

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Li et al. 3

Energy consumption

4

1000

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 150: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 100 × 100 m2 area.

54

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1000 500

6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

1

(e) β = 5

1

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

EMA−SIC Li et al.

400

200

1500

EMA−SIC Li et al.

1000

500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 4000

(b) β = 2

EMA−SIC Li et al.

3000 2000 1000

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 152: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 120 × 120 m2 area.

10000

EMA−SIC Li et al.

8000 6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

5

Energy consumption

1

600

(d) β = 4

5

2

1

Fig. 153: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 120 × 120 m2 area.

Number of nodes

(c) β = 3

3

2

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

0.5

3 x 10 EMA−SIC Li et al.

5

Energy consumption

5

1

(b) β = 2

4

10

(d) β = 4 7

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

Number of nodes

(a) β = 1 15 x 10 EMA−SIC Li et al.

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

0 100 200 300 400 500 600 700 800 900 1000

4

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

5

7

(f) β = 6

Fig. 151: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 120 × 120 m2 area.

10

(c) β = 3 Energy consumption

2

Number of nodes

Number of nodes

Energy consumption

2

15 x 10 EMA−SIC Li et al.

Number of nodes

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

4

6

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

0.5

Energy consumption

Energy consumption

1

Energy consumption

8 x 10 EMA−SIC Li et al. 6

4

2 x 10 EMA−SIC Li et al. 1.5

(b) β = 2

6

(d) β = 4

4

2

Number of nodes

Energy consumption

1500

EMA−SIC Li et al.

(d) β = 4 4

4

Energy consumption

2000

8000

4

(a) β = 1

Energy consumption

EMA−SIC Li et al.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 Energy consumption

Energy consumption

2500

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

2

Energy consumption

500

6

Energy consumption

1000

3 x 10 EMA−SIC Li et al.

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

6

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 154: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 140 × 140 m2 area.

55

4 2

Fig. 155: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 140 × 140 m2 area.

0.5

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

4 2

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 5

15 x 10 EMA−SIC Li et al. 10

5

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

5

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 156: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 140 × 140 m2 area.

Energy consumption

Energy consumption

1 0.5

1

(d) β = 4

7

1.5

2

4

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

7

2.5 x 10 EMA−SIC 2 Li et al.

4 x 10 EMA−SIC Li et al. 3

Number of nodes

(c) β = 3

(f) β = 6

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

6

5

1

0 100 200 300 400 500 600 700 800 900 1000

4

10

2

Fig. 157: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 160 × 160 m2 area.

(b) β = 2 15 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3

(e) β = 5

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

1

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

Energy consumption

2

1.5

Energy consumption

Energy consumption

Energy consumption

4

6

4

4

4

6 x 10 EMA−SIC Li et al.

(a) β = 1 8 x 10 EMA−SIC Li et al. 6

(d) β = 4

4

2.5 x 10 EMA−SIC 2 Li et al.

6

Number of nodes

Number of nodes

(c) β = 3

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

5000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

EMA−SIC Li et al.

10000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

1

15000

Energy consumption

6

6

Number of nodes

(b) β = 2

Energy consumption

Energy consumption

Energy consumption

10 x 10 EMA−SIC 8 Li et al.

Number of nodes

2

1000

(d) β = 4 5

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Li et al.

2000

Energy consumption

(c) β = 3

2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

5

4

3000

Number of nodes

Number of nodes

6 x 10 EMA−SIC Li et al.

EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

500

Energy consumption

1

1000

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

2

EMA−SIC Li et al.

1500

Number of nodes

4000

5

3 x 10 EMA−SIC Li et al.

2000

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

4

5

200

Number of nodes

(a) β = 1

10

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Li et al.

Energy consumption

2

EMA−SIC Li et al.

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

600

Energy consumption

1

8 x 10 EMA−SIC Li et al. 6

Energy consumption

2

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 158: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 160 × 160 m2 area.

56

0 100 200 300 400 500 600 700 800 900 1000

4

2

Number of nodes

(b) β = 2

5

5

(c) β = 3

1 0.5

4 x 10 EMA−SIC Li et al. 3 2 1

8 x 10 EMA−SIC Li et al. 6 4 2

(e) β = 5

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(f) β = 6

Fig. 159: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 160 × 160 m2 area.

15 x 10 EMA−SIC Li et al.

Number of nodes

Number of nodes

Number of nodes

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

5

Energy consumption

1.5

Energy consumption

7

(f) β = 6

Fig. 161: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 180 × 180 m2 area.

200

EMA−SIC Li et al.

1500 1000 500

0 100 200 300 400 500 600 700 800 900 1000

2

1

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

6

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

Energy consumption

400

Energy consumption

Energy consumption

EMA−SIC Li et al.

2000

Energy consumption

6

600

1

(c) β = 3

(d) β = 4 7

2.5 x 10 EMA−SIC 2 Li et al.

2

Number of nodes

Number of nodes

Number of nodes

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

10

Energy consumption

10

Energy consumption

2

Energy consumption

4

2

5

15 x 10 EMA−SIC Li et al.

6

15 x 10 EMA−SIC Li et al.

4

(a) β = 1 4

8 x 10 EMA−SIC Li et al. 6

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

6

Energy consumption

1

Number of nodes

(a) β = 1

Energy consumption

2

4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 x 10 EMA−SIC Li et al. 3

Energy consumption

1

6 x 10 EMA−SIC Li et al.

Energy consumption

2

4

6

Energy consumption

Energy consumption

6

3 x 10 EMA−SIC Li et al.

Number of nodes

(a) β = 1

(b) β = 2

(b) β = 2

1000

10000

5000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

4

Energy consumption

Energy consumption

2

2

(d) β = 4

4

3 x 10 EMA−SIC Li et al.

4

15 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

Energy consumption

2000

EMA−SIC Li et al.

6

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 160: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 180 × 180 m2 area.

7

7

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

3000

15000

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

4000

Energy consumption

6

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 162: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 180 × 180 m2 area.

57

1000 500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

3000 2000 1000

15000

EMA−SIC Li et al.

10000

5000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

2

1

6

4

Energy consumption

Energy consumption

4

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 5

Energy consumption

10

5

Energy consumption

4

15 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

Energy consumption

Energy consumption

5

8 x 10 EMA−SIC Li et al. 6

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 164: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 200 × 200 m2 area.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

7

Energy consumption

1

(d) β = 4

7

Energy consumption

2

8 x 10 EMA−SIC Li et al. 6

Number of nodes

(c) β = 3

4

Energy consumption

Energy consumption

4

2

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000 4 x 10 EMA−SIC Li et al. 3

4

(a) β = 1

(f) β = 6

Fig. 163: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 200 × 200 m2 area.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

6

Energy consumption

EMA−SIC Li et al.

4000

Energy consumption

Energy consumption

5000

(b) β = 2

Energy consumption

200

EMA−SIC Li et al.

1500

Energy consumption

400

2000

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 165: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 200 × 200 m2 area.

58

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3 4000

EMA−SIC Li et al.

3000 2000 1000 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

6000

Energy consumption

Energy consumption

5000

(d) β = 4

15 x 10 EMA−SIC Li et al. 10

5

(b) β = 2 5

2 x 10 EMA−SIC Li et al. 1.5 1 0.5

0 100 200 300 400 500 600 700 800 900 1000

EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

4000

(d) β = 4

5

2000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 166: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 100 × 100 m2 area.

2

Number of nodes

4

1000

4

(a) β = 1

3 x 10 EMA−SIC Li et al. 2

1

5

Energy consumption

500

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 167: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 100 × 100 m2 area.

6

3 x 10 EMA−SIC Li et al. 2

1

6

Energy consumption

1000

0 100 200 300 400 500 600 700 800 900 1000 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al.

Energy consumption

1500

3000

1

10

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 7

7

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

2000

(b) β = 2

2

4

Energy consumption

(a) β = 1

Energy consumption

Number of nodes

4 x 10 EMA−SIC Li et al. 3

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

500

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

1000

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 168: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 100 × 100 m2 area.

59

1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

4000

2000

15000

5000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 169: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 120 × 120 m2 area.

2 1

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

1 0.5

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 171: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 120 × 120 m2 area.

600

EMA−SIC Li et al.

400

200

1500

EMA−SIC Li et al.

1000

500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

2000

EMA−SIC Li et al.

1500 1000 500

(b) β = 2 6000

EMA−SIC Li et al.

4000

2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

5

Energy consumption

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Li et al. 1.5

(d) β = 4

5

1

(d) β = 4 7

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

2

Number of nodes

(a) β = 1

2 x 10 EMA−SIC Li et al. 1.5

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

7

5

Energy consumption

Energy consumption

5

5

(c) β = 3

(b) β = 2

4

10

10

Number of nodes

Number of nodes

(a) β = 1 15 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al.

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

2

EMA−SIC Li et al.

10000

0 100 200 300 400 500 600 700 800 900 1000

4

Energy consumption

EMA−SIC Li et al.

8 x 10 EMA−SIC Li et al. 6

6

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4 Energy consumption

Energy consumption

6000

(b) β = 2

6

Energy consumption

500

2000

2

Number of nodes

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 170: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 120 × 120 m2 area.

4

15000

EMA−SIC Li et al.

10000

5000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1000

EMA−SIC Li et al.

Energy consumption

1500

3000

4

(a) β = 1

Energy consumption

EMA−SIC Li et al.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 Energy consumption

Energy consumption

2000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

1

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

6

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Li et al.

Energy consumption

500

Energy consumption

1000

Energy consumption

200

EMA−SIC Li et al.

Energy consumption

400

6

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 172: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 140 × 140 m2 area.

60

0.5

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al. 4

2

(f) β = 6

Fig. 173: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 140 × 140 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2000 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4 4

15000

EMA−SIC Li et al.

10000

5000

2

(f) β = 6

Fig. 175: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 160 × 160 m2 area.

4 x 10 EMA−SIC Li et al. 3 2 1

4

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

5

5

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(d) β = 4

5

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 174: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 140 × 140 m2 area.

Energy consumption

1

Energy consumption

7

Number of nodes

(c) β = 3

(d) β = 4 7

2 x 10 EMA−SIC Li et al. 1.5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

0.5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

10

5

2 x 10 EMA−SIC Li et al. 1.5

Energy consumption

15 x 10 EMA−SIC Li et al.

Energy consumption

Energy consumption

2

10

2

Number of nodes

4

4

4

(a) β = 1

6

15 x 10 EMA−SIC Li et al.

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6

1

(e) β = 5

Number of nodes

(a) β = 1

2

Number of nodes

Number of nodes

Energy consumption

4

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Energy consumption

4000

4

6 x 10 EMA−SIC Li et al.

EMA−SIC Li et al.

6000

Number of nodes

6

Energy consumption

Energy consumption

6

Energy consumption

8000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

500

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

Number of nodes

(b) β = 2

(d) β = 4

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Li et al.

1000

5

Energy consumption

Energy consumption

1

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

(c) β = 3

2

1500

Number of nodes

5

4 x 10 EMA−SIC Li et al. 3

EMA−SIC Li et al.

2000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

500

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

1

EMA−SIC Li et al.

1000

Number of nodes

2500

5

2 x 10 EMA−SIC Li et al. 1.5

1500

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

4

5

200

Number of nodes

(a) β = 1

10

400

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Li et al.

Energy consumption

2

EMA−SIC Li et al.

Energy consumption

0 100 200 300 400 500 600 700 800 900 1000

4

600

Energy consumption

1

8 x 10 EMA−SIC Li et al. 6

Energy consumption

2

4

Energy consumption

Energy consumption

4

4 x 10 EMA−SIC Li et al. 3

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 176: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 160 × 160 m2 area.

61

0 100 200 300 400 500 600 700 800 900 1000

4

2

1

10

5

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

2

Energy consumption

4

15 x 10 EMA−SIC Li et al.

6

15 x 10 EMA−SIC Li et al.

10

5

(b) β = 2 2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

5

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

Number of nodes

Energy consumption

6

4

(a) β = 1

(b) β = 2 4

8 x 10 EMA−SIC Li et al. 6

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 Energy consumption

2

4

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 x 10 EMA−SIC Li et al. 3

Energy consumption

1

6 x 10 EMA−SIC Li et al.

Energy consumption

2

4

6

Energy consumption

Energy consumption

6

3 x 10 EMA−SIC Li et al.

Number of nodes

(c) β = 3

(d) β = 4

(d) β = 4

5

2

1

4

2

(e) β = 5

(f) β = 6

Fig. 177: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 160 × 160 m2 area.

6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

Number of nodes

10 x 10 EMA−SIC 8 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al.

Energy consumption

0.5

Energy consumption

Energy consumption

1

3 x 10 EMA−SIC Li et al.

Energy consumption

7

7

2 x 10 EMA−SIC Li et al. 1.5

5

Number of nodes

(e) β = 5

(f) β = 6

Fig. 179: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 180 × 180 m2 area.

200

EMA−SIC Li et al.

1000

500

0 100 200 300 400 500 600 700 800 900 1000

2

1

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

6

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

Energy consumption

400

Energy consumption

Energy consumption

EMA−SIC Li et al.

1500

Energy consumption

6

600

Number of nodes

(a) β = 1

(b) β = 2

(b) β = 2

2000

1000

10000

EMA−SIC Li et al.

8000 6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

Number of nodes

(c) β = 3

(d) β = 4

4

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 178: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 180 × 180 m2 area.

7

7

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1

4 x 10 EMA−SIC Li et al. 3

Energy consumption

2 x 10 EMA−SIC Li et al. 1.5

Energy consumption

4

Energy consumption

2

15 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

4

6

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

3000

Energy consumption

6

8 x 10 EMA−SIC Li et al. 6

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 180: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 180 × 180 m2 area.

62

200

EMA−SIC Li et al.

1000

500

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 EMA−SIC Li et al.

3000 2000 1000

10000

Energy consumption

EMA−SIC Li et al.

8000 6000 4000 2000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

1 0.5

4 x 10 EMA−SIC Li et al. 3 2 1

Energy consumption

1.5

Energy consumption

Energy consumption

6

4

4

2.5 x 10 EMA−SIC 2 Li et al.

3 x 10 EMA−SIC Li et al. 2

1

(e) β = 5

(b) β = 2

6

8 x 10 EMA−SIC Li et al. 6 4 2

6

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

4

10

5

5

Energy consumption

Energy consumption

15 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

Energy consumption

Energy consumption

5

6 x 10 EMA−SIC Li et al.

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 182: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 200 × 200 m2 area.

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Energy consumption

1

4

(d) β = 4 7

7

Energy consumption

2

8 x 10 EMA−SIC Li et al. 6

Number of nodes

(c) β = 3

4

Energy consumption

Energy consumption

4

2

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000 4 x 10 EMA−SIC Li et al. 3

4

(a) β = 1

(f) β = 6

Fig. 181: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 200 × 200 m2 area.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

Number of nodes

6

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Energy consumption

Energy consumption

4000

(b) β = 2

Energy consumption

400

1500

Energy consumption

EMA−SIC Li et al.

Energy consumption

Energy consumption

600

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 183: A separate comparison of energy consumption (joule) between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 200 × 200 m2 area.

63

times energy

2

18

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Number of nodes

(c) β = 3

(d) β = 4

10

times energy

Number of nodes

(e) β = 5

(f) β = 6

Fig. 184: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 100 × 100 m2 area. times energy α−1

times energy α−1

times energy

(f) β = 6

α−1

times energy

9

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

times energy α−1

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

Fig. 185: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 100 × 100 m2 area.

11

times energy

Number of nodes

(d) β = 4

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

times energy

15

6 x 10 EMA−SIC Cell−AS Li et al. 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(b) β = 2 10

Latency

Number of nodes

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

(e) β = 5

Number of nodes

Number of nodes

(d) β = 4

α−1

Number of nodes

1 0.5

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

15

1

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0.5

(c) β = 3

2

1 0.5

Latency

times energy α−1

1

Latency

times energy α−1

Latency

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

9

times energy

(b) β = 2

α−1

Number of nodes

14

0 100 200 300 400 500 600 700 800 900 1000

0.5

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

2

1.5

5

Number of nodes

6 x 10 EMA−SIC Cell−AS Li et al. 4

1

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Fig. 186: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 100 × 100 m2 area.

Latency

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(e) β = 5

times energy

1

(d) β = 4 20

Number of nodes

α−1

2

13

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

Latency

α−1

times energy

13

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Number of nodes

α−1

Number of nodes

2

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

0 100 200 300 400 500 600 700 800 900 1000

α−1

0.5

Latency

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

times energy

2

5

Number of nodes

α−1

Latency

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

(b) β = 2 19

20

α−1

α−1

6

11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latency

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

2

Number of nodes

α−1

times energy

1

4

(a) β = 1

α−1

2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

18

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

5

Latency

times energy

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

(b) β = 2 10

1

Latency

(a) β = 1

18

times energy

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

1

α−1

times energy

3

Latency

0 100 200 300 400 500 600 700 800 900 1000

α−1

times energy α−1

Latency

5

9

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 187: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 120 × 120 m2 area.

64

Number of nodes

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy α−1

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 189: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 120 × 120 m2 area.

times energy α−1

times energy α−1

Latency

times energy α−1

Latency

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

times energy

14

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 15

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

3

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

20

Latency

1

5

(d) β = 4

20

2

times energy

Number of nodes

(c) β = 3 4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

0.5

(d) β = 4

15

α−1

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

15

α−1

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1 times energy

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(d) β = 4 11

Number of nodes

20

Latency

times energy α−1

Latency

2

0.5

(b) β = 2

19

4 x 10 EMA−SIC Cell−AS 3 Li et al.

1

13

Number of nodes

(a) β = 1

0 100 200 300 400 500 600 700 800 900 1000

Fig. 190: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 140 × 140 m2 area.

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

(e) β = 5

Latency

times energy α−1

5

4

Number of nodes

Number of nodes

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

Latency

α−1

times energy

18

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(f) β = 6

Fig. 188: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 120 × 120 m2 area.

6

times energy

(e) β = 5

1

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

11

times energy

0 100 200 300 400 500 600 700 800 900 1000

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(c) β = 3

Latency

α−1

times energy

15 x 10 EMA−SIC Cell−AS Li et al. 10

(b) β = 2 10

Number of nodes

15

Latency

times energy α−1

Latency

2

2

(d) β = 4

15

Number of nodes

α−1

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

6 x 10 EMA−SIC Cell−AS Li et al. 4

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

10

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

2

(a) β = 1

Latency

times energy α−1

2

Latency

times energy α−1

4

Latency

(b) β = 2 3 x 10 EMA−SIC Cell−AS Li et al. 2

4

Number of nodes

15

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

6

Latency

(a) β = 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

16

times energy

Number of nodes

1

α−1

0 100 200 300 400 500 600 700 800 900 1000

9

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

α−1 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

9

3 x 10 EMA−SIC Cell−AS Li et al. 2

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy α−1

2

Latency

times energy α−1

2

Latency

14

4 x 10 EMA−SIC Cell−AS 3 Li et al.

times energy

13

6 x 10 EMA−SIC Cell−AS Li et al. 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 191: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 140 × 140 m2 area.

65

Number of nodes

0.5 0 100 200 300 400 500 600 700 800 900 1000

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 193: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 160 × 160 m2 area.

times energy α−1

times energy α−1

Latency

times energy α−1

Latency

Number of nodes

(f) β = 6

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

times energy

19

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0.5 0 100 200 300 400 500 600 700 800 900 1000

times energy

1

α−1

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(b) β = 2 20

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

3

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

20

11

Latency

times energy α−1

Latency

1

5

(d) β = 4

11

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

20

Number of nodes

(c) β = 3

1

(a) β = 1

α−1

α−1

1

2

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

11

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

0.5

(b) β = 2

10

2

times energy α−1

Number of nodes

(a) β = 1

4

1

18

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

1.5

(d) β = 4 16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Fig. 194: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 160 × 160 m2 area.

times energy

0 100 200 300 400 500 600 700 800 900 1000

0.5

2 0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

α−1

1

Latency

α−1

1

4

Number of nodes

Number of nodes

10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

times energy

times energy Latency

α−1

9

2

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(f) β = 6

Fig. 192: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 140 × 140 m2 area.

3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0.5

15 x 10 EMA−SIC Cell−AS Li et al. 10

21

times energy

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6

(c) β = 3

α−1

5

(b) β = 2 15

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

Latency

times energy

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

0 100 200 300 400 500 600 700 800 900 1000

1

16

20

Latency

Latency

α−1

times energy

20

2

1.5

(d) β = 4 times energy

(c) β = 3

4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Number of nodes

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

times energy

2

Number of nodes

15

Latency

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

Latency

times energy α−1

α−1

6

Latency

(b) β = 2

Latency

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

5

Number of nodes

20

19

0.5

α−1

(a) β = 1

1

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

Latency

1

14

times energy

times energy

2

α−1

2

19

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

Latency

α−1

times energy

18

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 195: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 160 × 160 m2 area.

66

times energy α−1

Latency times energy

α−1

Latency

times energy

Number of nodes

(d) β = 4

α−1

times energy

16

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 197: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 180 × 180 m2 area.

19

2

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

times energy

4 x 10 EMA−SIC Cell−AS 3 Li et al.

times energy

Fig. 196: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 180 × 180 m2 area.

1

α−1

(f) β = 6

2

Latency

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

20

times energy

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

16

α−1

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

0.5

(c) β = 3

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

1

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

20

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3 21

times energy

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

times energy α−1

4

0 100 200 300 400 500 600 700 800 900 1000

11

Latency

α−1

times energy

11

8 x 10 EMA−SIC Cell−AS 6 Li et al.

2

Latency

(d) β = 4

16

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

times energy

(c) β = 3

15

6 x 10 EMA−SIC Cell−AS Li et al. 4

α−1

Number of nodes

Number of nodes

Number of nodes

(b) β = 2

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

α−1

1

1 0.5

Number of nodes

Latency

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latency

α−1

4

times energy

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Latency

0 100 200 300 400 500 600 700 800 900 1000

11

Latency

times energy α−1

Latency

(b) β = 2

1

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

(d) β = 4 21

times energy

(a) β = 1 10

1.5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

Number of nodes

Number of nodes

2

3

15

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

2

Latency

0 100 200 300 400 500 600 700 800 900 1000

times energy

1

14

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latency

α−1

2

Latency

times energy α−1

Latency

4

times energy

10

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

3

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 198: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 180 × 180 m2 area.

67

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

(d) β = 4

α−1

times energy

16

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy α−1

Latency

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latency

α−1

times energy times energy

times energy

times energy

(b) β = 2 10

5

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 200: Latency-energy tradeoff comparison under α = 4 and uniform distribution in a 200 × 200 m2 area.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3 8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

4

10

times energy

0.5

Number of nodes

16

15 x 10 EMA−SIC Cell−AS Li et al. 10

times energy

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10

1

1

0 100 200 300 400 500 600 700 800 900 1000

α−1

(b) β = 2 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

3

(a) β = 1

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

Number of nodes

Latency

1 0.5

16

Latency

times energy α−1

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

9

(a) β = 1

4

2

1.5

Number of nodes

15

8 x 10 EMA−SIC Cell−AS 6 Li et al.

4

times energy

α−1

6

9

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

8

α−1

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(f) β = 6

Fig. 201: Latency-energy tradeoff comparison under α = 5 and uniform distribution in a 200 × 200 m2 area.

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

15

Latency

1

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5 (f) β = 6

times energy

times energy α−1

Latency

2

5

Number of nodes

Fig. 199: Latency-energy tradeoff comparison under α = 3 and uniform distribution in a 200 × 200 m2 area.

3

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

14

1

0.5

(e) β = 5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

times energy α−1

2

3

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

1

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

21

α−1

Number of nodes

Number of nodes

(d) β = 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

21

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

times energy

12

α−1

2

1 0.5

(c) β = 3 (d) β = 4

Latency

times energy α−1

Latency

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

Number of nodes

11

2

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3 8 x 10 EMA−SIC Cell−AS 6 Li et al.

6 x 10 EMA−SIC Cell−AS Li et al. 4

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 21

Latency

2

Latency

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

11

α−1

2

Number of nodes

20

times energy

Latency

4

1

(a) β = 1 (b) β = 2

Latency

times energy α−1

6

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

times energy

10

times energy

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1 10 x 10 EMA−SIC 8 Cell−AS Li et al.

1

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

Latency

2

α−1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

5

Latency

10

times energy

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

times energy

20

19

4 x 10 EMA−SIC Cell−AS 3 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 202: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 100 × 100 m2 area.

68

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 204: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 100 × 100 m2 area.

times energy α−1

Latency times energy

α−1

Latency

times energy

times energy α−1

Latency

times energy α−1

Latency

times energy

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 times energy α−1

4 2

Latency

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

15

6 x 10 EMA−SIC Cell−AS Li et al. 4

15

times energy

times energy

2

1 0.5

Number of nodes

α−1

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

1.5

15

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

20

α−1

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

(d) β = 4

Latency

Latency

α−1

times energy

19

4

14

Number of nodes

(c) β = 3

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1 times energy

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 11

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

Number of nodes

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

α−1

0.5

2

0 100 200 300 400 500 600 700 800 900 1000

Fig. 205: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 120 × 120 m2 area.

19

3

2

(e) β = 5

(b) β = 2 5 x 10 EMA−SIC 4 Cell−AS Li et al.

(b) β = 2 10

6 x 10 EMA−SIC Cell−AS Li et al. 4

Number of nodes

Number of nodes

Latency

times energy α−1

Latency

1

0 100 200 300 400 500 600 700 800 900 1000

α−1

2

(a) β = 1 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2

Latency

times energy α−1

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

19

4

13

18

Latency

Latency

α−1

times energy

18

0 100 200 300 400 500 600 700 800 900 1000

6

(f) β = 6

Fig. 203: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 100 × 100 m2 area.

Number of nodes

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

(e) β = 5

1

times energy α−1

1

2

(c) β = 3

α−1

2

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

times energy

15

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

0.5

(d) β = 4

15

1

1

Number of nodes

(c) β = 3 3 x 10 EMA−SIC Cell−AS Li et al. 2

10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

Latency

1

5

9

Number of nodes

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

times energy

times energy α−1

2

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

14

3

0.5

Number of nodes

(a) β = 1 5 x 10 EMA−SIC 4 Cell−AS Li et al.

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1.5

times energy

0 100 200 300 400 500 600 700 800 900 1000

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latency

1

5

Latency

Latency

α−1

times energy

times energy α−1

2

13

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

13

4 x 10 EMA−SIC Cell−AS 3 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 206: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 120 × 120 m2 area.

69

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 208: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 140 × 140 m2 area.

times energy α−1

Latency times energy

α−1

times energy α−1

times energy

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 20

2

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

20

times energy

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latency

α−1

4

0 100 200 300 400 500 600 700 800 900 1000

α−1

times energy α−1

0.5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

Latency

times energy α−1

Latency

1

1

19

11

11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

Number of nodes

(c) β = 3

0.5

(f) β = 6

Number of nodes

α−1

4

1

Number of nodes

α−1

times energy

6

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

times energy α−1

Latency

1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(b) β = 2

10

3 x 10 EMA−SIC Cell−AS Li et al. 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

19

18

Number of nodes

(a) β = 1

(d) β = 4 16

times energy

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4

21

α−1

0.5

5

1

Fig. 209: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 140 × 140 m2 area.

α−1

times energy α−1

Latency

1

Latency

times energy α−1

1.5

3

(e) β = 5

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

times energy

9

(b) β = 2 15

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Number of nodes

(f) β = 6

Fig. 207: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 120 × 120 m2 area.

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Number of nodes

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy

1

5

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

3

2

15

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

0.5

(c) β = 3

Latency

Latency

1

(d) β = 4

α−1

α−1

1.5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

times energy

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

5

Number of nodes

20

14

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latency

α−1

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3 20

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

5

Number of nodes

14

times energy

α−1

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

Latency

1

(b) β = 2

times energy

2

3

Number of nodes

19

times energy

times energy α−1

19

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

(a) β = 1 5 x 10 EMA−SIC 4 Cell−AS Li et al.

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5

Latency

0.5

Latency

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

times energy

1

α−1

1

13

19

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Cell−AS Li et al. 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 210: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 140 × 140 m2 area.

70

4 2 0 100 200 300 400 500 600 700 800 900 1000

α−1

times energy

times energy times energy α−1

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

(b) β = 2

15

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 212: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 160 × 160 m2 area.

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

11

times energy

times energy

4 x 10 EMA−SIC Cell−AS 3 Li et al.

α−1

(d) β = 4

α−1

Number of nodes

Latency

2 0 100 200 300 400 500 600 700 800 900 1000

1.5

(d) β = 4

11

4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(c) β = 3

6

times energy

times energy α−1

times energy

4

11

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latency

(b) β = 2

α−1

(e) β = 5

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

Number of nodes

2

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

2

Latency

0 100 200 300 400 500 600 700 800 900 1000

10

10

4

Latency

times energy α−1

Latency

times energy α−1

Latency

1

6 x 10 EMA−SIC Cell−AS Li et al. 4

6

16

0.5

Latency

9

Latency

14

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(f) β = 6

Fig. 213: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 160 × 160 m2 area. times energy

(f) β = 6

(c) β = 3 16

1

Number of nodes

α−1

Number of nodes

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Latency

0 100 200 300 400 500 600 700 800 900 1000

15

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

Number of nodes

(a) β = 1

1

Number of nodes

(e) β = 5

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

α−1

6

2

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

times energy α−1

4

14

0.5

2

3

Latency

11

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Fig. 211: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 160 × 160 m2 area.

1

(b) β = 2 20

5 x 10 EMA−SIC 4 Cell−AS Li et al.

21

20

times energy

(d) β = 4

(e) β = 5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

α−1

times energy times energy α−1

times energy α−1

Number of nodes

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

11

1

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

2

5

5

Latency

times energy α−1

Latency

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

Number of nodes

(b) β = 2

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

10

0 100 200 300 400 500 600 700 800 900 1000

3

(a) β = 1

(a) β = 1

2

2

Number of nodes

0.5

Number of nodes

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

1

Latency

Latency

1

5

times energy

2

3

10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

times energy

times energy α−1

9

5 x 10 EMA−SIC 4 Cell−AS Li et al.

19

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 214: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 180 × 180 m2 area.

71

0 100 200 300 400 500 600 700 800 900 1000

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

Latency

α−1

times energy

times energy

times energy α−1

Latency

times energy

0 100 200 300 400 500 600 700 800 900 1000

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 216: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 180 × 180 m2 area.

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

15

16

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1

α−1

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

21

2

times energy α−1

1

(d) β = 4 4 x 10 EMA−SIC Cell−AS 3 Li et al.

1.5

Latency

times energy

2

15

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latency

2

times energy

(f) β = 6

(a) β = 1

Number of nodes

α−1

1

Number of nodes

Number of nodes

α−1

4

Latency

Latency

α−1

times energy

21

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

16

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

3

16

times energy

times energy

6

(c) β = 3 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

times energy

20

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

5

Fig. 217: Latency-energy tradeoff comparison under α = 3 and poisson distribution in a 200 × 200 m2 area.

Latency

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

(e) β = 5

(b) β = 2

α−1

0.5

2

Number of nodes

Latency

times energy α−1

Latency

1

4

α−1

0 100 200 300 400 500 600 700 800 900 1000

20

(d) β = 4 11

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

2

(a) β = 1

Number of nodes

Number of nodes

α−1

4

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

6

1

0 100 200 300 400 500 600 700 800 900 1000

14

19

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3

(f) β = 6

18

Number of nodes

Number of nodes

Number of nodes

Fig. 215: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 180 × 180 m2 area.

5

0 100 200 300 400 500 600 700 800 900 1000

α−1

2

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

11

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

α−1

0.5

6

Latency

times energy α−1

1

Latency

times energy α−1

Latency

1.5

1

11

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(d) β = 4 6 x 10 EMA−SIC Cell−AS Li et al. 4

2

(a) β = 1

Number of nodes

16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

0.5

10

Number of nodes

α−1

α−1

1

(c) β = 3 16

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

2

10

16

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

4

(b) β = 2

15

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Number of nodes

(a) β = 1 3 x 10 EMA−SIC Cell−AS Li et al. 2

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

times energy α−1

1

9

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

Latency

α−1

times energy

14

3 x 10 EMA−SIC Cell−AS Li et al. 2

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 218: Latency-energy tradeoff comparison under α = 4 and poisson distribution in a 200 × 200 m2 area.

72

20

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

Number of nodes

(e) β = 5

(f) β = 6

Fig. 219: Latency-energy tradeoff comparison under α = 5 and poisson distribution in a 200 × 200 m2 area. 8

α−1

times energy

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

6

(e) β = 5

times energy α−1

times energy

times energy α−1

α−1

1

times energy

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

times energy

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

18

times energy

2

3

4

times energy

(d) β = 4 10

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

Number of nodes

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

1 19

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 220: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 100 × 100 m2 area.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1

(d) β = 4 19

times energy

Number of nodes

18

α−1

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0.5

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

α−1

1

Number of nodes

Fig. 221: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 100 × 100 m2 area.

2

Latency

times energy α−1

Latency

1.5

2 0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

4

(c) β = 3 2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

4

α−1

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

10

6

Latency

(b) β = 2 9

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

14

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Number of nodes

times energy

Number of nodes

9

1

Number of nodes

(d) β = 4

18

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1

2

0 100 200 300 400 500 600 700 800 900 1000

5

Number of nodes

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2

0 100 200 300 400 500 600 700 800 900 1000

α−1

0 100 200 300 400 500 600 700 800 900 1000

1 0.5

14

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latency

2

Latency

Latency

α−1

times energy

8

1.5

(c) β = 3

α−1

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

(b) β = 2 14

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Number of nodes

α−1

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latency

times energy α−1

3 x 10 EMA−SIC Cell−AS Li et al. 2

5

21

Latency

Latency

α−1

times energy

21

2

Number of nodes

13

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

0 100 200 300 400 500 600 700 800 900 1000

4

(a) β = 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

1

Latency

Latency

α−1

times energy

times energy α−1

2

0 100 200 300 400 500 600 700 800 900 1000

20

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

20

4 x 10 EMA−SIC Cell−AS 3 Li et al.

1

Latency

(b) β = 2

times energy

Number of nodes

(a) β = 1

Latency

Number of nodes

2

13

Latency

times energy

0 100 200 300 400 500 600 700 800 900 1000

α−1

0 100 200 300 400 500 600 700 800 900 1000

0.5

Latency

1

13

4 x 10 EMA−SIC Cell−AS 3 Li et al.

α−1

α−1

1

Latency

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latency

Latency

α−1

times energy

19

3 x 10 EMA−SIC Cell−AS Li et al. 2

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 222: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 100 × 100 m2 area.

73

times energy α−1

Number of nodes

(e) β = 5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 224: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 120 × 120 m2 area.

times energy α−1

Latency

times energy α−1

Latency times energy

α−1

Latency

times energy α−1

Latency

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 times energy α−1

4

10

2

Number of nodes

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

10

15

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

14

5

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 15 x 10 EMA−SIC Cell−AS Li et al. 10

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

9

Latency

1

0 100 200 300 400 500 600 700 800 900 1000

(a) β = 1 times energy

times energy

2

3

5

Number of nodes

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

α−1

5

4

0 100 200 300 400 500 600 700 800 900 1000

14

Latency

times energy α−1

Latency

15 x 10 EMA−SIC Cell−AS Li et al. 10

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(b) β = 2 5 x 10 EMA−SIC 4 Cell−AS Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

9

8

Number of nodes

(a) β = 1 13

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 19

6 x 10 EMA−SIC Cell−AS Li et al. 4

10

times energy

2

Number of nodes

Fig. 225: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 120 × 120 m2 area.

α−1

4

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

Latency

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Cell−AS 6 Li et al.

5

Number of nodes

13

Latency

Latency

α−1

times energy

13

1

2

(f) β = 6

Fig. 223: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 120 × 120 m2 area.

2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy α−1

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

19

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

(b) β = 2 18

(c) β = 3

Latency

times energy α−1

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Number of nodes

Number of nodes

10

Latency

times energy α−1

Latency

2

4

(d) β = 4

10

4 x 10 EMA−SIC Cell−AS 3 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

18

α−1

5

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1

Latency

α−1

times energy

15 x 10 EMA−SIC Cell−AS Li et al. 10

18

Number of nodes

times energy

9

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

(b) β = 2

9

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 6 x 10 EMA−SIC Cell−AS Li et al. 4

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

1 0.5

Latency

times energy

1.5

α−1

2

18

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Cell−AS Li et al. 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 226: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 140 × 140 m2 area.

74

Number of nodes

(e) β = 5

1

times energy α−1

Latency times energy

α−1

times energy α−1

Latency

times energy

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

(b) β = 2

2

1

α−1

3

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

times energy

5 x 10 EMA−SIC 4 Cell−AS Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

15

20

times energy

14

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Number of nodes

(d) β = 4

α−1

0 100 200 300 400 500 600 700 800 900 1000

1

Number of nodes

Latency

times energy α−1

Latency

5

(f) β = 6

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

19

Number of nodes

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 228: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 140 × 140 m2 area.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2

15

times energy

times energy

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

14

19

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

2

0 100 200 300 400 500 600 700 800 900 1000

α−1

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(b) β = 2

α−1

5

0.5

Fig. 229: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 160 × 160 m2 area.

Number of nodes

Latency

Latency

α−1

times energy

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

α−1

2

(a) β = 1

1

(e) β = 5

α−1

times energy α−1

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4 11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

13

18

Latency

Latency

α−1

times energy

18

Number of nodes

Number of nodes

(f) β = 6

Fig. 227: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 140 × 140 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Cell−AS Li et al. 2

times energy

times energy

1

5

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

3

1

0 100 200 300 400 500 600 700 800 900 1000

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

times energy

0.5

α−1

times energy Latency

α−1

1

3

(c) β = 3

15

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2

Number of nodes

(d) β = 4

Latency

15

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

5

(b) β = 2 10

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

2

Number of nodes

times energy α−1

4

α−1

0 100 200 300 400 500 600 700 800 900 1000

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

0.5

9

Latency

times energy α−1

6

Latency

times energy α−1

Latency

1

2

(a) β = 1

14

10 x 10 EMA−SIC 8 Cell−AS Li et al.

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

14

1.5

0 100 200 300 400 500 600 700 800 900 1000

α−1

(a) β = 1 2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

5

Number of nodes

9

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy α−1

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

Latency

1

Latency

Latency

α−1

6

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latency

times energy

times energy α−1

2

13

10 x 10 EMA−SIC 8 Cell−AS Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

13

4 x 10 EMA−SIC Cell−AS 3 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 230: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 160 × 160 m2 area.

75

18

(b) β = 2

times energy

(c) β = 3

(e) β = 5

(f) β = 6

Fig. 231: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 160 × 160 m2 area. times energy

4

times energy α−1

times energy times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

1.5

19

2

0 100 200 300 400 500 600 700 800 900 1000

1

19

times energy

6 x 10 EMA−SIC Cell−AS Li et al. 4

α−1

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(b) β = 2

Latency

(d) β = 4

Number of nodes

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 232: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 180 × 180 m2 area.

(d) β = 4

20

3 x 10 EMA−SIC Cell−AS Li et al. 2

20

times energy

times energy

1

α−1

(e) β = 5

2

19

Number of nodes

11

α−1

Number of nodes

18

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

Fig. 233: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 180 × 180 m2 area.

2

(c) β = 3

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

(e) β = 5

α−1

(b) β = 2 10

Number of nodes

Number of nodes

Latency

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

times energy

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

Latency

1 0.5

1

2

(a) β = 1 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

15

15 x 10 EMA−SIC Cell−AS Li et al. 10

4

Number of nodes

10

2

3

α−1

0 100 200 300 400 500 600 700 800 900 1000

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

0.5

0 100 200 300 400 500 600 700 800 900 1000

6

α−1

1

9

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

Latency

α−1

times energy

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

1 0.5

Number of nodes

α−1

Number of nodes

1.5

(d) β = 4

Latency

1

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(c) β = 3 15

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

times energy

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

α−1

0.5

2

times energy

α−1

1

times energy

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(d) β = 4 20

Latency

Latency

α−1

times energy

20

(b) β = 2 15

α−1

Number of nodes

α−1

Number of nodes

4

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

2

Number of nodes

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

(a) β = 1

times energy

0.5

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

times energy

6

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

19

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

Latency

α−1

times energy

19

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2

times energy

Number of nodes

(a) β = 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

times energy

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

4

Latency

1

14

13

6 x 10 EMA−SIC Cell−AS Li et al. 4

α−1

α−1

6

Latency

times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Cell−AS Li et al. 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 234: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 180 × 180 m2 area.

76

9

α−1

times energy

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 times energy

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6

α−1

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

18

2 1

4 2

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

Number of nodes

(f) β = 6

(b) β = 2

19

Latency

α−1

Fig. 235: Latency-energy tradeoff comparison under α = 3 and cluster distribution in a 200 × 200 m2 area.

1.5

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

19

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

times energy

8 x 10 EMA−SIC Cell−AS 6 Li et al.

α−1

α−1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

times energy α−1

1.5

(d) β = 4 11

Latency

Latency

α−1

times energy

11

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latency

(c) β = 3

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4

20

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

α−1

1

2

0 100 200 300 400 500 600 700 800 900 1000

Latency

3 x 10 EMA−SIC Cell−AS Li et al. 2

10

Latency

Latency

α−1

times energy

10

times energy

0 100 200 300 400 500 600 700 800 900 1000

5

times energy

1

Latency

Latency

α−1

times energy

9

3 x 10 EMA−SIC Cell−AS Li et al. 2

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

α−1

times energy

15

5

0 100 200 300 400 500 600 700 800 900 1000

Latency

Latency

α−1

times energy

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

(b) β = 2 6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 times energy α−1

15 x 10 EMA−SIC Cell−AS Li et al. 10

16

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

Latency

α−1

times energy

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 236: Latency-energy tradeoff comparison under α = 4 and cluster distribution in a 200 × 200 m2 area.

2

α−1

3

20

times energy

times energy

1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

α−1

2

20

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latency

times energy α−1

5

(d) β = 4

14

Latency

Latency

α−1

times energy

13

15 x 10 EMA−SIC Cell−AS Li et al. 10

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 237: Latency-energy tradeoff comparison under α = 5 and cluster distribution in a 200 × 200 m2 area.

77

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy

times energy α−1

Latency

(b) β = 2 14

times energy

2 x 10 EMA−SIC Li et al. 1.5

α−1

1

Number of nodes

(d) β = 4

(c) β = 3

Latency

0 100 200 300 400 500 600 700 800 900 1000

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

14

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 238: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 100 × 100 m2 area.

3 x 10 EMA−SIC Li et al. 2

14

times energy

15 x 10 EMA−SIC Li et al.

α−1

times energy α−1

4

5

Number of nodes

9

Latency

Latency

α−1

times energy

9

2

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

(c) β = 3 8 x 10 EMA−SIC Li et al. 6

times energy α−1

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

4

0 100 200 300 400 500 600 700 800 900 1000

13

15 x 10 EMA−SIC Li et al.

times energy

α−1

0.5

Latency

times energy Latency

α−1

1

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

9

4 x 10 EMA−SIC Li et al. 3

13

Number of nodes

(b) β = 2 times energy

9

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 2 x 10 EMA−SIC Li et al. 1.5

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

2

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

10

α−1

4

8

15 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 239: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 100 × 100 m2 area.

78

times energy α−1

Latency

times energy

1

α−1

0 100 200 300 400 500 600 700 800 900 1000

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

1

1

0 100 200 300 400 500 600 700 800 900 1000

Fig. 240: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 100 × 100 m2 area.

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(f) β = 6

4

α−1

2

14

8 x 10 EMA−SIC Li et al. 6

times energy

4 x 10 EMA−SIC Li et al. 3

α−1

times energy

2

(e) β = 5

2 x 10 EMA−SIC Li et al. 1.5

14

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 14

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

(d) β = 4

α−1

0.5

10

Number of nodes

Latency

times energy Latency

α−1

1

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

19

19

2 x 10 EMA−SIC Li et al. 1.5

15 x 10 EMA−SIC Li et al.

α−1

5

(c) β = 3

2

(a) β = 1

Latency

times energy

10

Number of nodes

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

18

α−1

0 100 200 300 400 500 600 700 800 900 1000

13

8 x 10 EMA−SIC Li et al. 6

13

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

18

4

1

Number of nodes

(a) β = 1 8 x 10 EMA−SIC Li et al. 6

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

1

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

4

α−1

2

13

18

6 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

Number of nodes

(e) β = 5

(f) β = 6

Fig. 242: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 120 × 120 m2 area.

Number of nodes

(e) β = 5

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 241: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 120 × 120 m2 area.

times energy α−1

times energy

15 x 10 EMA−SIC Li et al.

α−1

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4 19

19

2 x 10 EMA−SIC Li et al. 1.5 1

times energy

times energy

(b) β = 2 18

Latency

0 100 200 300 400 500 600 700 800 900 1000

10

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

(d) β = 4

Latency

times energy α−1

Latency

5

4

Number of nodes

9

10

8 x 10 EMA−SIC Li et al. 6

α−1

2

(c) β = 3 15 x 10 EMA−SIC Li et al.

times energy α−1

4

Latency

times energy

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

0.5

4

18

9

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

Latency

Latency

α−1

times energy

9

1

Number of nodes

(a) β = 1 2 x 10 EMA−SIC Li et al. 1.5

2

18

6 x 10 EMA−SIC Li et al.

Latency

α−1 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

α−1

2

3 x 10 EMA−SIC Li et al.

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy

10

Latency

times energy α−1

Latency

4

8

15 x 10 EMA−SIC Li et al.

times energy

18

8

6 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 243: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 120 × 120 m2 area.

79

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0.5 0 100 200 300 400 500 600 700 800 900 1000

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

1

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 245: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 140 × 140 m2 area.

times energy α−1

times energy α−1

Latency

times energy α−1

Latency

times energy α−1

2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 9

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

14

times energy

times energy α−1

Latency

2

3

4 x 10 EMA−SIC Li et al. 3

(d) β = 4

14

5 x 10 EMA−SIC 4 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

9

Number of nodes

(c) β = 3

5

α−1

1

10

(a) β = 1 times energy

2 x 10 EMA−SIC Li et al. 1.5

8

15 x 10 EMA−SIC Li et al.

Number of nodes

α−1

times energy α−1

Latency

5

2

0 100 200 300 400 500 600 700 800 900 1000

14

Latency

times energy α−1

10

4

(b) β = 2

13

15 x 10 EMA−SIC Li et al.

(f) β = 6

8

6 x 10 EMA−SIC Li et al.

Number of nodes

(a) β = 1

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

times energy

2

2

Fig. 246: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 140 × 140 m2 area.

α−1

4

3 x 10 EMA−SIC Li et al.

(e) β = 5

times energy

8 x 10 EMA−SIC Li et al. 6

(d) β = 4

Number of nodes

10

α−1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

1

0.5

13

Latency

Latency

α−1

times energy

13

2

1

(f) β = 6

Fig. 244: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 140 × 140 m2 area.

4 x 10 EMA−SIC Li et al. 3

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

(e) β = 5

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Li et al. 1.5 1

(d) β = 4 10

times energy

1

5

19

19

α−1

2

10

(c) β = 3 times energy

4 x 10 EMA−SIC Li et al. 3

(b) β = 2 18

15 x 10 EMA−SIC Li et al.

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Latency

α−1

0.5

2

10

Latency

times energy Latency

α−1

1

4

(d) β = 4 times energy

10

2 x 10 EMA−SIC Li et al. 1.5

8 x 10 EMA−SIC Li et al. 6

Number of nodes

(c) β = 3

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy α−1

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

4

2

18

Latency

times energy α−1

1

Latency

times energy α−1

Latency

2

4

(a) β = 1

9

8 x 10 EMA−SIC Li et al. 6

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

9

3 x 10 EMA−SIC Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

2

18

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

Latency

α−1

5

Latency

0 100 200 300 400 500 600 700 800 900 1000

times energy

times energy α−1

2

Latency

10

Latency

times energy α−1

4

18

8

15 x 10 EMA−SIC Li et al.

Latency

8

6 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 247: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 160 × 160 m2 area.

80

13

times energy

4

α−1

Latency

α−1

2

13

8 x 10 EMA−SIC Li et al. 6

1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

4 x 10 EMA−SIC Li et al. 3

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

13

5

Latency

times energy

1.5

α−1

α−1

10

2.5 x 10 EMA−SIC 2 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

1

Latency

times energy

15 x 10 EMA−SIC Li et al.

14

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

14

2

Latency

times energy

6

α−1

α−1

4

14

10 x 10 EMA−SIC 8 Li et al.

0 100 200 300 400 500 600 700 800 900 1000

4

Latency

times energy

6 x 10 EMA−SIC Li et al.

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 248: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 160 × 160 m2 area.

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 times energy

15 x 10 EMA−SIC Li et al. 10

α−1

4

18

2

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

Latency

α−1

times energy

18

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

α−1

1

times energy

2 x 10 EMA−SIC Li et al. 1.5

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy α−1

Latency

(d) β = 4 19

19

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 249: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 160 × 160 m2 area.

81

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

Latency

(b) β = 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 250: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 180 × 180 m2 area.

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0.5 0 100 200 300 400 500 600 700 800 900 1000

times energy

14

times energy

4

1

Number of nodes

14

6 x 10 EMA−SIC Li et al.

2 x 10 EMA−SIC Li et al. 1.5

(d) β = 4

α−1

times energy

2

14

α−1

(c) β = 3

3

2

Number of nodes

Latency

times energy

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

α−1

1 0.5

5

Number of nodes

10

Latency

times energy α−1

Latency

1.5

10

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

4

0 100 200 300 400 500 600 700 800 900 1000

Latency

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3 10

15 x 10 EMA−SIC Li et al.

α−1

2

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

Latency

4

Number of nodes

2.5 x 10 EMA−SIC 2 Li et al.

1

Number of nodes

times energy

times energy

6

α−1

0 100 200 300 400 500 600 700 800 900 1000

13

13

9

10 x 10 EMA−SIC 8 Li et al.

Latency

times energy α−1

Latency

1

2

(b) β = 2

9

2

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 4 x 10 EMA−SIC Li et al. 3

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

0.5

Latency

2

13

Latency

times energy

1

α−1

4

9

2 x 10 EMA−SIC Li et al. 1.5

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 251: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 180 × 180 m2 area.

82

0 100 200 300 400 500 600 700 800 900 1000

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

Latency

times energy

times energy α−1

Latency times energy

times energy

α−1

2

α−1

2

0 100 200 300 400 500 600 700 800 900 1000

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

5 x 10 EMA−SIC 4 Li et al.

2

3

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 253: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution in a 200 × 200 m2 area.

18

times energy

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

10

times energy

1.5

4

(d) β = 4

α−1

2.5 x 10 EMA−SIC 2 Li et al.

6 x 10 EMA−SIC Li et al.

Number of nodes

Latency

times energy

10

18

4 x 10 EMA−SIC Li et al. 3

α−1

5

(c) β = 3

(f) β = 6

(a) β = 1

α−1

10

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

times energy

times energy

15 x 10 EMA−SIC Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

18

9

Latency

times energy α−1

Latency

1

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

9

2

0.5

Number of nodes

(a) β = 1 4 x 10 EMA−SIC Li et al. 3

1

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 x 10 EMA−SIC Li et al. 1.5

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

(d) β = 4 19

19

α−1

2

4

Fig. 254: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution in a 200 × 200 m2 area.

Latency

times energy α−1

10

6

(e) β = 5

18

8

15 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

8

14

10 x 10 EMA−SIC 8 Li et al.

Number of nodes

(f) β = 6

Fig. 252: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 180 × 180 m2 area.

4

1

0 100 200 300 400 500 600 700 800 900 1000

times energy

(e) β = 5

6 x 10 EMA−SIC Li et al.

3

Number of nodes

Number of nodes

Number of nodes

(d) β = 4

Latency

0 100 200 300 400 500 600 700 800 900 1000

α−1

1

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

14

5 x 10 EMA−SIC 4 Li et al.

Latency

0.5

3 x 10 EMA−SIC Li et al.

0.5

(c) β = 3

Latency

times energy α−1

α−1

1

Latency

(d) β = 4

Latency

times energy

2 x 10 EMA−SIC Li et al. 1.5

1

Number of nodes

19

19

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

Latency

5

Latency

Number of nodes

10

times energy

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2 14

2 x 10 EMA−SIC Li et al. 1.5

2 x 10 EMA−SIC Li et al. 1.5 1

times energy

0 100 200 300 400 500 600 700 800 900 1000

5

Number of nodes

13

15 x 10 EMA−SIC Li et al.

α−1

2

2

(a) β = 1

α−1

times energy α−1

10

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

18

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

Latency

4

1

(b) β = 2

18

6 x 10 EMA−SIC Li et al.

13

8 x 10 EMA−SIC Li et al. 6

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

(a) β = 1

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

3 x 10 EMA−SIC Li et al.

Latency

0 100 200 300 400 500 600 700 800 900 1000

Latency

1

times energy

0 100 200 300 400 500 600 700 800 900 1000

Latency

0.5

α−1

α−1

1

Latency

2

Latency

times energy α−1

1.5

13

18

4 x 10 EMA−SIC Li et al. 3

times energy

18

2.5 x 10 EMA−SIC 2 Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 255: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution in a 200 × 200 m2 area.

83

2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 256: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 100 × 100 m2 area.

times energy Latency

α−1

(c) β = 3

0.5

times energy

1

α−1

(d) β = 4

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(d) β = 4

14

2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

14

times energy

3 x 10 EMA−SIC Li et al.

5 x 10 EMA−SIC 4 Li et al.

α−1

1

(b) β = 2 14

2 x 10 EMA−SIC Li et al. 1.5

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 x 10 EMA−SIC Li et al. 1.5

2

Number of nodes

2

Latency

times energy α−1

4

5

Number of nodes

10

Latency

Latency

α−1

times energy

9

6

times energy α−1

(c) β = 3 10 x 10 EMA−SIC 8 Li et al.

times energy

times energy

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

3

4

0 100 200 300 400 500 600 700 800 900 1000

13

15 x 10 EMA−SIC Li et al.

times energy

0.5

α−1

1

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

9

5 x 10 EMA−SIC 4 Li et al.

13

Number of nodes

(b) β = 2

Latency

Latency

α−1

times energy

9

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 2 x 10 EMA−SIC Li et al. 1.5

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

2

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

10

α−1

4

8

15 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

3

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 257: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 100 × 100 m2 area.

84

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 259: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 120 × 120 m2 area.

times energy α−1

Latency

times energy times energy

times energy α−1

times energy α−1

Latency times energy

α−1

8 x 10 EMA−SIC Li et al. 6 4 2

0 100 200 300 400 500 600 700 800 900 1000

Latency

18

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3 times energy

(d) β = 4 19

19

α−1

5

2

(b) β = 2

18

Latency

α−1

10

Number of nodes

(a) β = 1

(d) β = 4 3 x 10 EMA−SIC Li et al.

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

times energy

15 x 10 EMA−SIC Li et al.

4

times energy

4

Latency

times energy

1

10

9

α−1

2

18

6 x 10 EMA−SIC Li et al.

α−1

8 x 10 EMA−SIC Li et al. 6

(c) β = 3

Latency

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

9

Number of nodes

(f) β = 6

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Fig. 260: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 120 × 120 m2 area.

α−1

1 0.5

2

(e) β = 5

(b) β = 2

α−1

1.5

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

times energy

2.5 x 10 EMA−SIC 2 Li et al.

8 x 10 EMA−SIC Li et al. 6

2 x 10 EMA−SIC Li et al. 1.5 1

times energy

0 100 200 300 400 500 600 700 800 900 1000

Latency

Latency

α−1

times energy

9

1

0 100 200 300 400 500 600 700 800 900 1000

α−1

5

(a) β = 1

14

α−1

times energy

10

Number of nodes

Number of nodes

(d) β = 4

18

8

15 x 10 EMA−SIC Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

3

(f) β = 6

Latency

times energy α−1

Latency

2

5 x 10 EMA−SIC 4 Li et al.

Number of nodes

8

4

times energy

1

Fig. 258: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 100 × 100 m2 area.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3

α−1

2

(e) β = 5

1 0.5

14

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

times energy

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

0.5

5

(d) β = 4

Latency

times energy Latency

α−1

1

10

Number of nodes

19

19

(b) β = 2 14

Latency

5

(c) β = 3 2 x 10 EMA−SIC Li et al. 1.5

15 x 10 EMA−SIC Li et al.

α−1

10

Number of nodes

Number of nodes

(a) β = 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

times energy

15 x 10 EMA−SIC Li et al.

α−1

2

4

13

18

Latency

times energy α−1

Latency

4

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

18

8 x 10 EMA−SIC Li et al. 6

1

Number of nodes

(a) β = 1

13

8 x 10 EMA−SIC Li et al. 6

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

Number of nodes

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

1

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 261: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 120 × 120 m2 area.

85

Number of nodes

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

times energy α−1

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 263: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 140 × 140 m2 area.

times energy α−1

Latency

times energy α−1

Latency

times energy α−1

Latency

times energy

times energy

times energy Latency

α−1

2 1

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 9

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

14

Latency

times energy α−1

Latency

4

4 x 10 EMA−SIC Li et al. 3

(d) β = 4 10 x 10 EMA−SIC 8 Li et al.

0.5

9

Number of nodes

14

1

α−1

1.5

(c) β = 3 6 x 10 EMA−SIC Li et al.

2

(a) β = 1

10

α−1

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Li et al.

9

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

α−1

times energy α−1

Latency

5

4

0 100 200 300 400 500 600 700 800 900 1000

14

Latency

times energy α−1

10

6 x 10 EMA−SIC Li et al.

(b) β = 2

13

15 x 10 EMA−SIC Li et al.

8

Number of nodes

(a) β = 1

(f) β = 6

times energy

Number of nodes

1

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

2.5 x 10 EMA−SIC 2 Li et al. 1.5

(d) β = 4 10

times energy

2

(d) β = 4 19

4 x 10 EMA−SIC Li et al. 3

Fig. 264: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 140 × 140 m2 area.

α−1

4

0 100 200 300 400 500 600 700 800 900 1000

(e) β = 5

Latency

times energy α−1

8 x 10 EMA−SIC Li et al. 6

5

Number of nodes

Number of nodes

13

Latency

Latency

α−1

times energy

13

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

(f) β = 6

Fig. 262: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 140 × 140 m2 area.

1

0.5

Number of nodes

(e) β = 5

2

1

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

4 x 10 EMA−SIC Li et al. 3

19

2 x 10 EMA−SIC Li et al. 1.5

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

10

(c) β = 3

Latency

α−1

0.5

4 x 10 EMA−SIC Li et al. 3

(b) β = 2 18

15 x 10 EMA−SIC Li et al.

Number of nodes

10

Latency

times energy Latency

α−1

1

2

(d) β = 4 times energy

10

2 x 10 EMA−SIC Li et al. 1.5

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

Number of nodes

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

2

Number of nodes

times energy α−1

4

4

0 100 200 300 400 500 600 700 800 900 1000

18

Latency

0 100 200 300 400 500 600 700 800 900 1000

1

6 x 10 EMA−SIC Li et al.

(a) β = 1

Latency

times energy α−1

6

18

Number of nodes

9

10 x 10 EMA−SIC 8 Li et al.

Latency

times energy α−1

Latency

1

2

(b) β = 2

9

2

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 4 x 10 EMA−SIC Li et al. 3

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

2

Latency

times energy

10

α−1

4

18

8

15 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 265: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 160 × 160 m2 area.

86

13

times energy

4

α−1

Latency

α−1

2

13

8 x 10 EMA−SIC Li et al. 6

1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

4 x 10 EMA−SIC Li et al. 3

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1 times energy α−1

α−1

10

5

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

15 x 10 EMA−SIC Li et al.

Latency

(b) β = 2 14

13

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

14

10

2

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

Latency

α−1

α−1

4

14

15 x 10 EMA−SIC Li et al.

times energy

times energy

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 266: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 160 × 160 m2 area.

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

10

α−1

4

15 x 10 EMA−SIC Li et al.

times energy

8 x 10 EMA−SIC Li et al. 6

18

2

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

Latency

α−1

times energy

18

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

α−1

1

times energy

2 x 10 EMA−SIC Li et al. 1.5

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy α−1

Latency

(d) β = 4 19

19

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 267: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 160 × 160 m2 area.

87

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

Latency

(b) β = 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 268: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 180 × 180 m2 area.

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

times energy

14

times energy

4

1.5

Number of nodes

14

8 x 10 EMA−SIC Li et al. 6

2.5 x 10 EMA−SIC 2 Li et al.

(d) β = 4

α−1

times energy

2

14

α−1

(c) β = 3

3

2

Number of nodes

Latency

times energy

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

α−1

1 0.5

5

Number of nodes

10

Latency

times energy α−1

Latency

1.5

10

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

4

0 100 200 300 400 500 600 700 800 900 1000

Latency

(c) β = 3 10

15 x 10 EMA−SIC Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

Latency

5

Number of nodes

2.5 x 10 EMA−SIC 2 Li et al.

1

Number of nodes

times energy

times energy

10

α−1

0 100 200 300 400 500 600 700 800 900 1000

13

13

9

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

Latency

1

2

(b) β = 2

9

2

4 x 10 EMA−SIC Li et al. 3

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1 4 x 10 EMA−SIC Li et al. 3

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

0.5

Latency

2

13

Latency

times energy

1

α−1

4

9

2 x 10 EMA−SIC Li et al. 1.5

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 269: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 180 × 180 m2 area.

88

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

Latency

times energy

times energy α−1

times energy

times energy

α−1

times energy

times energy

6 x 10 EMA−SIC Li et al. 4

α−1

1

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 271: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution in a 200 × 200 m2 area.

Number of nodes

(c) β = 3

(d) β = 4 19

19

α−1

1

0 100 200 300 400 500 600 700 800 900 1000

2 x 10 EMA−SIC Li et al. 1.5 1

times energy

times energy

2

5

Number of nodes

10

3

10

Latency

2

(d) β = 4 5 x 10 EMA−SIC 4 Li et al.

15 x 10 EMA−SIC Li et al.

α−1

4

18

times energy

8 x 10 EMA−SIC Li et al. 6

Number of nodes

α−1

1

2

18

α−1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

Latency

2

(f) β = 6

Number of nodes

Latency

5

10

3 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

α−1

10

(c) β = 3

Number of nodes

Fig. 272: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution in a 200 × 200 m2 area.

times energy

15 x 10 EMA−SIC Li et al.

Number of nodes

0 100 200 300 400 500 600 700 800 900 1000

18

9

α−1

0 100 200 300 400 500 600 700 800 900 1000

5

(e) β = 5

(b) β = 2

Latency

Latency

1

10

Number of nodes

Number of nodes

times energy

times energy α−1

2

14

15 x 10 EMA−SIC Li et al.

Latency

0 100 200 300 400 500 600 700 800 900 1000

9

3

2

0 100 200 300 400 500 600 700 800 900 1000

α−1

0.5

(a) β = 1 5 x 10 EMA−SIC 4 Li et al.

4

Latency

times energy

1

Number of nodes

Number of nodes

(d) β = 4

18

9

2 x 10 EMA−SIC Li et al. 1.5

α−1

0 100 200 300 400 500 600 700 800 900 1000

6 x 10 EMA−SIC Li et al.

(f) β = 6

Latency

times energy α−1

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

Latency

1

Fig. 270: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 180 × 180 m2 area.

4

1 0.5

(c) β = 3

Number of nodes

8

1.5

Number of nodes

α−1

2

(e) β = 5

2.5 x 10 EMA−SIC 2 Li et al.

14

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

5

(d) β = 4

Latency

times energy Latency

α−1

1 0.5

10

Number of nodes

19

19

(b) β = 2 14

Latency

0 100 200 300 400 500 600 700 800 900 1000

(c) β = 3 2 x 10 EMA−SIC Li et al. 1.5

15 x 10 EMA−SIC Li et al.

Latency

5

Number of nodes

Number of nodes

(a) β = 1

α−1

10

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

15 x 10 EMA−SIC Li et al.

Latency

Latency

2

4

13

18

times energy

times energy α−1

4

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

18

8 x 10 EMA−SIC Li et al. 6

1

Number of nodes

(a) β = 1

13

8 x 10 EMA−SIC Li et al. 6

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

Number of nodes

2

α−1

0 100 200 300 400 500 600 700 800 900 1000

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

1

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 273: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution in a 200 × 200 m2 area.

89

0 100 200 300 400 500 600 700 800 900 1000

α−1

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

times energy

times energy α−1

Latency

α−1

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

(d) β = 4

14

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 274: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 100 × 100 m2 area.

3 x 10 EMA−SIC Li et al. 2

14

times energy

9

times energy

times energy α−1

1

(b) β = 2 14

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

(d) β = 4

9

2

0 100 200 300 400 500 600 700 800 900 1000

times energy

(c) β = 3

3

5

Number of nodes

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

times energy α−1

1

α−1

0 100 200 300 400 500 600 700 800 900 1000

10

Latency

0.5

2

2

Number of nodes

13

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

α−1

1

Latency

times energy

2 x 10 EMA−SIC Li et al. 1.5

3 x 10 EMA−SIC Li et al.

4

(a) β = 1

9

9

Latency

(b) β = 2

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

α−1

(a) β = 1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

1

Latency

5

13

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

α−1

2

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

10

α−1

4

8

15 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

8

6 x 10 EMA−SIC Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 275: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 100 × 100 m2 area.

90

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 277: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 120 × 120 m2 area.

times energy α−1

Latency

α−1

times energy α−1

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 15 x 10 EMA−SIC Li et al.

times energy

4

18

10

2

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3 times energy

(d) β = 4 19

19

α−1

times energy

times energy

times energy

times energy times energy

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Latency

0 100 200 300 400 500 600 700 800 900 1000

10

2

18

9

15 x 10 EMA−SIC Li et al.

4

α−1

1

6 x 10 EMA−SIC Li et al.

(a) β = 1

(d) β = 4

α−1

2

1

Number of nodes

Latency

times energy α−1

Latency

4

2

18

Number of nodes

α−1

2

0 100 200 300 400 500 600 700 800 900 1000

9

6 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

3 x 10 EMA−SIC Li et al.

(c) β = 3

(f) β = 6

Fig. 278: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 120 × 120 m2 area.

(b) β = 2

Number of nodes

Number of nodes

(e) β = 5

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy α−1

Latency

0.5

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

9

1

1

0 100 200 300 400 500 600 700 800 900 1000

α−1

5

(a) β = 1 2 x 10 EMA−SIC Li et al. 1.5

2

14

4 x 10 EMA−SIC Li et al. 3

2 x 10 EMA−SIC Li et al. 1.5 1

times energy

times energy

10

Number of nodes

9

3 x 10 EMA−SIC Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4

18

8

15 x 10 EMA−SIC Li et al.

α−1

2

Number of nodes

(c) β = 3

(f) β = 6

Latency

times energy α−1

Latency

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

8

1 0.5

Latency

1

Fig. 276: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 100 × 100 m2 area.

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

α−1

2

(e) β = 5

2 x 10 EMA−SIC Li et al. 1.5

14

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 14

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

times energy α−1

0.5

10

(d) β = 4

Latency

times energy Latency

α−1

1

15 x 10 EMA−SIC Li et al.

Number of nodes

19

19

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

Latency

5

(c) β = 3

2

(a) β = 1

α−1

10

Number of nodes

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Latency

times energy

15 x 10 EMA−SIC Li et al.

α−1

0 100 200 300 400 500 600 700 800 900 1000

13

8 x 10 EMA−SIC Li et al. 6

13

18

Latency

times energy α−1

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

18

4

1

Number of nodes

(a) β = 1 8 x 10 EMA−SIC Li et al. 6

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

1

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 279: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 120 × 120 m2 area.

91

times energy α−1

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

6 x 10 EMA−SIC Li et al. 4

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 281: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 140 × 140 m2 area.

times energy

times energy α−1

times energy α−1

times energy α−1

Latency

times energy α−1

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8 x 10 EMA−SIC Li et al. 6 4

α−1

1.5

(b) β = 2 9

times energy

2.5 x 10 EMA−SIC 2 Li et al.

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

14

Latency

times energy α−1

Latency

2

9

(d) β = 4

14

3 x 10 EMA−SIC Li et al.

5

(a) β = 1

Number of nodes

(c) β = 3

10

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

8

15 x 10 EMA−SIC Li et al.

Number of nodes

α−1 0.5

(f) β = 6

(d) β = 4 10

9

15 x 10 EMA−SIC Li et al. 10

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

times energy

times energy α−1

5

Latency

1

Latency

times energy α−1

10

4

14

2 x 10 EMA−SIC Li et al. 1.5

Number of nodes

8

6 x 10 EMA−SIC Li et al.

(b) β = 2

13

1

Fig. 282: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 140 × 140 m2 area.

Number of nodes

(a) β = 1

2

0 100 200 300 400 500 600 700 800 900 1000

α−1

2

(d) β = 4 3 x 10 EMA−SIC Li et al.

(e) β = 5

α−1

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

15 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

times energy

8 x 10 EMA−SIC Li et al. 6

Number of nodes

Number of nodes

α−1

0 100 200 300 400 500 600 700 800 900 1000

0.5

Latency

α−1

1

Latency

times energy Latency

α−1

2

1

13

times energy

13

4 x 10 EMA−SIC Li et al. 3

2 x 10 EMA−SIC Li et al. 1.5

(f) β = 6

Fig. 280: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 140 × 140 m2 area.

0 100 200 300 400 500 600 700 800 900 1000

19

19

Number of nodes

(e) β = 5

5

Latency

α−1 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10

(c) β = 3

α−1

0.5

(b) β = 2 18

15 x 10 EMA−SIC Li et al.

Number of nodes

Latency

times energy

1

α−1

0 100 200 300 400 500 600 700 800 900 1000

2

10

2 x 10 EMA−SIC Li et al. 1.5

Latency

times energy α−1

Latency

2

4

(d) β = 4

9

4

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(c) β = 3

6

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

10 x 10 EMA−SIC 8 Li et al.

times energy

times energy

1

Latency

0 100 200 300 400 500 600 700 800 900 1000

2

3

Number of nodes

18

times energy

0.5

α−1

times energy Latency

α−1

1

2

(a) β = 1

9

5 x 10 EMA−SIC 4 Li et al.

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

Latency

9

2 x 10 EMA−SIC Li et al. 1.5

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

2

18

6 x 10 EMA−SIC Li et al.

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

3 x 10 EMA−SIC Li et al.

Latency

α−1

5

Latency

0 100 200 300 400 500 600 700 800 900 1000

times energy

times energy α−1

2

Latency

10

Latency

times energy α−1

4

18

8

15 x 10 EMA−SIC Li et al.

Latency

8

6 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 283: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 160 × 160 m2 area.

92

13

times energy

4

α−1

Latency

α−1

2

13

8 x 10 EMA−SIC Li et al. 6

1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

4 x 10 EMA−SIC Li et al. 3

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2

13

5

Latency

times energy

1

α−1

α−1

10

2 x 10 EMA−SIC Li et al. 1.5

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

15 x 10 EMA−SIC Li et al.

14

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

14

times energy

4

α−1

Latency

α−1

2

14

8 x 10 EMA−SIC Li et al. 6

1

0 100 200 300 400 500 600 700 800 900 1000

Latency

times energy

4 x 10 EMA−SIC Li et al. 3

2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 284: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 160 × 160 m2 area.

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(a) β = 1

(b) β = 2 times energy

15 x 10 EMA−SIC Li et al. 10

α−1

4

18

2

0 100 200 300 400 500 600 700 800 900 1000

5

Latency

Latency

α−1

times energy

18

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

α−1

1

times energy

2 x 10 EMA−SIC Li et al. 1.5

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

times energy α−1

Latency

(d) β = 4 19

19

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 285: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 160 × 160 m2 area.

93

0 100 200 300 400 500 600 700 800 900 1000

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

times energy α−1

(b) β = 2

α−1

times energy

14

Number of nodes

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 286: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 180 × 180 m2 area.

1 0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

5 x 10 EMA−SIC 4 Li et al.

2

3

14

times energy

2

1.5

(d) β = 4

14

3 x 10 EMA−SIC Li et al.

2.5 x 10 EMA−SIC 2 Li et al.

(c) β = 3

α−1

times energy α−1

0.5

Latency

times energy α−1

Latency

1

0 100 200 300 400 500 600 700 800 900 1000

(d) β = 4 10

10

2 x 10 EMA−SIC Li et al. 1.5

5

Number of nodes

(c) β = 3

2

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy α−1

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

10

α−1

1

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

4

Latency

times energy α−1

Latency

2

8 x 10 EMA−SIC Li et al. 6

4

(a) β = 1 13

9

8 x 10 EMA−SIC Li et al. 6

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2

9

3 x 10 EMA−SIC Li et al.

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(a) β = 1

2

13

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

Number of nodes

4 x 10 EMA−SIC Li et al. 3

Latency

α−1

5

Latency

0 100 200 300 400 500 600 700 800 900 1000

13

times energy

times energy α−1

2

Latency

10

Latency

times energy α−1

4

8

15 x 10 EMA−SIC Li et al.

Latency

8

6 x 10 EMA−SIC Li et al.

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 287: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 180 × 180 m2 area.

94

times energy α−1

Latency

times energy α−1

(d) β = 4 14

10 x 10 EMA−SIC 8 Li et al.

times energy

4

2

0 100 200 300 400 500 600 700 800 900 1000

6 4 2 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(e) β = 5

(f) β = 6

Fig. 290: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution in a 200 × 200 m2 area.

Number of nodes

(e) β = 5

times energy α−1

Number of nodes

(b) β = 2

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 289: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution in a 200 × 200 m2 area.

times energy α−1

times energy α−1

1

10

5

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

Number of nodes

(c) β = 3

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

α−1

1

times energy

19

19

α−1

2

15 x 10 EMA−SIC Li et al.

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

10

4 x 10 EMA−SIC Li et al. 3

18

0.5 0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(e) β = 5

Latency

0 100 200 300 400 500 600 700 800 900 1000

4

(d) β = 4

Latency

0.5

Latency

0 100 200 300 400 500 600 700 800 900 1000

8 x 10 EMA−SIC Li et al. 6

Number of nodes

α−1

1

2

(a) β = 1

4 2

times energy

2 x 10 EMA−SIC Li et al. 1.5

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

6

10 x 10 EMA−SIC 8 Li et al.

(c) β = 3

18

6 x 10 EMA−SIC Li et al.

18

9

Number of nodes

10

0 100 200 300 400 500 600 700 800 900 1000

times energy

0 100 200 300 400 500 600 700 800 900 1000

1

(b) β = 2

Latency

1

2

Number of nodes

α−1

2

3 x 10 EMA−SIC Li et al.

Latency

0 100 200 300 400 500 600 700 800 900 1000

times energy

4 x 10 EMA−SIC Li et al. 3

times energy α−1

5

Latency

2

9

times energy

6 x 10 EMA−SIC Li et al.

Latency

times energy α−1

10

Latency

times energy α−1

Latency

4

8

15 x 10 EMA−SIC Li et al.

(a) β = 1

α−1

Number of nodes

(c) β = 3

(f) β = 6

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

18

6 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

times energy

1 0.5

Number of nodes

Number of nodes

8

α−1

1.5

α−1

1

Fig. 288: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 180 × 180 m2 area.

Latency

0 100 200 300 400 500 600 700 800 900 1000

α−1

times energy

2

(e) β = 5

2.5 x 10 EMA−SIC 2 Li et al.

14

3 x 10 EMA−SIC Li et al.

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(b) β = 2 14

Latency

0 100 200 300 400 500 600 700 800 900 1000

5

(d) β = 4

α−1

0.5

10

Number of nodes

Latency

times energy Latency

α−1

1

Number of nodes

Latency

0 100 200 300 400 500 600 700 800 900 1000

19

19

2 x 10 EMA−SIC Li et al. 1.5

15 x 10 EMA−SIC Li et al.

α−1

5

(c) β = 3

2

(a) β = 1

Latency

times energy

10

Number of nodes

4

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

times energy

18

α−1

0 100 200 300 400 500 600 700 800 900 1000

13

8 x 10 EMA−SIC Li et al. 6

13

15 x 10 EMA−SIC Li et al.

Latency

times energy α−1

Latency

2

0 100 200 300 400 500 600 700 800 900 1000

(b) β = 2

18

4

1

Number of nodes

(a) β = 1 8 x 10 EMA−SIC Li et al. 6

times energy α−1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

2

times energy

0 100 200 300 400 500 600 700 800 900 1000

2

Latency

1

13

4 x 10 EMA−SIC Li et al. 3

Latency

times energy

4

α−1

2

18

6 x 10 EMA−SIC Li et al.

Latency

Latency

α−1

times energy

18

3 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100 200 300 400 500 600 700 800 900 1000

Number of nodes

(f) β = 6

Fig. 291: A separate comparison of Latency-energy tradeoff between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution in a 200 × 200 m2 area.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 293: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 294: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 292: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

95

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 295: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 297: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 298: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 296: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

96

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 299: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

97

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 300: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

160

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

Network scale

180

800

(d) β = 4

800

(e) β = 5

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 301: Impact of network scale on the aggregation latency under α = 3 and uniform distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 303: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 304: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 302: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

98

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 305: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 307: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 308: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 306: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

99

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 309: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

100

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 310: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 311: Impact of network scale on the aggregation latency under α = 4 and uniform distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 313: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 314: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 312: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

101

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 315: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 317: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 318: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 316: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

102

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 319: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

103

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 320: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 321: Impact of network scale on the aggregation latency under α = 5 and uniform distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 323: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 324: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 322: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

104

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 325: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 327: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 328: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 326: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

105

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 329: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

106

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 330: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 331: Impact of network scale on the aggregation latency under α = 3 and poisson distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 333: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 334: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 332: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

107

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 335: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 337: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 338: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 336: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

108

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 339: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

109

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 340: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 341: Impact of network scale on the aggregation latency under α = 4 and poisson distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 343: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 344: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 342: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

110

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 345: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 400 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

180

200

Aggregation latency

Aggregation latency

600

160

(e) β = 5

160

180

180

200

600

160

Network scale

180

200

(f) β = 6

Fig. 347: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 600 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

200

(d) β = 4

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

1000

(a) β = 1

600

Network scale

200

600

(d) β = 4

800

Network scale

200

800

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Fig. 348: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 700 nodes.

Aggregation latency

Aggregation latency

Aggregation latency

800

160

1000

(b) β = 2

1000

Network scale

160

Network scale

(a) β = 1

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

160

Network scale

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

Network scale

200

Fig. 346: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 500 nodes.

800

600

(b) β = 2

800

(f) β = 6

1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

1000

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

Aggregation latency

Aggregation latency

800

Network scale

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

(a) β = 1

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

111

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 349: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

112

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 350: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 351: Impact of network scale on the aggregation latency under α = 5 and poisson distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 353: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 354: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 352: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

113

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 355: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 400 nodes.

180

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

600

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

800

600

160

Network scale

(f) β = 6

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

800

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

114

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

160

180

200

180

200

800

600

160

Network scale

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 357: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 600 nodes.

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

600

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

800

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Fig. 358: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 700 nodes.

Aggregation latency

Fig. 356: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 500 nodes.

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 359: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

115

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 360: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 361: Impact of network scale on the aggregation latency under α = 3 and cluster distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 363: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 364: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 362: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

116

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 365: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 400 nodes.

180

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

600

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

800

600

160

Network scale

(f) β = 6

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

800

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

117

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

160

180

200

180

200

800

600

160

Network scale

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 367: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 600 nodes.

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

600

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

800

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Fig. 368: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 700 nodes.

Aggregation latency

Fig. 366: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 500 nodes.

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 369: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

118

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 370: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 371: Impact of network scale on the aggregation latency under α = 4 and cluster distribution with 1000 nodes.

600

160

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

600

160

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

180

200

800

600

600

160

Network scale

180

200

600

Aggregation latency

Aggregation latency

800

600

160

(e) β = 5

160

Network scale

180

180

200

600

160

180

200

(f) β = 6

Fig. 373: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 200 nodes.

180

200

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Aggregation latency

Network scale

1000

800

600

400 100

200

EMA−SIC Cell−AS Li et al. 120 140

160

Network scale

(f) β = 6

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

800

Network scale

180

1000

Network scale

200

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

200

(d) β = 4

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(a) β = 1

800

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

Aggregation latency

800

180

1000

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

160

160

Network scale

(e) β = 5

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

EMA−SIC Cell−AS Li et al. 400 100 120 140

Fig. 374: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 300 nodes.

Aggregation latency

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Network scale

Aggregation latency

Aggregation latency

Aggregation latency

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

600

(f) β = 6

800

600

(b) β = 2

800

Network scale

200

Fig. 372: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 100 nodes. 1000

200

800

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

1000

(d) β = 4

800

Network scale

160

Network scale

160

Network scale

1000

(a) β = 1

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

200

Aggregation latency

800

Network scale

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

200

600

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

119

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 375: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 400 nodes.

180

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

600

180

200

180

200

800

600

160

Network scale

(e) β = 5

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

800

600

160

Network scale

(f) β = 6

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

800

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

800

1000

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Aggregation latency

120

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

200

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

(e) β = 5

160

180

200

180

200

800

600

160

Network scale

180

200

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 377: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 600 nodes.

800

600

160

Network scale

(e) β = 5

180

200

180

200

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(d) β = 4

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(b) β = 2

(c) β = 3

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

1000

(a) β = 1

800

Network scale

Aggregation latency

Aggregation latency

600

160

EMA−SIC Cell−AS Li et al. 400 100 120 140

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

600

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

800

Aggregation latency

180

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

160

Network scale

600

Aggregation latency

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

Aggregation latency

600

Aggregation latency

800

1000

Aggregation latency

1000

Aggregation latency

Fig. 378: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 700 nodes.

Aggregation latency

Fig. 376: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 500 nodes.

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 379: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 800 nodes.

1000

Aggregation latency

Aggregation latency

121

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

Aggregation latency

Aggregation latency

800

600

160

Network scale

180

200

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

800

600

160

180

800

(d) β = 4

1000

Network scale

200

1000

(c) β = 3

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2

1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(a) β = 1

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(e) β = 5

(f) β = 6

1000

Aggregation latency

Aggregation latency

Fig. 380: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 900 nodes.

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

800

600

160

Network scale

180

200

(e) β = 5

200

180

200

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

Aggregation latency

Aggregation latency

600

160

180

800

(d) β = 4

800

Network scale

200

1000

(c) β = 3 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

180

(b) β = 2 Aggregation latency

Aggregation latency

(a) β = 1 1000

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

180

200

1000

800

600

EMA−SIC Cell−AS Li et al. 400 100 120 140

160

Network scale

(f) β = 6

Fig. 381: Impact of network scale on the aggregation latency under α = 5 and cluster distribution with 1000 nodes.

122

4

120

140

160

Network scale

180

0 100

200

120

(a) β = 1

2000 0 100

200

140

160

180

0 100

200

120

Energy consumption

1.5 1 0.5

140

160

Network scale

180

200

120

140

160

Network scale

(e) β = 5

180

200

120

120

1

140

160

Network scale

180

200

(f) β = 6

140

160

Network scale

180

Fig. 382: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 100 nodes.

200

180

200

180

200

1 0.5 0 100

120

140

160

Network scale

(d) β = 4 5

2 1

120

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

(b) β = 2

2

0 100

140

Network scale

5

5

2

120

0 100

200

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

180

4

5

5

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

140

Network scale

1

(a) β = 1 Energy consumption

2

Network scale

120

3 x 10 EMA−SIC Cell−AS Li et al. 2

4

5

(c) β = 3 Energy consumption

4000

4

4

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

120

160

6000

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Energy consumption

0.5

8000

Energy consumption

0 100

1

EMA−SIC Cell−AS Li et al.

Energy consumption

1000

10000

Energy consumption

2000

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

3000

140

160

Network scale

(e) β = 5

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 383: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 200 nodes.

123

120

140

160

Network scale

180

0 100

200

120

(a) β = 1

2

140

160

Network scale

180

0 100

120

(c) β = 3 Energy consumption

Energy consumption

1 140

160

180

140

160

Network scale

180

120

2

120

140

160

Network scale

180

140

160

Network scale

180

0 100

140

160

Network scale

180

180

0 100

120

(a) β = 1 Energy consumption

Energy consumption 140

160

Network scale

180

1 0.5 120

Energy consumption 140

160

(e) β = 5

180

200

140

160

Network scale

180

140

160

Network scale

180

2

0 100

200

5

0 100

4 2 140

160

Network scale

180

200

120

140

160

Network scale

180

0 100

200

(f) β = 6

Fig. 385: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 400 nodes.

180

200

120

160

180

200

180

200

(d) β = 4

4 2

120

140

Network scale

5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

1

(c) β = 3

6

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

5

120

120

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

160

5

(d) β = 4

2

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1

5

Network scale

120

4

(c) β = 3 Energy consumption

0 100

200

1.5

5

120

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

160

5000

(b) β = 2

2

120

140

Network scale

EMA−SIC Cell−AS Li et al.

10000

5

4

120

(f) β = 6

Energy consumption

1

4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

15000

2

0 100

200

200

5

Energy consumption

160

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

Energy consumption

140

Network scale

160

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

120

140

Network scale

Fig. 386: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 500 nodes.

Energy consumption

0 100

120

(e) β = 5

Energy consumption

Energy consumption

Energy consumption

5000

200

1 0.5

4

10000

180

1.5

(d) β = 4

2

(f) β = 6

200

5

4

120

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

160

5

2 0 100

140

Network scale

(b) β = 2

(c) β = 3

4

(e) β = 5

EMA−SIC Cell−AS Li et al.

120

5

Fig. 384: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 300 nodes.

15000

2

0 100

200

4

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

180

6

(d) β = 4

2

160

10 x 10 EMA−SIC 8 Cell−AS Li et al.

5

3

140

Network scale

4

0.5

5

Network scale

120

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1

1

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

0 100

200

5

4

0 100

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

4

120

160

5000

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Energy consumption

1

10000

Energy consumption

0 100

2

EMA−SIC Cell−AS Li et al.

Energy consumption

5000

15000

Energy consumption

10000

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

4

15000

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 387: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 600 nodes.

124

4

180

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

1

120

2 120

140

160

Network scale

180

140

160

Network scale

180

200

120

140

160

Network scale

180

0.5

140

160

Network scale

180

200

120

(a) β = 1

160

180

5

140

160

Network scale

180

120

(c) β = 3

160

180

0 100

2 140

160

Network scale

(e) β = 5

180

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 389: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 800 nodes.

180

200

Energy consumption

1

0 100

200

120

140

160

Network scale

(f) β = 6

120

140

160

Network scale

180

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

200

120

180

200

180

200

180

200

(b) β = 2

120

140

160

Network scale

180

2 1

0 100

200

120

160

6

4 2

120

140

Network scale

(d) β = 4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

5

5

0 100

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

0.5

200

0.5

5

1

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

160

0.5

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

160

4

(d) β = 4

6

140

Network scale

1

6

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

140

Network scale

140

Network scale

6

(a) β = 1

1

0 100

200

120

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

4

2

180

1

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

200

Fig. 390: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 900 nodes.

Energy consumption

Energy consumption

Energy consumption

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

(b) β = 2

5

Energy consumption

140

Network scale

180

2

0 100

200

2

5

4

180

4

2

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(e) β = 5

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

160

4

Energy consumption

Energy consumption

Energy consumption

1

140

Network scale

6

4

120

120

0 100

140

Network scale

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

Fig. 388: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 700 nodes.

4

120

5

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

200

5

0.5 0 100

2

(c) β = 3

1

(e) β = 5

0 100

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

4

180

5

6

6

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

5

140

Network scale

4

2

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

120

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

5

0 100

0 100

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3 Energy consumption

180

5

4

0 100

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Energy consumption

160

Energy consumption

140

Network scale

1 0.5

Energy consumption

120

2

Energy consumption

0 100

4

Energy consumption

0.5

4

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

Energy consumption

Energy consumption

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 391: Impact of network scale on the energy consumption (joule) under α = 3 and uniform distribution with 1000 nodes.

125

1 120

140

160

Network scale

180

5

0 100

200

120

(a) β = 1

0 100

120

140

160

Network scale

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

200

120

140

160

Network scale

180

4 2 120

(c) β = 3

140

160

Network scale

180

200

2

120

120

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

0 100

200

6 4 2 0 100

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 392: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 100 nodes.

1

120

160

180

200

180

200

(d) β = 4

2

0 100

140

Network scale

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

Energy consumption

1

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

5

7

3 x 10 EMA−SIC Cell−AS Li et al. 2

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

(d) β = 4 7

160

6

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

(b) β = 2 Energy consumption

Energy consumption

6

0 100

200

120

(a) β = 1 6

Energy consumption

Energy consumption

2

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

7

Energy consumption

180

4

6

1

0 100

160

6

(b) β = 2

6

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

6

Energy consumption

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

3

0 100

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

5

Energy consumption

Energy consumption

5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 393: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 200 nodes.

126

6

120

140

160

Network scale

180

1

0 100

200

120

(a) β = 1

4 2 120

140

160

Network scale

180

120

1 120

140

160

Network scale

180

140

160

Network scale

180

0 100

120

(e) β = 5

140

160

Network scale

180

120

140

160

Network scale

180

160

180

4 2

120

(a) β = 1

140

160

Network scale

180

120

140

160

Network scale

180

0 100

200

0.5

140

160

Network scale

180

0 100

200

120

(c) β = 3 7

160

Network scale

180

Energy consumption 140

160

Network scale

(e) β = 5

180

200

120

180

140

160

Network scale

180

200

(f) β = 6

Fig. 395: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 400 nodes.

160

180

200

4 2

0 100

200

120

160

180

200

180

200

180

200

7

120

140

160

Network scale

180

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

200

120

160

8

6 4 2 120

140

Network scale

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

7

5

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

5

0 100

200

7

2

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

6 x 10 EMA−SIC Cell−AS Li et al. 4

120

140

Energy consumption

120

120

6

Energy consumption

Energy consumption

5

120

(a) β = 1

1

160

6

1

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

Network scale

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

7

6

120

Fig. 396: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 500 nodes.

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

1

5

Energy consumption

0 100

200

180

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

140

Network scale

200

7

Energy consumption

120

180

(d) β = 4

6

2

160

3

(e) β = 5

6 x 10 EMA−SIC Cell−AS Li et al. 4

140

Network scale

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

Energy consumption

Energy consumption

Energy consumption

5

120

(b) β = 2

6

5

Energy consumption

0 100

200

7

(f) β = 6

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2

7

5

200

4

(c) β = 3

Fig. 394: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 300 nodes.

0 100

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2

180

5

7

3

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

7

140

Network scale

6

1

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

120

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3

0 100

0 100

200

Energy consumption

Energy consumption

Energy consumption

6

Energy consumption

180

7

6

0 100

160

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

5

Energy consumption

0 100

2

Energy consumption

2

3

Energy consumption

4

6

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

Energy consumption

6

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 397: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 600 nodes.

127

6

180

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

2 1 120

140

160

Network scale

180

200

2 140

160

Network scale

180

0.5 0 100

120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

2

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

120

140

160

Network scale

180

Energy consumption 140

160

Network scale

(e) β = 5

180

200

1

120

140

160

Network scale

180

200

(f) β = 6

Fig. 399: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 800 nodes.

140

160

Network scale

200

Energy consumption

8

140

160

Network scale

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

200

120

140

160

Network scale

(f) β = 6

0.5

120

140

160

Network scale

180

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

200

120

160

180

200

180

200

180

200

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

120

140

160

Network scale

180

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

200

120

160

8

5

120

140

Network scale

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

(b) β = 2

7

0.5 0 100

120

(c) β = 3

8

5

180

1

(d) β = 4

1

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

2

0 100

200

6

1

180

3

(a) β = 1

2

200

6

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

0 100

200

3

0 100

200

7

Energy consumption

180

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

0 100

160

Energy consumption

0 100

140

Network scale

Energy consumption

Energy consumption

Energy consumption

5

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

7

6

160

5

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

6

4

180

Fig. 400: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 900 nodes.

Energy consumption

Energy consumption

Energy consumption

0.5

120

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

6

1

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

140

Network scale

(b) β = 2

0.5

0 100

200

Fig. 398: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 700 nodes.

6

120

7

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2

0 100

200

7

1

200

4

(c) β = 3

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

4

180

1

0 100

8

6

160

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(d) β = 4

7

140

Network scale

7

3

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

120

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

5

0 100

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3 Energy consumption

180

7

6

0 100

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Energy consumption

160

0.5

Energy consumption

140

Network scale

1

Energy consumption

120

2

Energy consumption

0 100

4

Energy consumption

0.5

6

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

Energy consumption

Energy consumption

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 401: Impact of network scale on the energy consumption (joule) under α = 4 and uniform distribution with 1000 nodes.

128

8

120

140

160

Network scale

180

1

0 100

200

120

(a) β = 1

0 100

200

2 120

140

160

Network scale

120

180

1

120

(c) β = 3

160

180

140

160

Network scale

180

120

140

160

Network scale

180

120

140

160

Network scale

(e) β = 5

180

200

5

0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 402: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 100 nodes.

200

180

200

180

200

1 0.5 120

140

160

Network scale

(d) β = 4

6 4 2 120

180

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

160

1.5

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

200

Energy consumption

Energy consumption

2

120

(b) β = 2

9

4

2

9

5

0 100

200

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

4

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

2

0 100

200

140

Network scale

8

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

Energy consumption

4

Energy consumption

180

5

(b) β = 2

6

0 100

160

8

8

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

Energy consumption

0 100

2

Energy consumption

1

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2

8

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

Energy consumption

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 403: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 200 nodes.

129

120

(a) β = 1

120

140

160

Network scale

180

2 1 120

120

140

160

Network scale

180

160

180

120

(e) β = 5

140

160

Network scale

180

0.5

140

160

Network scale

180

2

0 100

200

120

160

180

140

160

Network scale

180

120

(c) β = 3 9

120

140

160

Network scale

180

160

(e) β = 5

180

200

1 0.5 0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 405: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 400 nodes.

160

180

200

180

200

10

140

160

Network scale

180

1.5 1 0.5 0 100

200

120

140

160

Network scale

(f) β = 6

8

140

160

Network scale

180

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

200

120

1

140

160

Network scale

180

200

180

200

180

200

2

120

140

160

Network scale

(d) β = 4 10

5

120

180

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

9

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

140

Network scale

(b) β = 2

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(c) β = 3 Energy consumption

Energy consumption 140

Network scale

120

0 100

200

10

5

120

(d) β = 4

1

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

200

9

2

0 100

200

180

(a) β = 1 Energy consumption

Energy consumption

5

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

200

Fig. 406: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 500 nodes.

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

(b) β = 2

8

Energy consumption

140

Network scale

180

2

8

4

160

6 x 10 EMA−SIC Cell−AS Li et al. 4

(e) β = 5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

Energy consumption

0 100

Energy consumption

Energy consumption

Energy consumption

1

140

Network scale

5

8

8

120

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

Fig. 404: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 300 nodes.

0 100

1

0 100

140

Network scale

9

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

(b) β = 2

9

0.5 0 100

2

(c) β = 3

1

200

4

0 100

200

0.5

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption 140

Network scale

180

1.5

10

5

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(d) β = 4

9

140

Network scale

9

3

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

120

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

5

0 100

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3 Energy consumption

180

9

8

0 100

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Energy consumption

0 100

200

Energy consumption

180

Energy consumption

160

1

Energy consumption

140

Network scale

2

8

3 x 10 EMA−SIC Cell−AS Li et al. 2

Energy consumption

120

4

Energy consumption

0 100

Energy consumption

Energy consumption

Energy consumption

5

8

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

8

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

(e) β = 5

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 407: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 600 nodes.

130

120

(a) β = 1

140

160

Network scale

180

2

120

160

180

160

180

120

140

160

Network scale

180

140

160

Network scale

180

0 100

120

(a) β = 1

140

160

Network scale

180

200

180

(c) β = 3

Energy consumption 140

160

Network scale

180

140

160

Network scale

180

120

140

160

Network scale

180

0 100

200

120

140

160

Network scale

(e) β = 5

180

200

0 100

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 409: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 800 nodes.

160

180

200

2 1

140

160

Network scale

180

200

180

200

180

200

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(d) β = 4 10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

160

9

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

140

Network scale

(b) β = 2

(c) β = 3

1

140

Network scale

5

10

Energy consumption

Energy consumption

0.5

120

8

1

(d) β = 4

1

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

180

(f) β = 6

2

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

160

1

0 100

200

9

2

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

4

120

120

(d) β = 4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

10

Energy consumption

160

Energy consumption

1

0 100

140

Network scale

9

Energy consumption

Energy consumption

9

120

120

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

8

5

200

180

2

Fig. 410: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 900 nodes.

Energy consumption

1

160

0.5

8

2

200

4

(e) β = 5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

8

180

10

(f) β = 6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

Network scale

1

0 100

160

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

Fig. 408: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 700 nodes.

0 100

120

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3

0.5 0 100

120

(b) β = 2

10

1

(e) β = 5

120

0 100

200

1.5

200

0 100

200

Energy consumption

Energy consumption

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption 140

Network scale

180

1

(d) β = 4

0.5

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

10

1

140

Network scale

9

4

(c) β = 3 10

0 100

120

5

(a) β = 1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

0 100

200

Energy consumption

1

0 100

180

9

Energy consumption

Energy consumption

9

120

160

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

1

Energy consumption

0 100

200

2

Energy consumption

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

160

8

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

140

Network scale

5

Energy consumption

120

Energy consumption

1

0 100

8

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

8

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

140

160

Network scale

(e) β = 5

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 411: Impact of network scale on the energy consumption (joule) under α = 5 and uniform distribution with 1000 nodes.

131

140

160

Network scale

180

0 100

200

120

(a) β = 1

140

160

Network scale

180

200

2

120

140

160

Network scale

180

200

0.5

140

160

Network scale

(e) β = 5

180

200

120

1

140

160

Network scale

180

200

140

160

Network scale

180

(f) β = 6

Fig. 412: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 100 nodes.

180

200

180

200

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

200

120

140

160

Network scale

(d) β = 4 5

1

120

160

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

(b) β = 2

5

2

120

120

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

0 100

200

2

0 100

Energy consumption

Energy consumption

1

180

4

5

5

160

0.5

4

4

0 100

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

120

1

(a) β = 1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3 Energy consumption

0 100

200

Energy consumption

1

0 100

180

2000

4

Energy consumption

Energy consumption

4

120

160

4000

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

Energy consumption

5000

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

120

10000

EMA−SIC Cell−AS Li et al.

Energy consumption

2000

6000

EMA−SIC Cell−AS Li et al.

Energy consumption

4000

0 100

4

15000

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

6000

140

160

Network scale

(e) β = 5

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 413: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 200 nodes.

132

120

140

160

Network scale

180

0 100

200

120

(a) β = 1

2

140

160

Network scale

180

0 100

0 100

Energy consumption

1

140

160

180

120

140

160

Network scale

180

180

2 120

2

120

140

160

Network scale

180

140

160

Network scale

180

0 100

1 140

160

Network scale

180

160

180

120

4

Energy consumption

Energy consumption

4 2 140

160

Network scale

180

0 100

1 0.5 120

Energy consumption

2 1 140

160

(e) β = 5

180

200

140

160

Network scale

180

160

180

200

1

2 120

4 2 140

160

180

200

140

160

Network scale

180

(f) β = 6

Fig. 415: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 400 nodes.

200

180

200

180

200

1 0.5 120

140

160

Network scale

(d) β = 4 5

2

120

180

1.5

0 100

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

160

5

4

0 100

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(c) β = 3

6

Network scale

120

(b) β = 2

5

120

160

2

0 100

200

6

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(d) β = 4

3

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1

5

Network scale

120

4

(c) β = 3 Energy consumption

5000

200

1.5

5

120

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

10000

5

6

120

140

Network scale

EMA−SIC Cell−AS Li et al.

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

(f) β = 6

Energy consumption

1

(a) β = 1

0 100

15000

2

0 100

200

180

5

0 100

Energy consumption

140

Network scale

160

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

120

140

Network scale

Fig. 416: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 500 nodes.

Energy consumption

0 100

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

Energy consumption

2000

200

1 0.5

(e) β = 5

Energy consumption

Energy consumption

Energy consumption

4000

180

1.5

(d) β = 4

2

(f) β = 6

200

5

3

120

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

160

5

4

0 100

140

Network scale

(b) β = 2

4

6000

120

(c) β = 3

4

(e) β = 5

EMA−SIC Cell−AS Li et al.

1

5

Fig. 414: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 300 nodes.

8000

2

0 100

200

6

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

160

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(d) β = 4

2

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1

5

Network scale

120

4

(c) β = 3 5

Energy consumption

2000

200

5

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

4000

4

4

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

120

160

6000

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Energy consumption

1

8000

Energy consumption

0 100

2

EMA−SIC Cell−AS Li et al.

Energy consumption

2000

10000

Energy consumption

4000

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

4

6000

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 417: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 600 nodes.

133

120

140

160

Network scale

180

0 100

200

120

(a) β = 1 Energy consumption

Energy consumption

6 4 2

200

120

140

160

Network scale

180

120

140

160

Network scale

180

0 100

200

160

180

0 100

2 1

0 100

200

120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

160

180

200

2

140

160

Network scale

180

0 100

200

1 180

200

120

140

160

Network scale

180

1

120

5

140

160

Network scale

180

Energy consumption

2

140

160

Network scale

(e) β = 5

180

200

180

200

140

160

Network scale

2 1

0 100

200

120

180

200

(f) β = 6

Fig. 419: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 800 nodes.

160

180

200

180

200

180

200

5

120

140

160

Network scale

180

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

200

120

160

5

6 4 2 120

140

Network scale

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

5

5

120

160

3

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

5

0 100

200

5

4

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

120

4

2

0 100

200

140

Network scale

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(a) β = 1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3 Energy consumption

160

5000

0 100

Energy consumption

Energy consumption

Energy consumption

5

0 100

140

Network scale

10000

5

4

120

(f) β = 6

EMA−SIC Cell−AS Li et al.

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

5

Energy consumption

2

(a) β = 1

0 100

15000

3

120

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

140

Network scale

160

5

4

120

140

Network scale

(d) β = 4

Energy consumption

120

120

4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

180

1

(e) β = 5

Energy consumption

0 100

200

Fig. 420: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 900 nodes.

Energy consumption

Energy consumption

Energy consumption

5000

180

2

4

10000

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(f) β = 6

EMA−SIC Cell−AS Li et al.

140

Network scale

(b) β = 2 5

120

0 100

200

Fig. 418: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 700 nodes.

15000

120

5

5

200

3

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption 140

Network scale

180

5

5

2

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

5

140

Network scale

4

1

0 100

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

120

120

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(a) β = 1

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3 Energy consumption

180

0 100

5

4

0 100

160

5000

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

Energy consumption

1

10000

Energy consumption

0 100

2

EMA−SIC Cell−AS Li et al.

Energy consumption

5000

15000

Energy consumption

10000

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

4

15000

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 421: Impact of network scale on the energy consumption (joule) under α = 3 and poisson distribution with 1000 nodes.

134

6

140

160

Network scale

180

200

120

(a) β = 1

4 2

140

160

Network scale

180

200

0 100

Energy consumption

0.5

1 140

160

Network scale

180

200

140

160

Network scale

(e) β = 5

180

200

120

120

2 1 140

160

Network scale

180

200

(f) β = 6

140

160

Network scale

180

Fig. 422: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 100 nodes.

200

180

200

180

200

6 4 2 0 100

200

120

140

160

Network scale

(d) β = 4 7

2 1

120

180

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

6

2

0 100

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

7

3

120

0 100

200

(c) β = 3

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

180

4

7

1

160

0.5

(a) β = 1

2

120

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

120

1

6

3

0 100

7

Energy consumption

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

0 100

180

2

6

Energy consumption

Energy consumption

6

120

160

4

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

0 100

Energy consumption

120

0.5

Energy consumption

0 100

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

0.5

Energy consumption

Energy consumption

Energy consumption

1

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 423: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 200 nodes.

135

6

180

0 100

200

120

(a) β = 1

2

140

160

Network scale

180

0 100

Energy consumption

Energy consumption

1 140

160

180

120

140

160

Network scale

180

2

120

140

160

Network scale

180

140

160

Network scale

180

2 1

0 100

200

120

(a) β = 1

140

160

Network scale

180

120

160

2 1 140

160

Network scale

(e) β = 5

180

200

Energy consumption 160

180

0 100

200

120

140

160

Network scale

120

200

(f) β = 6

Fig. 425: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 400 nodes.

160

180

200

120

140

160

Network scale

180

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

200

120

180

200

180

200

180

200

(b) β = 2

4 2 120

140

160

Network scale

180

1

0 100

200

120

160

(d) β = 4

2

120

140

Network scale

7

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

160

7

6

0 100

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3

180

140

Network scale

(f) β = 6

7

5

0 100

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

5

(a) β = 1

180

180

6

(d) β = 4 Energy consumption

Energy consumption

140

Network scale

160

(d) β = 4

1

0 100

200

0.5

140

Network scale

7

5

7

3

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

6

(c) β = 3

120

180

1

7

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

200

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

Energy consumption

2

180

1 0.5

Fig. 426: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 500 nodes.

7

4

160

2

(b) β = 2

6

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

140

Network scale

200

1.5

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(e) β = 5

Energy consumption

120

140

Network scale

3

120

160

7

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

Energy consumption

Energy consumption

Energy consumption

2

0 100

120

6

4

0 100

0 100

140

Network scale

(b) β = 2

2

(f) β = 6

6

120

(c) β = 3

4

(e) β = 5

5

1

7

Fig. 424: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 300 nodes.

10 x 10 EMA−SIC 8 Cell−AS Li et al.

2

0 100

200

4

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

180

6

(d) β = 4

2

160

10 x 10 EMA−SIC 8 Cell−AS Li et al.

7

3

140

Network scale

6

(c) β = 3 7

Network scale

120

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1

5

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

0 100

200

6

4

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

6

120

160

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Energy consumption

160

Energy consumption

140

Network scale

Energy consumption

120

Energy consumption

0 100

1

5

Energy consumption

2

2

Energy consumption

4

6

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

Energy consumption

6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 427: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 600 nodes.

136

6

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

1

120

140

160

Network scale

180

2 140

160

Network scale

180

0 100

120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

120

160

180

Energy consumption

6 4 2 140

160

(e) β = 5

180

200

0 100

120

Energy consumption 140

160

Network scale

180

140

160

Network scale

180

200

(f) β = 6

Fig. 429: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 800 nodes.

200

180

200

5

0 100

200

120

140

160

Network scale

(f) β = 6

6

5

120

140

160

Network scale

180

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

200

120

160

180

200

180

200

180

200

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

120

140

160

Network scale

180

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

200

120

160

7

6 4 2 120

140

Network scale

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

7

5

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

200

7

Network scale

120

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

140

120

(d) β = 4

6

Network scale

180

1

(a) β = 1

1

200

2

0 100

200

2

0 100

200

2

0 100

200

7

Energy consumption

180

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

0 100

160

Energy consumption

0 100

140

Network scale

Energy consumption

Energy consumption

Energy consumption

5

180

4

7

6

160

6

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

5

2

180

Fig. 430: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 900 nodes.

Energy consumption

Energy consumption

Energy consumption

5

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(e) β = 5

6 x 10 EMA−SIC Cell−AS Li et al. 4

140

Network scale

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

6

120

120

0 100

200

Fig. 428: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 700 nodes.

5

120

7

(f) β = 6

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

200

7

5

200

2

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

4

180

5

0 100

200

7

6

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

7

140

Network scale

6

2

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

120

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

5

0 100

0 100

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3 Energy consumption

180

7

6

0 100

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Energy consumption

180

Energy consumption

160

5

Energy consumption

140

Network scale

1

Energy consumption

120

2

Energy consumption

0 100

3

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

5

6

5

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 431: Impact of network scale on the energy consumption (joule) under α = 4 and poisson distribution with 1000 nodes.

137

7

140

160

Network scale

180

0 100

200

120

(a) β = 1

0 100

200

120

140

160

Network scale

180

200

0 100

120

9

140

160

Network scale

180

200

Energy consumption

2 1 140

160

Network scale

(e) β = 5

180

200

0 100

200

140

160

Network scale

120

120

180

200

(f) β = 6

Fig. 432: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 100 nodes.

180

200

180

200

180

200

(b) β = 2

140

160

Network scale

180

5

0 100

200

120

140

160

Network scale

(d) β = 4 9

3 2 1 120

160

8

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

9

2

120

1

(c) β = 3

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

180

2

0 100

9

3

160

6 x 10 EMA−SIC Cell−AS Li et al. 4

(d) β = 4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

140

Network scale

2

(a) β = 1

2

(c) β = 3

120

120

4 x 10 EMA−SIC Cell−AS 3 Li et al.

8

6 x 10 EMA−SIC Cell−AS Li et al. 4

Energy consumption

Energy consumption

Energy consumption

1 0.5

Energy consumption

180

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

160

8

8

0 100

140

Network scale

Energy consumption

2

5

Energy consumption

120

4

Energy consumption

0 100

6

Energy consumption

1

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

2

8

7

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 433: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 200 nodes.

138

8

160

180

0 100

200

120

(a) β = 1 8

160

180

0 100

120

160

180

140

160

180

200

120

(e) β = 5

140

160

180

120

140

160

Network scale

180

1

0 100

200

120

(a) β = 1

Energy consumption

0 100

200

120

140

160

Network scale

180

2 140

160

Network scale

180

140

160

Network scale

180

0 100

200

120

(c) β = 3

0 100

120

160

Network scale

180

120

140

160

Network scale

180

Energy consumption

2

6 4 2 0 100

140

160

Network scale

(e) β = 5

180

200

0 100

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 435: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 400 nodes.

160

180

200

180

200

180

200

9

Energy consumption

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5

120

140

160

Network scale

180

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

200

120

160

(d) β = 4

6 4 2 120

140

Network scale

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

5

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

9

9

4

160

8

0.5

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

(f) β = 6

1

(d) β = 4

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

140

Energy consumption

120

180

0.5

9

1

200

1

(a) β = 1 Energy consumption

Energy consumption

5

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

160

10

4

120

140

Network scale

(d) β = 4

9

8

Energy consumption

180

6

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

160

Energy consumption

0.5

0 100

140

Network scale

8

2

200

Fig. 436: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 500 nodes.

Energy consumption

Energy consumption

Energy consumption

1

180

1

(e) β = 5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

8

8

140

Network scale

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

120

0 100

200

Fig. 434: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 300 nodes.

0 100

120

9

9

Network scale

2

(c) β = 3

5

0 100

4

0 100

200

5

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

140

Network scale

9

2

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

4

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

0.5

200

140

Network scale

8

1

9

Network scale

120

Energy consumption

Energy consumption

Energy consumption 140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(c) β = 3

0 100

180

9

5

120

160

0.5

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

1

Energy consumption

140

Network scale

1

Energy consumption

120

2

Energy consumption

0 100

8

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

Energy consumption

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 437: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 600 nodes.

139

8

160

180

0 100

200

120

(a) β = 1

0.5

120

140

160

Network scale

180

2 1 120

120

140

160

Network scale

180

160

120

(e) β = 5

140

160

Network scale

180

0.5 160

Network scale

180

0 100

200

120

(a) β = 1

Energy consumption 160

180

160

Network scale

180

0 100

200

140

160

Network scale

180

120

140

160

Network scale

180

200

0 100

120

(c) β = 3

140

160

Network scale

180

0 100

120

140

160

Network scale

180

4 2

120

140

160

Network scale

(e) β = 5

180

200

0 100

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 439: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 800 nodes.

160

180

200

180

200

180

200

9

1 0.5 120

140

160

Network scale

180

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

200

120

160

(d) β = 4

5

120

140

Network scale

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

(b) β = 2

(c) β = 3

1

200

6

0 100

9

Energy consumption

Energy consumption

5

160

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

1.5

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

Network scale

8

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

10

9

120

(f) β = 6

1

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

1 0.5

Energy consumption

1

200

1.5

200

Energy consumption

0.5

2

160

(d) β = 4

9

3

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1 Energy consumption

1

120

10

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

180

1

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

120

(b) β = 2

9

Energy consumption

140

200

2

8

5

180

Fig. 440: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 900 nodes.

Energy consumption

1

160

3

(e) β = 5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

1.5

0 100

0 100

8

8

140

Network scale

5

(f) β = 6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

140

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

Fig. 438: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 700 nodes.

120

0.5 0 100

140

Network scale

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

0.5 0 100

120

(b) β = 2

9

1

200

0 100

200

9

(d) β = 4

180

180

1

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption 140

Network scale

160

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

10

5

140

Network scale

9

3

(c) β = 3

0 100

120

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

0 100

200

Energy consumption

Energy consumption

Energy consumption

1

9

Energy consumption

180

9

9

0 100

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

5

Energy consumption

140

Network scale

1 0.5

Energy consumption

120

1.5

Energy consumption

0 100

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

0.5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

1

8

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 441: Impact of network scale on the energy consumption (joule) under α = 5 and poisson distribution with 1000 nodes.

140

120

140

160

Network scale

180

0 100

200

120

(a) β = 1

500 0 100

200

0.5

120

140

160

Network scale

180

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

(c) β = 3

140

160

Network scale

180

Energy consumption

2 140

160

Network scale

(e) β = 5

180

200

0 100

0.5

140

160

Network scale

120

180

200

1.5 1 0.5 120

140

160

Network scale

180

(f) β = 6

Fig. 442: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 100 nodes.

160

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

200

120

160

180

200

180

200

(d) β = 4

5

120

140

Network scale

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

(b) β = 2

(c) β = 3

1

120

2000

200

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

180

4000

4

(d) β = 4

4

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

200

5

6

140

Network scale

6000

4

1

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

120

EMA−SIC Cell−AS Li et al.

8000

(a) β = 1 Energy consumption

Energy consumption

Energy consumption

1

4

Energy consumption

180

1000

4

4

0 100

160

1500

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Energy consumption

2000

Energy consumption

0 100

4000

10000

EMA−SIC Cell−AS Li et al.

2000

Energy consumption

500

6000

Energy consumption

1000

2500

EMA−SIC Cell−AS Li et al.

Energy consumption

8000

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

1500

140

160

Network scale

(e) β = 5

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 443: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 200 nodes.

141

180

0 100

200

120

(a) β = 1

160

180

120

Energy consumption

140

160

Network scale

180

200

160

180

1 0.5 0 100

120

(e) β = 5

140

160

Network scale

180

1000

140

160

Network scale

180

EMA−SIC Cell−AS Li et al.

10000

5000

0 100

200

120

(a) β = 1

160

180

200

Energy consumption

Energy consumption 140

160

Network scale

(e) β = 5

180

200

120

140

160

Network scale

180

Energy consumption 140

160

180

0.5 120

140

160

Network scale

180

200

(f) β = 6

Fig. 445: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 400 nodes.

180

200

1.5 1 0.5 0 100

200

120

140

160

Network scale

(f) β = 6

15000

140

160

Network scale

180

EMA−SIC Cell−AS Li et al.

10000

5000

0 100

200

120

1

140

160

Network scale

180

180

200

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

200

120

140

160

Network scale

(d) β = 4 5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

160

4

2

120

140

Network scale

(b) β = 2

(c) β = 3

1

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

5

1.5

140

Network scale

(d) β = 4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

120

5

Network scale

120

(d) β = 4

0.5

120

0 100

5

1

0 100

200

1000

Energy consumption

2

200

2

(a) β = 1

4

180

4

4

(c) β = 3 2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

180

2000

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

5

0 100

180

Energy consumption

Energy consumption

Energy consumption 140

Network scale

160

EMA−SIC Cell−AS Li et al.

3000

4

1

120

160

200

Fig. 446: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 500 nodes.

(b) β = 2

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

180

6

(e) β = 5

Energy consumption

Energy consumption

Energy consumption

2000

120

120

4000

15000

140

Network scale

0.5 0 100

160

(b) β = 2

1

(f) β = 6

EMA−SIC Cell−AS Li et al.

0 100

120

140

Network scale

4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

Fig. 444: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 300 nodes.

3000

120

10 x 10 EMA−SIC 8 Cell−AS Li et al.

5

1.5

200

0 100

200

(c) β = 3

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption 140

180

1

0 100

5

5

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

(d) β = 4

4

140

Network scale

(a) β = 1

2

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Network scale

120

4

Energy consumption

Energy consumption

Energy consumption 140

Network scale

120

0 100

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

0 100

180

4

1

120

160

(b) β = 2

4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

5000

Energy consumption

160

1000

Energy consumption

140

Network scale

2000

EMA−SIC Cell−AS Li et al.

10000

Energy consumption

120

4000

2000

Energy consumption

0 100

6000

15000

EMA−SIC Cell−AS Li et al.

Energy consumption

1000

Energy consumption

2000

3000

EMA−SIC Cell−AS Li et al.

8000

Energy consumption

10000

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

3000

120

140

160

Network scale

(e) β = 5

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 447: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 600 nodes.

142

140

160

Network scale

180

200

120

(a) β = 1

1

140

160

Network scale

180

0 100

200

0 100

120

Energy consumption

1

140

160

180

200

140

160

Network scale

180

0 100

200

1

120

140

160

Network scale

140

160

Network scale

180

(f) β = 6

0 100

140

160

Network scale

180

140

160

Network scale

180

200

120

(a) β = 1

1

140

160

Network scale

180

0 100

200

5

120

1

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

120

140

160

Network scale

180

0 100

200

120

200

1

120

140

160

Network scale

180

200

140

160

Network scale

180

0 100

120

(f) β = 6

Fig. 449: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 800 nodes.

140

160

Network scale

180

200

180

200

(d) β = 4 5

1

120

200

5

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

1

160

(b) β = 2

2

0 100

140

Network scale

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

2

120

160

0.5

(a) β = 1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

1

4

5

Energy consumption

Energy consumption

0 100

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

1000

200

5

(c) β = 3

0 100

180

2000

4

2

120

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

0 100

140

Network scale

3000

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

0.5

EMA−SIC Cell−AS Li et al.

4000

Energy consumption

120

1

0 100

180

1

4

5000

Energy consumption

0 100

200

(f) β = 6

Energy consumption

1000

180

2

0 100

200

Energy consumption

2000

160

Fig. 450: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 900 nodes.

Energy consumption

3000

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(e) β = 5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

4000

120

(d) β = 4

4

EMA−SIC Cell−AS Li et al.

200

5

1

120

180

5

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

180

140

Network scale

(b) β = 2

5

Fig. 448: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 700 nodes.

5000

120

4

2

0 100

200

1

(e) β = 5

180

0.5

(a) β = 1

2

120

160

1

4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

Network scale

120

(d) β = 4

5

Energy consumption

1000

200

5

(c) β = 3

120

2000

4

2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

4

120

160

3000

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

Energy consumption

0 100

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

120

5000

4000

Energy consumption

0 100

10000

EMA−SIC Cell−AS Li et al.

Energy consumption

1000

Energy consumption

2000

5000

EMA−SIC Cell−AS Li et al.

Energy consumption

3000

4

15000

EMA−SIC Cell−AS Li et al.

Energy consumption

Energy consumption

4000

140

160

Network scale

(e) β = 5

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 451: Impact of network scale on the energy consumption (joule) under α = 3 and cluster distribution with 1000 nodes.

143

5

120

140

160

Network scale

180

0 100

200

120

(a) β = 1

120

140

160

Network scale

180

120

140

160

Network scale

180

Energy consumption

1

140

160

Network scale

(e) β = 5

180

200

5

140

160

Network scale

Energy consumption

2

0 100

200

120

180

200

(f) β = 6

Fig. 452: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 100 nodes.

160

180

200

180

200

180

200

(b) β = 2

120

140

160

Network scale

180

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

200

120

160

6

5

120

140

Network scale

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

6

6

120

4

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

180

5

0 100

200

6

2

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

6

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

1

0 100

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

120

5

Energy consumption

Energy consumption

Energy consumption

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3 Energy consumption

180

0.5

(b) β = 2

5

0 100

160

1

6

5

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Energy consumption

0 100

2

1.5

Energy consumption

0.5

5

5

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

1

6 x 10 EMA−SIC Cell−AS Li et al. 4

Energy consumption

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 453: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 200 nodes.

144

120

(a) β = 1

0.5

140

160

Network scale

180

3 2 1 120

(c) β = 3

120

140

160

Network scale

180

160

180

120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

4 2 0 100

200

120

(a) β = 1

0.5

140

160

Network scale

180

180

2

120

120

140

160

Network scale

180

160

(e) β = 5

120

180

200

Energy consumption 180

120

Energy consumption 140

160

Network scale

180

200

(f) β = 6

Fig. 455: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 400 nodes.

180

200

180

200

1

120

140

160

Network scale

(f) β = 6

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

180

1 0.5 0 100

200

120

140

160

Network scale

180

200

180

200

180

200

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(d) β = 4 7

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

160

6

1

120

140

Network scale

(b) β = 2

(c) β = 3

0.5

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

6

1

0 100

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

(d) β = 4 Energy consumption

140

Network scale

140

Network scale

1

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

(d) β = 4

6

7

5

120

(a) β = 1

4

(c) β = 3 15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

6

Energy consumption

160

Energy consumption

Energy consumption

Energy consumption

1

0 100

140

Network scale

6

6

120

120

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

180

5

Energy consumption

1

160

2

Fig. 456: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 500 nodes.

5

6

200

4

(e) β = 5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

Network scale

5

0 100

180

7

(f) β = 6

Fig. 454: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 300 nodes.

120

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

160

6

0.5 0 100

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3

1

0 100

120

(b) β = 2

6

0.5

200

0 100

200

1

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption 140

Network scale

180

1.5

7

5

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(d) β = 4

6

140

Network scale

6

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

120

(a) β = 1

Energy consumption

120

0 100

0 100

200

Energy consumption

Energy consumption

Energy consumption

1

Energy consumption

180

6

6

0 100

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Energy consumption

0 100

200

5

Energy consumption

180

1

Energy consumption

160

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

140

Network scale

4

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

Energy consumption

120

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

1

0 100

5

5

Energy consumption

Energy consumption

5

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

140

160

Network scale

(e) β = 5

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 457: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 600 nodes.

145

120

(a) β = 1

140

160

Network scale

180

4 2 0 100

120

140

160

Network scale

180

120

140

160

Network scale

180

4 x 10 EMA−SIC Cell−AS 3 Li et al.

200

1

120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

0 100

200

120

(a) β = 1

160

180

1

140

160

180

4 2 0 100

200

120

(c) β = 3

140

160

Network scale

180

0.5

120

140

160

Network scale

(e) β = 5

180

200

1

120

Energy consumption 180

140

160

Network scale

180

200

(f) β = 6

Fig. 459: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 800 nodes.

160

200

180

200

2 1

0 100

200

120

140

160

Network scale

(f) β = 6

5

1 120

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

180

5

0 100

200

120

1.5 1 0.5 140

160

Network scale

180

180

200

180

200

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

200

120

140

160

Network scale

(d) β = 4 7

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

160

6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

120

140

Network scale

(b) β = 2

7

2

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

2

0 100

200

Energy consumption

Energy consumption

1

140

Network scale

3

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

(d) β = 4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

7

7

0 100

200

6

6

180

2

(a) β = 1

6

Network scale

120

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

6

180

4

7

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

140

Network scale

200

6

5

Energy consumption

120

180

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(e) β = 5

5

160

Fig. 460: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 900 nodes.

Energy consumption

1

160

0.5

5

2

140

Network scale

1

0 100

140

Network scale

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

5

Energy consumption

120

(f) β = 6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

(b) β = 2

7

2

0 100

0 100

200

(c) β = 3

Fig. 458: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 700 nodes.

0 100

0 100

200

Energy consumption

Energy consumption

Energy consumption

0.5

180

1

7

1

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

(d) β = 4

7

140

Network scale

(a) β = 1

6

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

120

6

(c) β = 3

0 100

0 100

200

6

1

0 100

180

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

Energy consumption

6

120

160

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

Energy consumption

0 100

200

5

Energy consumption

180

2

Energy consumption

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

140

Network scale

Energy consumption

120

1 0.5

5

6 x 10 EMA−SIC Cell−AS Li et al. 4

Energy consumption

1

Energy consumption

2

0 100

5

6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

Energy consumption

5

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

140

160

Network scale

(e) β = 5

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 461: Impact of network scale on the energy consumption (joule) under α = 4 and cluster distribution with 1000 nodes.

146

7

140

160

Network scale

180

0 100

200

120

(a) β = 1 Energy consumption 140

160

Network scale

0 100

200

120

180

200

2 1 120

(c) β = 3

160

180

0 100

200

140

160

Network scale

180

200

5

120

140

160

Network scale

180

Energy consumption

2

120

140

160

Network scale

(e) β = 5

180

200

6 4 2 0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 462: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 100 nodes.

200

180

200

180

200

2 1 120

160

(d) β = 4

6 4 2 120

140

Network scale

8

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

180

3

0 100

200

8

4

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

(d) β = 4 10 x 10 EMA−SIC 8 Cell−AS Li et al.

140

Network scale

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

8

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

(b) β = 2

7

3

0 100

140

Network scale

5

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

8

Energy consumption

1 0.5

Energy consumption

Energy consumption

1

0 100

180

8

2

120

160

1.5

(b) β = 2

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

Energy consumption

120

Energy consumption

0 100

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

0.5

4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Energy consumption

Energy consumption

Energy consumption

1

7

7

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

7

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 463: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 200 nodes.

147

140

160

Network scale

180

0 100

200

120

(a) β = 1

1

140

160

Network scale

180

2

120

140

160

Network scale

180

180

120

140

160

Network scale

180

Network scale

180

1

140

160

Network scale

180

0 100

200

120

(a) β = 1

140

160

Network scale

180

Energy consumption

1

160

Network scale

180

200

4 2

120

(c) β = 3

Energy consumption 180

0 100

200

140

160

Network scale

180

1

120

140

160

Network scale

180

120

0 100

160

(e) β = 5

180

200

120

120

1

140

160

Network scale

180

140

160

Network scale

180

200

(f) β = 6

Fig. 465: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 400 nodes.

180

200

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

200

120

140

160

Network scale

(d) β = 4 9

Energy consumption

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

160

8

2

120

140

Network scale

(b) β = 2

(c) β = 3 Energy consumption

Energy consumption 140

Network scale

0 100

200

5

9

1

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(d) β = 4

5

160

7

2

0 100

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

Network scale

(f) β = 6

9

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

8

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

1

(a) β = 1

8

2

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

Energy consumption

8

140

120

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

200

2

7

5

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(e) β = 5

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

Fig. 466: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 500 nodes.

Energy consumption

2

120

(d) β = 4

0.5 0 100

180

2 0 100

200

1

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

Energy consumption

7

Energy consumption

160

9

(f) β = 6

Fig. 464: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 300 nodes.

Energy consumption

140

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

200

200

4

(c) β = 3 Energy consumption

0 100

200

180

6

Energy consumption

160

160

8

1

120

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Energy consumption

140

Network scale

120

(b) β = 2

9

5

(e) β = 5

0 100

0 100

200

2

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

2

120

180

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(d) β = 4

4

0 100

160

8

6

120

140

Network scale

8

4

(c) β = 3 8

0 100

120

5

(a) β = 1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

1

0 100

200

Energy consumption

2

0 100

180

8

Energy consumption

Energy consumption

8

120

160

2

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

5

7

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Energy consumption

0.5

Energy consumption

1

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

1.5

0 100

7

7

7

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

120

140

160

Network scale

(e) β = 5

180

200

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 467: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 600 nodes.

148

0 100

200

120

(a) β = 1

1

140

160

Network scale

180

0 100

200

120

140

160

Network scale

180

120

140

160

Network scale

180

3 2 1 120

(e) β = 5

140

160

Network scale

180

140

160

Network scale

180

0 100

200

120

(a) β = 1

180

1

140

160

Network scale

180

0 100

200

(c) β = 3

140

160

Network scale

140

160

Network scale

(e) β = 5

180

200

2 1 120

140

160

Network scale

180

200

(f) β = 6

Fig. 469: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 800 nodes.

180

200

1

Energy consumption

0 100

200

120

140

160

Network scale

(f) β = 6

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

180

5

0 100

200

120

2 1

140

160

Network scale

180

180

200

180

200

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

200

120

140

160

Network scale

(d) β = 4 9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

160

8

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

140

Network scale

(b) β = 2

(c) β = 3

3

0 100

180

9

Energy consumption

Energy consumption

0.5

120

120

(d) β = 4

1

160

2

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

4

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

200

2

(a) β = 1

180

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

8

120

140

Network scale

(d) β = 4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

5

9

Energy consumption

160

8

2

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

8

120

120

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

Fig. 470: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 900 nodes.

Energy consumption

120

0 100

200

0.5

Energy consumption

1

180

5

7

5

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

(e) β = 5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

2

180

1

7

3

160

1.5

0 100

180

9

(f) β = 6

7

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

200

160

8

1 120

140

Network scale

(b) β = 2

9

Fig. 468: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 700 nodes.

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

(c) β = 3

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

0 100

200

2

0 100

200

Energy consumption

Energy consumption

Energy consumption

0.5

180

3

(d) β = 4

1

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

8

(c) β = 3

0 100

120

(a) β = 1

5

9

0 100

0 100

200

8

2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

8

120

160

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

1 0.5

Energy consumption

180

2

Energy consumption

160

4

Energy consumption

140

Network scale

Energy consumption

120

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Energy consumption

1

5

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Energy consumption

2

Energy consumption

3

0 100

7

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

Energy consumption

Energy consumption

7

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

140

160

Network scale

(e) β = 5

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 471: Impact of network scale on the energy consumption (joule) under α = 5 and cluster distribution with 1000 nodes.

149

160

100 100

(a) β = 1

400 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

160

1000 500 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2000

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

3000

160

4000 2000

120

140

160

Network scale

Energy consumption

300 200 100

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 472: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 100 nodes.

160

3000 2000 1000 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 15000

6000 4000 2000 0 100

140

Network scale

4000

500

0 100

120

EMA−SIC Li et al. 180 200

(b) β = 2

8000

6000

0 100

140

Network scale

400

(c) β = 3

8000

4000

120

1000

(d) β = 4

5000

120

60 100

EMA−SIC Li et al. 180 200

1500

1500

(c) β = 3

1000

80

500

(a) β = 1

2000

600

0 100

140

Network scale

100

(b) β = 2

800

0 100

120

EMA−SIC Li et al. 180 200

Energy consumption

140

Network scale

150

Energy consumption

120

EMA−SIC Li et al. 180 200

200

120

Energy consumption

50

250

600

140

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

60

300

Energy consumption

70

40 100

Energy consumption

160

350

Energy consumption

Energy consumption

80

10000

5000

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 473: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 200 nodes.

150

120

140

160

Network scale

200 100

120

(a) β = 1

1500

1000

120

140

160

Network scale

100 100

EMA−SIC Li et al. 180 200

3000 2000 1000 0 100

120

140

160

Network scale

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1.5 1 0.5 0 100

120

140

160

Network scale

120

EMA−SIC Li et al. 180 200

2 x 10

140

160

Network scale

EMA−SIC Li et al. 180 200

(b) β = 2

1000

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 0 100

120

160

Network scale

EMA−SIC Li et al. 180 200

4

1.5 1 0.5 0 100

140

(d) β = 4

4

120

(f) β = 6

140

160

Network scale

EMA−SIC Li et al. 180 200

3 x 10

2

1

0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

250

400

100

160

Network scale

200 100

120

(a) β = 1

Network scale

1000

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

1500

120

5000 4000 3000 2000 1000 100

5000

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

10000

2.5 x 10

120

140

160

Network scale

EMA−SIC Li et al. 180 200

160

Network scale

800 600 400 100

2

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1500 1000 120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 475: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 400 nodes.

Network scale

EMA−SIC Li et al. 180 200

8000 6000 4000 2000 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 4

4

1.5 1 0.5 0 100

160

(b) β = 2

2000

2 x 10

140

10000

500 100

4

0.5

120

(c) β = 3

1.5

0 100

140

2500

(d) β = 4

15000

140

120

EMA−SIC Li et al. 180 200

1000

3000

(c) β = 3

120

100 100

1200

(a) β = 1

6000

2000

0 100

160

200

(b) β = 2

2500

500 100

140

EMA−SIC Li et al. 180 200

Energy consumption

140

400

Energy consumption

120

EMA−SIC Li et al. 180 200

600

300

Energy consumption

150

800

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

200

1400

Energy consumption

1000

Energy consumption

Fig. 476: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 500 nodes.

Energy consumption

Energy consumption

200 100

Fig. 474: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 300 nodes.

50 100

Energy consumption

400

(c) β = 3

2 x 10

(e) β = 5

Energy consumption

600

8000

500 100

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

5000

Network scale

1500

4

10000

160

2000

(d) β = 4

15000

140

800

2500

4000

(c) β = 3

0 100

120

EMA−SIC Li et al. 180 200

1000

(a) β = 1

5000

Energy consumption

Energy consumption

160

150

(b) β = 2

2000

500 100

140

Network scale

EMA−SIC Li et al. 180 200

200

Energy consumption

400

Energy consumption

50 100

EMA−SIC Li et al. 180 200

600

250

Energy consumption

100

800

1200

Energy consumption

150

Energy consumption

300

1000

Energy consumption

Energy consumption

200

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 477: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 600 nodes.

151

200

120

140

160

Network scale

500 0 100

120

(a) β = 1

1500 120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 100

140

160

Network scale

500 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3000

2000

1000 100

1 0.5 120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 2 1

0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

2.5 x 10

Energy consumption

2 1.5

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8000 6000 4000 2000 100

120

(c) β = 3

4 x 10

140

(b) β = 2

160

4

4

2 1.5 1 0.5 0 100

140

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

4

4

120

10000

(d) β = 4 Energy consumption

120

(f) β = 6

140

160

Network scale

EMA−SIC Li et al. 180 200

5 x 10 4 3 2 1

0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 478: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 700 nodes.

Fig. 480: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 900 nodes.

500

600

100 100

120

140

160

Network scale

500 0 100

120

(a) β = 1

Network scale

100 100

1000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8000 6000 4000 2000 100

120

(c) β = 3

1 0.5 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

2

0 100

140

160

Network scale

EMA−SIC Li et al. 180 200

2000 1000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 479: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 800 nodes.

Network scale

EMA−SIC Li et al. 180 200

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 4

3 x 10

2

1

0 100

160

5000

(c) β = 3

4

140

10000

4

1

120

(b) β = 2

3000

5 x 10

0 100

Network scale

4000

4

1.5

160

500

15000

(d) β = 4

4

140

1000

5000

10000

Energy consumption

2000

Energy consumption

2500

1500

120

EMA−SIC Li et al. 180 200

1500

(a) β = 1

12000

3000

0 100

160

200

(b) β = 2

3500

2.5 x 10

140

EMA−SIC Li et al. 180 200

300

Energy consumption

EMA−SIC Li et al. 180 200

400

Energy consumption

200

1000

2000

500

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

300

1500

Energy consumption

400

Energy consumption

2000

Energy consumption

Energy consumption

120

1000

4000

Energy consumption

2000

Energy consumption

Energy consumption

2500

0 100

Energy consumption

100 100

EMA−SIC Li et al. 180 200

1500

(a) β = 1

8000

(c) β = 3

Energy consumption

Network scale

200

10000

3000

1000 100

Energy consumption

160

300

(b) β = 2

3500

2.5 x 10

140

EMA−SIC Li et al. 180 200

400

Energy consumption

100 100

EMA−SIC Li et al. 180 200

1000

Energy consumption

300

1500

Energy consumption

400

2000

500

Energy consumption

600

2000

Energy consumption

Energy consumption

500

5 x 10 4 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 481: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 1000 nodes.

152

8000

1000

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 5000 4000 100

120

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

5 4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3.5 x 10

160

0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(b) β = 2

4

3 2.5 2 1.5 1 100

120

(d) β = 4

4

140

160

Network scale

EMA−SIC Li et al. 180 200

10 x 10

6 4 2 100

8 6 4 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

2 x 10 1.5 1 0.5 0 100

120

140

160

Network scale

120

(c) β = 3

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 482: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 100 nodes.

2.5 x 10

160

Network scale

EMA−SIC Li et al. 180 200

5

5

2 1.5 1 0.5 100

140

(d) β = 4

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

10

4

8

5

Energy consumption

Energy consumption

140

Network scale

1

(a) β = 1

6 x 10

(c) β = 3

2 100

120

EMA−SIC Li et al. 180 200

1.5

Energy consumption

1.5

12 x 10

2000

4

4

4

0.5 100

Network scale

4000

0 100

Energy consumption

2 x 10

160

6000

(b) β = 2 Energy consumption

Energy consumption

(a) β = 1

140

EMA−SIC Li et al. 180 200

Energy consumption

0 100

7000

Energy consumption

2000

2 x 10

8000

Energy consumption

3000

Energy consumption

Energy consumption

4000

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 483: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 200 nodes.

153

160

Network scale

120

(a) β = 1

4 3 2

140

160

Network scale

EMA−SIC Li et al. 180 200

14 x 10

Energy consumption

10 8 6 4 100

1 120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

5 4 3 2 120

140

160

Network scale

EMA−SIC Li et al. 180 200

0.5 100

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 3 2

120

2.5 x 10

140

160

Network scale

EMA−SIC Li et al. 180 200

4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1

120

(c) β = 3

1 0.5 0 100

140

160

Network scale

EMA−SIC Li et al. 180 200

2

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

4 2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 485: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 400 nodes.

160

160

Network scale

4 3 2

1 100

120

6

4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2.5 x 10

4 3

120

140

160

Network scale

(e) β = 5

Network scale

EMA−SIC Li et al. 180 200

2 1.5 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

5

1 100

160

5

0.5 100

6 x 10

2

140

(b) β = 2

5

8 x 10 6

140

Network scale

EMA−SIC Li et al. 180 200

5 x 10

(c) β = 3 Energy consumption

3

140

EMA−SIC Li et al. 180 200

8 x 10

2 100

5

Energy consumption

5

120

120

(d) β = 4

4 x 10

120

4

1.5

0.5 100

2

4

2

Energy consumption

3

Energy consumption

4

4

(f) β = 6

1.5

5

5

Network scale

EMA−SIC Li et al. 180 200

6

(a) β = 1

2 x 10

160

8 x 10

0 100

4

(b) β = 2

6 x 10

140

(d) β = 4

Energy consumption

2

0 100

120

Fig. 486: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 500 nodes.

4 x 10

(a) β = 1 Energy consumption

1

(e) β = 5

3

Network scale

EMA−SIC Li et al. 180 200

5

1 100

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1

Energy consumption

Network scale

160

1.5

4

4

0.5

1 100

160

5 x 10

(f) β = 6

1.5

2 100

140

140

5

(c) β = 3

Fig. 484: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 300 nodes.

0 100

120

EMA−SIC Li et al. 180 200

2 x 10

5

(e) β = 5

2 x 10

120

(b) β = 2

4

2 100

6 x 10

1 100

1

0 100

Energy consumption

Energy consumption

1.5

Network scale

6

5

2

160

2

8 x 10

(d) β = 4

2.5

140

3

4

12

5

0.5 100

120

4 x 10

(a) β = 1

4

(c) β = 3 3 x 10

Network scale

0 100

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

4

120

160

0.5

(b) β = 2

5 x 10

1 100

140

EMA−SIC Li et al. 180 200

Energy consumption

0.5 100

1

Energy consumption

140

1

Energy consumption

120

EMA−SIC Li et al. 180 200

1.5

4

4

1.5

Energy consumption

0 100

2

2 x 10

Energy consumption

5000

Energy consumption

Energy consumption

10000

4

Energy consumption

2.5 x 10

15000

10 x 10

5

8 6 4 2 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 487: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 600 nodes.

154

(a) β = 1 4

8 6 4 2 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2.5 x 10

1.5 1 0.5 100

120

4 3

160

Network scale

EMA−SIC Li et al. 180 200

12 x 10

Energy consumption

Energy consumption

5

140

140

160

Network scale

EMA−SIC Li et al. 180 200

12 x 10

6 4 2 100

8 6 4 120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

6 4 2

120

(a) β = 1 4

6

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1.5 1 120

4 2

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

6

140

160

12 x 10

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1

0 100

120

12 x 10

4 120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 489: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

5

10 8 6 4 2 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4

3 2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8 x 10 6 4 2

0 100

120

8 6 4 120

140

160

Network scale

EMA−SIC Li et al. 180 200

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

6

4

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

8 x 10

2 100

160

5

4

10

2 100

140

(b) β = 2

5

6

140

(f) β = 6

4 x 10

12 x 10

5

8

2 100

3

(c) β = 3

10

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

4

(d) β = 4

5

120

140

Network scale

EMA−SIC Li et al. 180 200

160

4 x 10

(a) β = 1

2.5

(c) β = 3 Energy consumption

Network scale

3 x 10

0.5 100

8 x 10

Network scale

6

0 100

Energy consumption

4

0 100

160

Energy consumption

8

2 100

140

EMA−SIC Li et al. 180 200

140

Fig. 490: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 900 nodes.

5

10

160

8 x 10

(b) β = 2 Energy consumption

Energy consumption

12 x 10

140

EMA−SIC Li et al. 180 200

4

8 x 10

0 100

120

(e) β = 5

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

0 100

120

0 100

4

1

0 100

(b) β = 2

8

Fig. 488: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 700 nodes.

2

2

5

(f) β = 6

3

4

(c) β = 3

10

4

Network scale

6

5

(e) β = 5

4 x 10

160

10

5

2 100

140

4

(d) β = 4

5

120

120

EMA−SIC Li et al. 180 200

8 x 10

(a) β = 1

2

6 x 10

1 100

Network scale

5

(c) β = 3

2

160

(b) β = 2 Energy consumption

Energy consumption

10 x 10

140

0 100

Energy consumption

120

1

Energy consumption

Network scale

1 100

2

Energy consumption

160

EMA−SIC Li et al. 180 200

3

Energy consumption

140

2

Energy consumption

120

3

Energy consumption

0 100

EMA−SIC Li et al. 180 200

4

4

4 x 10

EMA−SIC Li et al. 180 200

Energy consumption

1

5

Energy consumption

2

6 x 10

Energy consumption

4

4

Energy consumption

Energy consumption

4

3 x 10

15 x 10

5

10

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 491: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 1000 nodes.

155

140

160

Network scale

1 100

120

(a) β = 1

6

4

140

160

Network scale

EMA−SIC Li et al. 180 200

1.8 x 10

Energy consumption

Energy consumption

5

120

1.4 1.2 1 0.8 100

120

160

Network scale

EMA−SIC Li et al. 180 200

2 x 10

2.5 2

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Network scale

10

5

0 100

120

5 4 3 120

140

160

Network scale

EMA−SIC Li et al. 180 200

0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 2 1

0 100

120

(f) β = 6

Fig. 492: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 100 nodes.

6 5 4 3 120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

7 x 10

2 100

Network scale

EMA−SIC Li et al. 180 200

4 x 10

6

6

160

6

6

1

0 100

140

(b) β = 2

(c) β = 3

7 x 10

2 100

160

1.5

6

Energy consumption

Energy consumption

140

140

EMA−SIC Li et al. 180 200

5

(a) β = 1

1.6

6

120

120

(d) β = 4

3

1.5 100

Network scale

0 100

6

(c) β = 3 3.5 x 10

160

2

(b) β = 2

8 x 10

2 100

140

EMA−SIC Li et al. 180 200

4

Energy consumption

2

Energy consumption

120

3

Energy consumption

0 100

EMA−SIC Li et al. 180 200

4

15 x 10

Energy consumption

1

5

6 x 10

EMA−SIC Li et al. 180 200

Energy consumption

2

5

6 x 10

Energy consumption

5

Energy consumption

Energy consumption

5

3 x 10

12 x 10

6

10 8 6 4 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 493: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 200 nodes.

156

120

1.5 1

140

160

Network scale

EMA−SIC Li et al. 180 200

4 3 2

1 100

120

(c) β = 3 6

6 4 2 100

140

160

Network scale

120

140

160

Network scale

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

1.5

1

120

(e) β = 5

140

160

Network scale

120

140

160

Network scale

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

2

1

0 100

120

0 100

2 1

160

Network scale

EMA−SIC Li et al. 180 200

4 3

8 6

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

10

2.5 x 10

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Network scale

EMA−SIC Li et al. 180 200

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 495: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 400 nodes.

Network scale

EMA−SIC Li et al. 180 200

2.5 2 1.5 1 0.5 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

1.5 1 0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

3 2 1 120

140

160

Network scale

EMA−SIC Li et al. 180 200

Network scale

EMA−SIC Li et al. 180 200

8 x 10

6

4

2 100

120

7

1

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

1.5

0.5 100

160

6

4

2 x 10

140

(b) β = 2

5 x 10

0 100

7

1.5

160

7

(c) β = 3

2

0.5 100

160

140

6

(d) β = 4

6

140

3 x 10

(a) β = 1

5

1 100

120

(d) β = 4

6

2

2

0 100

6

0 100

Energy consumption

140

120

2 x 10

Energy consumption

3

4

Fig. 496: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 500 nodes.

6 x 10

(c) β = 3 Energy consumption

Network scale

Network scale

0.5

6

Energy consumption

Energy consumption

6

4 100

160

160

6

7

(b) β = 2

4 x 10

12 x 10

140

EMA−SIC Li et al. 180 200

160

8 x 10

(e) β = 5

1.5

0.5

140

1

6

(a) β = 1

120

120

EMA−SIC Li et al. 180 200

1.5

Energy consumption

5

2.5 x 10

Energy consumption

Energy consumption

10

0 100

1

(f) β = 6

5

0 100

2

2 x 10

140

Network scale

EMA−SIC Li et al. 180 200

6

0 100

Fig. 494: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 300 nodes.

15 x 10

120

(c) β = 3

2 x 10

0.5 100

1

(b) β = 2

3

7

8

160

4 x 10

(d) β = 4 Energy consumption

Energy consumption

10 x 10

140

Network scale

2

0 100

6

5 x 10

Energy consumption

2

120

120

3 x 10

(a) β = 1

6

6

0.5 100

160

(b) β = 2 Energy consumption

Energy consumption

2.5 x 10

140

Network scale

0 100

EMA−SIC Li et al. 180 200

Energy consumption

160

(a) β = 1

5

Energy consumption

140

Network scale

0 100

EMA−SIC Li et al. 180 200

Energy consumption

120

0.5

10

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1

Energy consumption

2

1.5

6

5

EMA−SIC Li et al. 180 200

Energy consumption

4

Energy consumption

6

15 x 10

Energy consumption

6

2 x 10

Energy consumption

Energy consumption

5

8 x 10

3 x 10

7

2.5 2 1.5 1 0.5 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 497: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 600 nodes.

157

(a) β = 1

4

2

140

160

Network scale

EMA−SIC Li et al. 180 200

10 x 10

Energy consumption

Energy consumption

6

120

120

140

160

Network scale

4

Energy consumption

Energy consumption

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1

120

140

160

Network scale

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1 120

(a) β = 1

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

0.5 100

6 4 2 100

120

1.5 1

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

6 4 2 100

120

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 499: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

EMA−SIC Li et al. 180 200

5 x 10 4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

2

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6 x 10

4

2

0 100

120

8 x 10 6 4 2

0 100

120

140

160

Network scale

2.5 x 10

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

6

10

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 7

7

2 1.5 1 0.5 100

140

(b) β = 2

(c) β = 3

3

140

6

0 100

4 x 10

1 100

8

(d) β = 4

3 x 10

7

7

10

6

8

Network scale

EMA−SIC Li et al. 180 200

6

(a) β = 1

10

Energy consumption

Energy consumption

Network scale

160

7

(d) β = 4

2

0.5 100

160

6

(c) β = 3 2.5 x 10

140

EMA−SIC Li et al. 180 200

140

Fig. 500: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 900 nodes.

Energy consumption

Energy consumption

4

120

12 x 10

7

(b) β = 2

6

Network scale

1

Energy consumption

3

12 x 10

0 100

6

4

6

1

(e) β = 5

5 x 10

0 100

8 x 10

160

1.5

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1 0.5

140

EMA−SIC Li et al. 180 200

2

6

2

0 100

120

2.5 x 10

Fig. 498: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 700 nodes.

0 100

2

(f) β = 6

1.5

2

(c) β = 3

3

(e) β = 5

6

3

(b) β = 2

4

0 100

4 x 10

0 100

Network scale

6

7

1

160

4

6

6

2 100

140

8 x 10

(d) β = 4

1.5

2.5 x 10

120

5 x 10

(a) β = 1

8

7

0.5 100

Network scale

0 100

6

(c) β = 3 2 x 10

160

(b) β = 2

6 x 10

0 100

140

Energy consumption

Network scale

120

EMA−SIC Li et al. 180 200

Energy consumption

160

0.5

Energy consumption

140

EMA−SIC Li et al. 180 200

1

Energy consumption

120

0 100

1.5

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1

2

Energy consumption

0.5

2

6

6

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

1

3

2.5 x 10

Energy consumption

1.5

4 x 10

Energy consumption

6

6

Energy consumption

Energy consumption

2 x 10

5 x 10 4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 501: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 1000 nodes.

158

160

0 100

(a) β = 1

400 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

160

1000 500 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2000

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

3000

160

4000 2000

120

140

160

Network scale

Energy consumption

300 200 100

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 502: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 100 nodes.

160

3000 2000 1000 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 15000

8000 6000 4000 2000 0 100

140

Network scale

4000

500

0 100

120

EMA−SIC Li et al. 180 200

(b) β = 2

10000

6000

0 100

140

Network scale

400

(c) β = 3

8000

4000

120

1000

(d) β = 4

5000

120

40 100

EMA−SIC Li et al. 180 200

1500

1500

(c) β = 3

1000

60

500

(a) β = 1

2000

600

0 100

140

Network scale

80

(b) β = 2

800

0 100

120

EMA−SIC Li et al. 180 200

Energy consumption

140

Network scale

100

Energy consumption

120

EMA−SIC Li et al. 180 200

200

100

Energy consumption

50

300

600

120

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

60

Energy consumption

70

40 100

Energy consumption

140

400

Energy consumption

Energy consumption

80

10000

5000

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 503: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 200 nodes.

159

160

Network scale

300 100

120

(a) β = 1

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

2000 1000 0 100

2.5 x 10

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

120

160

Network scale

600 400 100

160

1500 1000 500 100

2

1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 x 10

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 0 100

120

160

(d) β = 4 4

4

1.5 1 0.5 0 100

140

Network scale

EMA−SIC Li et al. 180 200

120

(f) β = 6

140

160

Network scale

EMA−SIC Li et al. 180 200

3 x 10

2

1

0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 504: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 300 nodes.

Fig. 506: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 500 nodes.

250

400

100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

600 400 200 100

120

(a) β = 1

160

Network scale

100 100

1500 1000

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

2000

140

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1200 1000 800 600 400 100

6000 4000 2000

120

(c) β = 3

140

160

Network scale

EMA−SIC Li et al. 180 200

2500 2000 1500 1000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 0 100

0 100

120

140

160

Network scale

(e) β = 5

1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 505: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 400 nodes.

1.5 1 0.5 0 100

140

160

Network scale

EMA−SIC Li et al. 180 200

4

4

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

EMA−SIC Li et al. 180 200

2

Energy consumption

5000

Energy consumption

10000

120

(d) β = 4

4

2 x 10

Network scale

EMA−SIC Li et al. 180 200

8000

(c) β = 3

3 x 10

160

10000

(d) β = 4

15000

140

(b) β = 2

3000

0 100

120

(a) β = 1

8000

120

200

(b) β = 2

2500

500 100

140

EMA−SIC Li et al. 180 200

300

Energy consumption

150

800

1400

Energy consumption

200

Energy consumption

1000

50 100

Energy consumption

140

8000

(c) β = 3

1.5

0.5

120

(b) β = 2

2000

4

(e) β = 5

Energy consumption

140

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

5000

140

2500

(d) β = 4

10000

140

120

EMA−SIC Li et al. 180 200

800

3000

3000

15000

120

100 100

Energy consumption

Energy consumption

1000

120

150

1000

(a) β = 1

4000

(c) β = 3

Energy consumption

Network scale

5000

1500

0 100

160

200

(b) β = 2

2000

500 100

140

EMA−SIC Li et al. 180 200

250

Energy consumption

140

400

Energy consumption

120

500

Energy consumption

50 100

EMA−SIC Li et al. 180 200

600

Energy consumption

100

700

Energy consumption

150

1200

Energy consumption

300

800

Energy consumption

Energy consumption

200

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 507: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 600 nodes.

160

1500

100 100

120

140

160

Network scale

500 100

120

(a) β = 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 100

120

0.5 0 100

160

120

140

160

Network scale

EMA−SIC Li et al. 180 200

500 100

2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

EMA−SIC Li et al. 180 200

(b) β = 2

2000

1000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

10000 8000 6000 4000 2000 100

120

160

Network scale

EMA−SIC Li et al. 180 200

4

3 x 10

2

1

0 100

140

(d) β = 4

4

3

0 100

120

(c) β = 3

4 x 10

(e) β = 5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

5 x 10 4 3 2 1

0 100

120

(e) β = 5

(f) β = 6

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 508: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 700 nodes.

Fig. 510: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 900 nodes.

500

600

100 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1000

500 100

120

(a) β = 1

Network scale

200 100 100

120

140

160

Network scale

8000 6000 4000 2000 100

120

(c) β = 3

0.5 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

140

160

Network scale

EMA−SIC Li et al. 180 200

4 3 2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2000 1000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 509: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

8000 6000 4000 2000 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 4

3 x 10

2

1

0 100

140

10000

4

1

120

(c) β = 3

5 x 10

0 100

500 100

(b) β = 2

3000

Energy consumption

1

Energy consumption

2

Network scale

4000

4

1.5

160

12000

(d) β = 4

4

140

1000

5000

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

2000

1000 100

120

EMA−SIC Li et al. 180 200

1500

(a) β = 1

10000

3000

0 100

160

300

(b) β = 2

4000

2.5 x 10

140

EMA−SIC Li et al. 180 200

400

Energy consumption

200

1500

Energy consumption

300

2000

500

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

400

Energy consumption

2000

Energy consumption

Energy consumption

140

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

1

Energy consumption

Energy consumption

2

Network scale

3000

4

1.5

160

12000

(d) β = 4

4

140

1000

4000

Energy consumption

2000

Energy consumption

Energy consumption

2500

1500

120

EMA−SIC Li et al. 180 200

1500

(a) β = 1

8000

(c) β = 3

Energy consumption

Network scale

100 100

10000

3000

1000 100

Energy consumption

160

200

(b) β = 2

3500

2.5 x 10

140

EMA−SIC Li et al. 180 200

300

Energy consumption

EMA−SIC Li et al. 180 200

400

Energy consumption

200

1000

2000

500

Energy consumption

300

600

Energy consumption

400

Energy consumption

Energy consumption

500

5 x 10 4 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 511: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 1000 nodes.

161

10000

2500 2000 120

140

160

Network scale

6000 4000 2000 100

120

160

1 0.8 0.6 100

120

(a) β = 1

1.4 1.2 120

140

160

Network scale

EMA−SIC Li et al. 180 200

6 x 10 5 4 3 2

1 100

120

(c) β = 3 4

10 8 6 4 120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

3.5 x 10

140

160

Network scale

EMA−SIC Li et al. 180 200

(b) β = 2

4

3

2.5

2 100

120

(d) β = 4

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

140

Network scale

1.2

2.5 x 10

160

Network scale

8 6 4 2 100

2

1

120

140

160

Network scale

120

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 512: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 100 nodes.

2.5 x 10

160

Network scale

EMA−SIC Li et al. 180 200

5

5

2 1.5 1 0.5 100

140

(d) β = 4

5

0.5

4

(c) β = 3

1.5

0 100

140

EMA−SIC Li et al. 180 200

10 x 10

Energy consumption

1.6

2 100

120

EMA−SIC Li et al. 180 200

1.4

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

1.8

12 x 10

2000

4

4

4

1 100

Network scale

4000

0 100

Energy consumption

2 x 10

160

6000

(b) β = 2 Energy consumption

Energy consumption

(a) β = 1

140

EMA−SIC Li et al. 180 200

Energy consumption

1500 100

EMA−SIC Li et al. 180 200

8000

Energy consumption

3000

1.6 x 10

8000

Energy consumption

3500

Energy consumption

Energy consumption

4000

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 513: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 200 nodes.

162

Network scale

120

(a) β = 1

4

3

140

160

Network scale

EMA−SIC Li et al. 180 200

14 x 10

Energy consumption

Energy consumption

4

120

8 6

1.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 3 2 120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

3 2.5 2 1.5 1 100

120

4

120

140

160

Network scale

Network scale

EMA−SIC Li et al. 180 200

0.5 100

120

140

160

Network scale

140

160

Network scale

(e) β = 5

Network scale

1 0.5 100

120

EMA−SIC Li et al. 180 200

10 x 10

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 515: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 400 nodes.

Network scale

EMA−SIC Li et al. 180 200

8 6 4 2 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4

0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

5 x 10 4 3 2

1 100

120

10 x 10

4

8 6 4 2 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2.5 x 10

4 3

120

140

160

Network scale

(e) β = 5

Network scale

EMA−SIC Li et al. 180 200

2 1.5 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

5

1 100

160

5

0.5 100

6 x 10

2

140

(b) β = 2

5

4

160

5

(c) β = 3

6

140

(f) β = 6

1

8 x 10

0 100

160

2

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

2

140

EMA−SIC Li et al. 180 200

1.5

5

3

1.5

4

(d) β = 4

4 x 10

5

(a) β = 1

1

5

Energy consumption

160

1.5

(c) β = 3

120

2

120

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

3

2.5 x 10

Energy consumption

5

Network scale

160

Fig. 516: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 500 nodes.

5

6

1 100

140

EMA−SIC Li et al. 180 200

2 x 10

Energy consumption

Energy consumption

4

160

4

(b) β = 2

7 x 10

140

140

2

5 x 10

4

(a) β = 1

3

120

EMA−SIC Li et al. 180 200

2.5 x 10

(e) β = 5

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1 0.5

2 100

4

1 100

Fig. 514: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 300 nodes.

0 100

5

(f) β = 6

1.5

120

(c) β = 3

5

3.5 x 10

1 100

5

1 100

4

2

(b) β = 2

6

3 100

6 x 10

(e) β = 5

2 x 10

Network scale

7

(d) β = 4

2.5

160

3

4

10

4 100

140

8 x 10

5

1

120

4 x 10

(a) β = 1

12

5

0.5 100

Network scale

0 100

4

(c) β = 3 3 x 10

160

(b) β = 2

5 x 10

2 100

140

Energy consumption

0.5 100

EMA−SIC Li et al. 180 200

Energy consumption

160

0.5

Energy consumption

140

EMA−SIC Li et al. 180 200

Energy consumption

120

1

1

Energy consumption

2000 100

EMA−SIC Li et al. 180 200

1.5

Energy consumption

4000

1.5

4

4

Energy consumption

6000

2

2 x 10

EMA−SIC Li et al. 180 200

Energy consumption

8000

Energy consumption

Energy consumption

10000

4

Energy consumption

2.5 x 10

12000

10 x 10

5

8 6 4 2 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 517: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 600 nodes.

163

120

(a) β = 1 4

4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 x 10 2.5 2 1.5 1 100

120

(c) β = 3

5 4 3

160

Network scale

EMA−SIC Li et al. 180 200

12 x 10

Energy consumption

Energy consumption

5

140

140

160

Network scale

EMA−SIC Li et al. 180 200

12 x 10

6 4 100

8 6 4 120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

6 4 2

0 100

120

6

140

160

Network scale

2 1.5 1 100

5 4 3 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

6

15 x 10

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1.5 1 100

120

Network scale

15 x 10

10

5

0 100

120

120

140

160

Network scale

(f) β = 6

Fig. 519: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 800 nodes.

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8 x 10

6

4

2 100

120

10 8 6 120

140

160

Network scale

EMA−SIC Li et al. 180 200

Network scale

EMA−SIC Li et al. 180 200

4 x 10

3

2

1 100

120

6

4

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

8 x 10

2 100

160

5

4

12

4 100

140

(b) β = 2

5

EMA−SIC Li et al. 180 200

160

(f) β = 6

3

14 x 10

5

5

140

Network scale

EMA−SIC Li et al. 180 200

5

(c) β = 3

10

0 100

3

4

(d) β = 4

7 x 10

160

EMA−SIC Li et al. 180 200

4 x 10

0 100

Energy consumption

120

EMA−SIC Li et al. 180 200

5

(a) β = 1

2.5

5

Energy consumption

Network scale

3 x 10

(c) β = 3

2 100

160

Energy consumption

8

Energy consumption

Energy consumption

10

2 100

140

EMA−SIC Li et al. 180 200

5

4

4

140

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

6

120

160

2.5

8 x 10

(b) β = 2

4

160

EMA−SIC Li et al. 180 200

3.5 x 10

4

8 x 10

(a) β = 1 12 x 10

140

Network scale

140

Fig. 520: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 900 nodes.

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1

120

(e) β = 5

4

0 100

120

2 100

Fig. 518: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 700 nodes.

2

2 100

(b) β = 2

8

(f) β = 6

3

4

(c) β = 3

10

4

Network scale

6

5

(e) β = 5

4 x 10

160

10

5

2 100

140

4

(d) β = 4

6 x 10

120

120

8 x 10

(a) β = 1 Energy consumption

6

2 100

Network scale

0 100

5

8

2 100

160

(b) β = 2 Energy consumption

Energy consumption

10 x 10

140

EMA−SIC Li et al. 180 200

Energy consumption

Network scale

1 100

1

Energy consumption

160

EMA−SIC Li et al. 180 200

2

Energy consumption

140

2

3

Energy consumption

120

3

Energy consumption

0 100

EMA−SIC Li et al. 180 200

4

4

4 x 10

EMA−SIC Li et al. 180 200

Energy consumption

1

5

Energy consumption

2

6 x 10

Energy consumption

4

4

Energy consumption

Energy consumption

4

3 x 10

15 x 10

5

10

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 521: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 1000 nodes.

164

1 120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 2

1 100

120

(a) β = 1

2

0 100

2 x 10

4

140

160

Network scale

EMA−SIC Li et al. 180 200

1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

6

120

3

2

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Network scale

10 8 6 4 2 100

6 5 4

120

140

160

Network scale

120

1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

3 x 10

(f) β = 6

Fig. 522: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 100 nodes.

5 4

140

160

Network scale

(e) β = 5

Network scale

EMA−SIC Li et al. 180 200

2.5 2 1.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

6

120

160

6

1 100

7 x 10

3 100

140

(b) β = 2

6

3

5

(c) β = 3

7 x 10

2 100

160

1.5

6

Energy consumption

6

140

6

(d) β = 4

4 x 10

120

120

EMA−SIC Li et al. 180 200

12 x 10

(a) β = 1

6

(c) β = 3 Energy consumption

Network scale

2 x 10

Energy consumption

Energy consumption

5

1 100

160

4

(b) β = 2

8 x 10

2 100

140

EMA−SIC Li et al. 180 200

Energy consumption

0.5 100

4

Energy consumption

1.5

5

Energy consumption

2

6 x 10

EMA−SIC Li et al. 180 200

Energy consumption

2.5

5

6 x 10

Energy consumption

5

5

Energy consumption

Energy consumption

3 x 10

14 x 10

6

12 10 8 6 4 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 523: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 200 nodes.

165

120

2 1.5 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4.5 x 10 4

3 2.5 120

Network scale

EMA−SIC Li et al. 180 200

6

120

1 100

140

160

Network scale

EMA−SIC Li et al. 180 200

1

0.5 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

1

120

(a) β = 1

0 100

2 1

140

160

Network scale

8 6 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

10

4 3

120

140

160

Network scale

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

2 1.5 1 100

120

1 0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 525: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 400 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

4 3 2

120

140

160

Network scale

1.8 x 10

EMA−SIC Li et al. 180 200

10 x 10

1.4 1.2

120

140

160

Network scale

(e) β = 5

Network scale

EMA−SIC Li et al. 180 200

8 6 4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

1.6

0.8 100

160

6

2 100

7

1

140

(b) β = 2

5 x 10

1 100

7

1.5

140

(f) β = 6

(c) β = 3

2

0.5 100

Network scale

Network scale

EMA−SIC Li et al. 180 200

6

(d) β = 4

12

160

160

7

(a) β = 1

5

2.5 x 10

140

140

2.5

6

2 100

6

120

3 x 10

6

0 100

Energy consumption

120

EMA−SIC Li et al. 180 200

3 100

(d) β = 4

1.5

6 x 10

(c) β = 3 Energy consumption

Network scale

EMA−SIC Li et al. 180 200

2 x 10

Energy consumption

3

4 100

160

4

Fig. 526: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 500 nodes.

6

Energy consumption

Energy consumption

6

0 100

120

(b) β = 2

4 x 10

14 x 10

140

5

(e) β = 5

1.5

0 100

Network scale

0.5

6

0.5

160

1

Energy consumption

5

Energy consumption

Energy consumption

10

2.5 x 10

160

6

7

(f) β = 6

5

140

EMA−SIC Li et al. 180 200

1.5

Fig. 524: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 300 nodes.

0 100

120

2 x 10

140

Network scale

7 x 10

(c) β = 3

1.5

(e) β = 5

15 x 10

2

2 x 10

Energy consumption

Energy consumption

160

120

EMA−SIC Li et al. 180 200

6

3

7

8

4 100

140

1

(b) β = 2

4 x 10

(d) β = 4

6

160

6

3.5

2 100

140

Network scale

2

(a) β = 1

6

(c) β = 3 10 x 10

120

Energy consumption

6

0.5 100

160

(b) β = 2 Energy consumption

Energy consumption

2.5 x 10

140

Network scale

3 x 10

0 100

Energy consumption

160

0 100

EMA−SIC Li et al. 180 200

Energy consumption

140

Network scale

(a) β = 1

5

Energy consumption

120

0 100

EMA−SIC Li et al. 180 200

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1 0.5

10

Energy consumption

2

1.5

6

5

EMA−SIC Li et al. 180 200

Energy consumption

4

Energy consumption

6

15 x 10

Energy consumption

6

2 x 10

Energy consumption

Energy consumption

5

8 x 10

3.5 x 10

7

3 2.5 2 1.5 1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 527: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 600 nodes.

166

(a) β = 1

5 4 3

120

140

160

Network scale

EMA−SIC Li et al. 180 200

10 x 10

Energy consumption

Energy consumption

6

1 100

4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

120

140

160

Network scale

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1 120

(a) β = 1

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

0.5 100

6 4 2 100

120

1.5 1

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

6 4 100

120

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 529: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

EMA−SIC Li et al. 180 200

5 x 10 4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

2

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6 x 10

4

2

0 100

120

8 x 10 6 4 2

0 100

120

140

160

Network scale

3 x 10

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

14 x 10

6

12 10 8 6 4 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 7

7

2.5 2 1.5 1 100

140

(b) β = 2

(c) β = 3

3

140

6

0 100

4 x 10

1 100

8

(d) β = 4

3 x 10

7

7

10

6

8

Network scale

EMA−SIC Li et al. 180 200

6

(a) β = 1

10

Energy consumption

Energy consumption

Network scale

160

7

(d) β = 4

2

0.5 100

160

6

(c) β = 3 2.5 x 10

140

EMA−SIC Li et al. 180 200

140

Fig. 530: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 900 nodes.

Energy consumption

Energy consumption

4

120

12 x 10

7

(b) β = 2

6

Network scale

1

Energy consumption

3

12 x 10

0 100

6

4

6

1

(e) β = 5

5 x 10

0 100

8 x 10

160

1.5

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1 0.5

140

EMA−SIC Li et al. 180 200

2

6

2

0 100

120

2.5 x 10

Fig. 528: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 700 nodes.

0 100

2

(f) β = 6

1.5

2

(c) β = 3

3

(e) β = 5

6

3

(b) β = 2

4

0 100

4 x 10

1 100

Network scale

6

7

1.5

160

4

6

6

2 100

140

8 x 10

(d) β = 4

2

2.5 x 10

120

5 x 10

(a) β = 1

8

7

0.5 100

Network scale

0 100

6

(c) β = 3 2.5 x 10

160

(b) β = 2

6 x 10

2

140

Energy consumption

Network scale

120

EMA−SIC Li et al. 180 200

Energy consumption

160

0.5

Energy consumption

140

EMA−SIC Li et al. 180 200

1

Energy consumption

120

0 100

1.5

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1

2

Energy consumption

0.5

2

6

6

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

1

3

2.5 x 10

Energy consumption

1.5

4 x 10

Energy consumption

6

6

Energy consumption

Energy consumption

2 x 10

5 x 10 4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 531: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 1000 nodes.

167

160

(a) β = 1

200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

160

1000 500 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1000

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

2000

160

4000 2000

120

140

160

Network scale

Energy consumption

200 100 0 100

EMA−SIC Li et al. 180 200

400 200 120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 532: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 100 nodes.

160

3000 2000 1000 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 15000

6000 4000 2000 0 100

140

Network scale

4000

600

0 100

120

EMA−SIC Li et al. 180 200

(b) β = 2

8000

6000

0 100

140

Network scale

300

(c) β = 3

8000

3000

120

800

(d) β = 4

4000

140

20 100

EMA−SIC Li et al. 180 200

1000

1500

(c) β = 3

120

40

400

(a) β = 1

2000

400

0 100

140

Network scale

60

(b) β = 2

600

0 100

120

EMA−SIC Li et al. 180 200

Energy consumption

140

Network scale

50 100

Energy consumption

120

EMA−SIC Li et al. 180 200

100

80

Energy consumption

20

150

500

100

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

40

200

Energy consumption

60

0 100

Energy consumption

120

250

Energy consumption

Energy consumption

80

10000

5000

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 533: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 200 nodes.

168

140

160

Network scale

0 100

120

(a) β = 1

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

3000 2000 1000 0 100

120

(c) β = 3

4000

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Network scale

5000

120

140

160

Network scale

200 120

EMA−SIC Li et al. 180 200

6000

1000 500 120

140

160

Network scale

EMA−SIC Li et al. 180 200

4000

2000

0 100

120

5000

120

(f) β = 6

140

160

Network scale

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

10000

0 100

160

(b) β = 2

1500

0 100

140

Network scale

EMA−SIC Li et al. 180 200

EMA−SIC Li et al. 180 200

2.5 x 10

4

2 1.5 1 0.5 0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 534: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 300 nodes.

Fig. 536: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 500 nodes.

250

400

100 50 120

140

160

Network scale

EMA−SIC Li et al. 180 200

400 200 0 100

120

(a) β = 1

160

Network scale

500

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

1000

140

100 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

3000 2000

0 100

120

(c) β = 3

140

160

Network scale

0 100

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

1500 1000 500 0 100

2 x 10

120

140

160

Network scale

EMA−SIC Li et al. 180 200

0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 535: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 400 nodes.

10000

5000

0 100

160

Network scale

EMA−SIC Li et al. 180 200

2000

120

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

EMA−SIC Li et al. 180 200

Energy consumption

1

140

4000

0 100

15000

1.5

0 100

120

(c) β = 3 Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

5000

200

6000

4

10000

400

(b) β = 2

2000

(d) β = 4

15000

600

0 100

2500

4000

1000

800

(a) β = 1

5000

1500

120

200

(b) β = 2

2000

0 100

140

EMA−SIC Li et al. 180 200

1000

300

Energy consumption

150

600

Energy consumption

800

200

0 100

Energy consumption

400

(c) β = 3

10000

0 100

Energy consumption

Energy consumption

160

Energy consumption

Energy consumption

6000

160

15000

(e) β = 5

Energy consumption

140

EMA−SIC Li et al. 180 200

15000

8000

140

Network scale

2000

(d) β = 4

10000

2000

120

600

0 100

2500

Energy consumption

Energy consumption

500

120

0 100

EMA−SIC Li et al. 180 200

800

(a) β = 1

4000

1000

0 100

160

100

(b) β = 2

1500

0 100

140

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

0 100

200

Energy consumption

50

400

200

Energy consumption

100

600

1000

Energy consumption

150

Energy consumption

300

800

Energy consumption

Energy consumption

200

2.5 x 10

4

2 1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 537: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 600 nodes.

169

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

500

0 100

120

(a) β = 1

140

160

Network scale

EMA−SIC Li et al. 180 200

4000 2000 0 100

120

0 100

160

120

140

160

Network scale

EMA−SIC Li et al. 180 200

0 100

Network scale

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 x 10

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8000 6000 4000 2000 0 100

120

(c) β = 3

2

0 100

120

(b) β = 2

1000

0 100

3 x 10

(e) β = 5

160

4

1.5 1 0.5 0 100

140

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

4

120

(f) β = 6

140

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 538: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 700 nodes.

Fig. 540: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 900 nodes.

500

600

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

500

0 100

120

(a) β = 1

Network scale

0 100

120

140

160

Network scale

6000 4000 2000 0 100

120

(c) β = 3

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

3 2 1

120

140

160

Network scale

120

1000

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 539: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 800 nodes.

2.5 x 10

120

140

160

Network scale

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

6000 4000 2000 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 4

4

2 1.5 1 0.5 0 100

140

8000

(c) β = 3

4 x 10

0 100

0 100

(b) β = 2

2000

0 100

Energy consumption

0.5

Energy consumption

1

Network scale

3000

4

1.5

160

10000

(d) β = 4

4

140

500

4000

Energy consumption

EMA−SIC Li et al. 180 200

Energy consumption

1000

0 100

120

EMA−SIC Li et al. 180 200

1000

(a) β = 1

8000

2000

0 100

160

200

(b) β = 2

3000

2 x 10

140

EMA−SIC Li et al. 180 200

Energy consumption

100

400

Energy consumption

200

1000

1500

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

300

Energy consumption

1500

400

Energy consumption

Energy consumption

140

EMA−SIC Li et al. 180 200

Energy consumption

0.5

Energy consumption

Energy consumption

1

Network scale

2000

4

1.5

160

10000

(d) β = 4

4

140

500

3000

Energy consumption

1000

Energy consumption

Energy consumption

2000

120

120

EMA−SIC Li et al. 180 200

(a) β = 1

6000

(c) β = 3

Energy consumption

Network scale

0 100

8000

0 100

Energy consumption

160

200

(b) β = 2

3000

2 x 10

140

EMA−SIC Li et al. 180 200

1000

Energy consumption

100

400

Energy consumption

200

1000

Energy consumption

300

1500

Energy consumption

600

1500

400

Energy consumption

Energy consumption

500

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 541: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 1000 nodes.

170

8000

1000

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4000 2000 0 100

120

1 0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(b) β = 2

4

4 3 2

1 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4

3 x 10

2

1

0 100

120

(d) β = 4

4

140

160

Network scale

EMA−SIC Li et al. 180 200

8 x 10 6 4 2

0 100

6 4 2 120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

2 x 10 1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 542: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 100 nodes.

2 x 10

160

Network scale

EMA−SIC Li et al. 180 200

5

5

1.5 1 0.5 0 100

140

(d) β = 4

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

8

120

(c) β = 3

5

Energy consumption

Energy consumption

120

1

(a) β = 1

5 x 10

(c) β = 3

0 100

0 100

EMA−SIC Li et al. 180 200

1.5

Energy consumption

1.5

10 x 10

2000

4

4

4

0 100

Network scale

4000

Energy consumption

2 x 10

160

6000

(b) β = 2 Energy consumption

Energy consumption

(a) β = 1

140

EMA−SIC Li et al. 180 200

Energy consumption

0 100

6000

Energy consumption

2000

2 x 10

8000

Energy consumption

3000

Energy consumption

Energy consumption

4000

3 x 10

2

1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 543: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 200 nodes.

171

160

Network scale

120

(a) β = 1

3 2 1

140

160

Network scale

EMA−SIC Li et al. 180 200

10 x 10

Energy consumption

6 4 2 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 3 2 1 120

140

160

Network scale

EMA−SIC Li et al. 180 200

0 100

140

160

Network scale

2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

120

4

3 2

140

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

Energy consumption

4

120

160

Network scale

2

1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

3 2 1

0 100

140

160

Network scale

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 545: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 400 nodes.

160

3 2 1

0 100

120

6 4 2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

2 x 10

5

1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 5

4 x 10 3 2 1

0 100

140

(b) β = 2

5

5 x 10 4

140

Network scale

4

(c) β = 3 Energy consumption

1

Energy consumption

2

120

EMA−SIC Li et al. 180 200

5 x 10

8 x 10

0 100

5

120

120

(d) β = 4

5

2

4

(c) β = 3 3 x 10

4

4

1.5

4

5

0 100

6

(a) β = 1

10

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

4

(b) β = 2

5 x 10

1

140

EMA−SIC Li et al. 180 200

160

8 x 10

0 100

Energy consumption

1

140

(d) β = 4

3

2.5 x 10

Energy consumption

2

0 100

120

Fig. 546: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 500 nodes.

4 x 10

(a) β = 1 Energy consumption

EMA−SIC Li et al. 180 200

1 0.5

(e) β = 5

3

Network scale

EMA−SIC Li et al. 180 200

5

0 100

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

0.5

Energy consumption

Network scale

160

1.5

4

4

1

0 100

160

4 x 10

(f) β = 6

1.5

0 100

140

140

5

(c) β = 3

Fig. 544: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 300 nodes.

0 100

120

2 x 10

5

(e) β = 5

2 x 10

120

(b) β = 2

2

0 100

5 x 10

0 100

1

0 100

Energy consumption

0 100

Energy consumption

Energy consumption

1

Network scale

4

5

2

160

2

6 x 10

(d) β = 4

1.5

140

3

4

8

5

0.5

120

4 x 10

(a) β = 1

4

(c) β = 3 2.5 x 10

Network scale

0 100

EMA−SIC Li et al. 180 200

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

4

120

160

0.5

(b) β = 2

4 x 10

0 100

140

EMA−SIC Li et al. 180 200

Energy consumption

0 100

Energy consumption

140

0.5

1

Energy consumption

120

EMA−SIC Li et al. 180 200

1

4

4

1.5

Energy consumption

0 100

2 1.5

2 x 10

Energy consumption

5000

Energy consumption

Energy consumption

10000

4

Energy consumption

2.5 x 10

15000

8 x 10 6 4 2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 547: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 600 nodes.

172

Network scale

0 100

120

(a) β = 1 4

6 4 2 120

140

160

Network scale

EMA−SIC Li et al. 180 200

1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

3 2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

0 100

6 4 2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

160

Network scale

4 2

0 100

120

6 4

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

8

2.5 x 10

1 0.5

3 2

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

4

120

120

120

140

160

Network scale

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

2 1.5 1 0.5 0 100

120

10 x 10

4

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 549: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

5

8 6 4 2 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4

3 2 1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

8 x 10 6 4 2

0 100

120

15 x 10

4

10

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2.5 x 10

2

140

160

Network scale

(e) β = 5

Network scale

EMA−SIC Li et al. 180 200

2 1.5 1 0.5 120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

4

120

160

5

0 100

6 x 10

0 100

140

(b) β = 2

5

6

140

(f) β = 6

4 x 10

0 100

5

2

5

(c) β = 3

8

0 100

140

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

2

(d) β = 4 10 x 10

2.5 x 10

(a) β = 1

2

5

0 100

Network scale

1.5

(c) β = 3

1

160

5

0 100

5 x 10

Network scale

EMA−SIC Li et al. 180 200

4

(b) β = 2

4

2

140

EMA−SIC Li et al. 180 200

Energy consumption

6

(a) β = 1 10 x 10

160

160

(b) β = 2

4

8 x 10

Energy consumption

140

140

140

Fig. 550: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 900 nodes.

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1

120

(e) β = 5

4

2

2

0 100

6 x 10

0 100

Fig. 548: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 700 nodes.

0 100

120

(f) β = 6

3

4

5

0 100

4

6

(c) β = 3

8 x 10

(e) β = 5

4 x 10

Network scale

5

Energy consumption

4

160

10

5

Energy consumption

5

140

4

(d) β = 4

5 x 10

1

120

8 x 10

(a) β = 1

2 x 10

(c) β = 3 Energy consumption

Network scale

0 100

5

8

0 100

160

(b) β = 2 Energy consumption

Energy consumption

10 x 10

140

EMA−SIC Li et al. 180 200

Energy consumption

160

1

Energy consumption

140

2

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

0 100

EMA−SIC Li et al. 180 200

2

3

Energy consumption

1

4

4

4 x 10

EMA−SIC Li et al. 180 200

Energy consumption

2

4

6 x 10

Energy consumption

4

Energy consumption

Energy consumption

4

3 x 10

10 x 10

5

8 6 4 2 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 551: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 1000 nodes.

173

140

160

Network scale

0 100

120

(a) β = 1

6 4 2

140

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

Energy consumption

Energy consumption

5

120

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2 x 10

1

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

2

Network scale

Energy consumption

5

0 100

120

4 3 2 120

140

160

Network scale

EMA−SIC Li et al. 180 200

0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

1

0 100

120

(f) β = 6

Fig. 552: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 100 nodes.

4

2

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

6 x 10

0 100

Network scale

EMA−SIC Li et al. 180 200

3 x 10

6

5

160

6

1

0 100

140

(b) β = 2

(c) β = 3

6 x 10

1 100

160

1.5

6

3

140

10

6

(d) β = 4

4 x 10

120

120

EMA−SIC Li et al. 180 200

5

(a) β = 1

10

6

Energy consumption

Network scale

0 100

5

(c) β = 3

0 100

160

2

(b) β = 2

8 x 10

0 100

140

EMA−SIC Li et al. 180 200

Energy consumption

120

2

Energy consumption

0 100

EMA−SIC Li et al. 180 200

4

15 x 10

Energy consumption

1

4

6 x 10

EMA−SIC Li et al. 180 200

Energy consumption

2

5

6 x 10

Energy consumption

5

Energy consumption

Energy consumption

5

3 x 10

10 x 10

6

8 6 4 2 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 553: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 200 nodes.

174

Network scale

0 100

120

(a) β = 1

1 0.5 120

140

160

Network scale

EMA−SIC Li et al. 180 200

3 2 1

0 100

120

(c) β = 3

6 4 2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

Energy consumption

Energy consumption

6

140

160

Network scale

EMA−SIC Li et al. 180 200

5

120

140

160

Network scale

120

140

160

Network scale

EMA−SIC Li et al. 180 200

2

1

0 100

120

0 100

2 1

140

160

Network scale

EMA−SIC Li et al. 180 200

4

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

1.5 1 0.5 0 100

120

140

160

Network scale

2

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 555: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 400 nodes.

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

EMA−SIC Li et al. 180 200

2 x 10

7

1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

6

1 0.5

120

140

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

4 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Network scale

EMA−SIC Li et al. 180 200

8 x 10 6 4 2

0 100

120

6

5

120

140

160

Network scale

(e) β = 5

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4

10

0 100

160

6

5 x 10

15 x 10

140

(b) β = 2

(c) β = 3

2 x 10

Network scale

EMA−SIC Li et al. 180 200

4

(a) β = 1

Energy consumption

6

Network scale

1.5

0 100

7

8

160

6

(d) β = 4

6

140

6

0 100

Energy consumption

Energy consumption

Network scale

6 x 10

(c) β = 3

2

120

2 x 10

Energy consumption

3

0 100

160

160

Fig. 556: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 500 nodes.

6

Energy consumption

Energy consumption

6

10 x 10

140

EMA−SIC Li et al. 180 200

140

6

(e) β = 5

1.5

0.5

120

8 x 10

6

(b) β = 2

4 x 10

120

120

EMA−SIC Li et al. 180 200

5

6

(a) β = 1

0 100

1

10

Energy consumption

EMA−SIC Li et al. 180 200

2.5 x 10

Energy consumption

Energy consumption

5

0 100

2

15 x 10

Fig. 554: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 300 nodes.

10

0 100

(b) β = 2

3

(f) β = 6

5

1

(c) β = 3

10

0 100

2

6

0 100

6

(e) β = 5

15 x 10

Network scale

4 x 10

(d) β = 4

8 x 10

160

6

4 x 10

Energy consumption

2 1.5

140

3 x 10

(a) β = 1

6

6

0 100

160

120

(b) β = 2 Energy consumption

Energy consumption

2.5 x 10

140

Network scale

0 100

EMA−SIC Li et al. 180 200

Energy consumption

160

5

Energy consumption

140

EMA−SIC Li et al. 180 200

Energy consumption

120

0.5

10

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1

Energy consumption

2

1.5

6

5

EMA−SIC Li et al. 180 200

Energy consumption

4

Energy consumption

6

15 x 10

Energy consumption

6

2 x 10

Energy consumption

Energy consumption

5

8 x 10

2.5 x 10

7

2 1.5 1 0.5 0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 557: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 600 nodes.

175

(a) β = 1

4

2

140

160

Network scale

EMA−SIC Li et al. 180 200

10 x 10

Energy consumption

Energy consumption

6

120

120

140

160

Network scale

4 2

Energy consumption

Energy consumption

EMA−SIC Li et al. 180 200

120

140

160

Network scale

EMA−SIC Li et al. 180 200

0 100

1

120

140

160

Network scale

140

160

Network scale

EMA−SIC Li et al. 180 200

2 1 120

(a) β = 1

2

120

140

160

Network scale

EMA−SIC Li et al. 180 200

Energy consumption

0 100

0 100

120

1 0.5

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

0 100

120

120

140

160

Network scale

EMA−SIC Li et al. 180 200

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 559: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 800 nodes.

160

Network scale

EMA−SIC Li et al. 180 200

140

160

Network scale

EMA−SIC Li et al. 180 200

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

2

1

120

140

160

Network scale

EMA−SIC Li et al. 180 200

6 x 10

4

2

0 100

120

8 x 10 6 4 2

0 100

120

140

160

Network scale

2 x 10

EMA−SIC Li et al. 180 200

160

Network scale

EMA−SIC Li et al. 180 200

15 x 10

6

10

5

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(d) β = 4 7

7

1.5 1 0.5 0 100

140

(b) β = 2

(c) β = 3

2

140

6

0 100

3 x 10

0 100

5

(d) β = 4

3 x 10

7

7

10

6

5

Network scale

EMA−SIC Li et al. 180 200

6

(a) β = 1

10

Energy consumption

Energy consumption

Network scale

160

7

(d) β = 4

1.5

0 100

160

6

(c) β = 3 2 x 10

140

EMA−SIC Li et al. 180 200

140

Fig. 560: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 900 nodes.

Energy consumption

Energy consumption

4

120

15 x 10

7

(b) β = 2

6

Network scale

0.5

Energy consumption

3

15 x 10

0 100

6

4

6

1

(e) β = 5

5 x 10

0 100

8 x 10

160

1

Energy consumption

120

EMA−SIC Li et al. 180 200

Energy consumption

Energy consumption

1 0.5

140

EMA−SIC Li et al. 180 200

1.5

6

2

0 100

120

2 x 10

Fig. 558: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 700 nodes.

0 100

2

(f) β = 6

1.5

2

(c) β = 3

2

(e) β = 5

6

3

(b) β = 2

4

3 x 10

0 100

Network scale

6

7

5

160

4

6

6

0 100

140

8 x 10

(d) β = 4

10

2.5 x 10

120

5 x 10

(a) β = 1

8

6

0 100

Network scale

0 100

6

(c) β = 3 15 x 10

160

(b) β = 2

6 x 10

0 100

140

Energy consumption

Network scale

120

EMA−SIC Li et al. 180 200

Energy consumption

160

0.5

Energy consumption

140

EMA−SIC Li et al. 180 200

1

Energy consumption

120

0 100

1.5

Energy consumption

0 100

EMA−SIC Li et al. 180 200

1

2

Energy consumption

0.5

2

6

6

120

140

160

Network scale

(e) β = 5

EMA−SIC Li et al. 180 200

Energy consumption

1

3

2.5 x 10

Energy consumption

1.5

4 x 10

Energy consumption

6

6

Energy consumption

Energy consumption

2 x 10

4 x 10 3 2 1

0 100

120

140

160

Network scale

EMA−SIC Li et al. 180 200

(f) β = 6

Fig. 561: Impact of network scale on the energy consumption (joule): a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 1000 nodes.

176

180

200

120

(a) β = 1

140

160

180

200

5

0 100

120

1 0.5 140

160

Network scale

(e) β = 5

180

200

160

180

200

180

200

2 1

140

160

Network scale

120

1 0.5 120

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 562: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 100 nodes.

200

180

200

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

10

1 0.5

120

180

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

160

9

1.5

0 100

140

Network scale

(b) β = 2

10

120

0 100

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

9

1.5

140

Network scale

9

(c) β = 3 9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

120

120

5

(a) β = 1 Latencyα−1 times energy

2

Network scale

0 100

200

8

4

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

8

120

160

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

1

Latencyα−1 times energy

0 100

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

160

0.5

8

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

1

Latencyα−1 times energy

120

Latencyα−1 times energy

1

0 100

8

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

7

3 x 10 EMA−SIC Cell−AS Li et al. 2

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 563: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 200 nodes.

177

120

(a) β = 1

140

160

Network scale

180

200

140

160

180

200

120

140

160

Network scale

180

2

140

160

Network scale

(e) β = 5

180

140

160

Network scale

180

200

2

0 100

120

140

160

Network scale

180

200

2 160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

120

180

200

120

0 100

160

Network scale

120

180

200

(f) β = 6

Fig. 565: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 400 nodes.

140

160

Network scale

180

200

10

140

160

Network scale

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

180

200

200

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

11

1

120

180

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

160

10

1

120

140

Network scale

(b) β = 2

2

times energy 140

200

1

11

0.5

180

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3

1

160

(f) β = 6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

160

1

0 100

α−1

4

140

Network scale

2

(d) β = 4

6

140

120

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latencyα−1 times energy

1

140

Network scale

(d) β = 4

(a) β = 1

2

120

120

Fig. 566: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 500 nodes.

200

11

Latencyα−1 times energy

Latencyα−1 times energy

180

200

2

0 100

10

(c) β = 3

120

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

180

11

(b) β = 2 Latencyα−1 times energy

Latencyα−1 times energy

120

200

9

10

5

0 100

0 100

160

6 x 10 EMA−SIC Cell−AS Li et al. 4

(e) β = 5

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1 9

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

0.5

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

5

160

1

200

9

8

140

Network scale

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(f) β = 6

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

140

Network scale

10

11

4

120

120

(c) β = 3

Fig. 564: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 300 nodes.

0 100

0 100

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

0 100

(b) β = 2

1

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

1

200

0.5

(d) β = 4

2

180

1.5

10

3

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

0.5 0 100

140

Network scale

10

(c) β = 3

Network scale

120

(a) β = 1

1

10

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

0 100

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

180

10

4

120

160

(b) β = 2

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

5

Latencyα−1 times energy

160

1 0.5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

140

Network scale

1.5

Latency

120

1

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latencyα−1 times energy

4

Latencyα−1 times energy

Latencyα−1 times energy

6

0 100

9

9

8

10 x 10 EMA−SIC 8 Cell−AS Li et al.

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 567: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 600 nodes.

178

120

(a) β = 1

160

180

200

0 100

120

(c) β = 3

140

160

Network scale

180

200

1 140

160

180

200

2

120

140

160

Network scale

(e) β = 5

180

200

160

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

(a) β = 1

4 2

120

140

160

Network scale

180

200

200

120

140

160

180

200

140

160

Network scale

(e) β = 5

180

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 569: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 800 nodes.

200

180

200

1 0.5 0 100

120

140

160

Network scale

160

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

10

5

0 100

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

6 4 2 120

140

160

Network scale

180

200

200

180

200

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(d) β = 4

4 2

120

180

12

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

11

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

5

0 100

140

Network scale

11

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

1.5

(d) β = 4

10

Network scale

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1

0.5

180

11

2

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

180

11

2

120

160

1

(c) β = 3 11

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

11

Latencyα−1 times energy

Latencyα−1 times energy

10

180

4

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

160

9

1

0 100

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

160

Fig. 570: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 900 nodes.

Latencyα−1 times energy

140

Network scale

120

0 100

Latencyα−1 times energy

120

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

2 0 100

140

Network scale

(b) β = 2

4

10

4

120

(e) β = 5

Fig. 568: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 700 nodes.

9

0 100

11

6

(f) β = 6

8 x 10 EMA−SIC Cell−AS 6 Li et al.

200

1

(c) β = 3

4

0 100

180

11

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

160

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

(d) β = 4

3

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

11

Network scale

120

10

5

11

0 100

0 100

Latencyα−1 times energy

140

Network scale

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

200

10

2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

10

120

160

(b) β = 2

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

2

Latencyα−1 times energy

160

4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

0.5

6

Latencyα−1 times energy

120

1

Latencyα−1 times energy

0 100

1.5

Latencyα−1 times energy

2

10

9

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

10

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

9

6 x 10 EMA−SIC Cell−AS Li et al. 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 571: Impact of network scale on the latency-energy tradeoff under α = 3 and uniform distribution with 1000 nodes.

179

140

160

Network scale

180

200

0 100

120

(a) β = 1

1

140

160

Network scale

180

200

0 100

120

160

180

200

4 2 120

140

160

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

180

2 1

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

0 100

α−1

2 0 100

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 572: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 100 nodes.

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

160

180

200

180

200

14

1

0 100

140

Network scale

(d) β = 4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

times energy

4

Latency

Latencyα−1 times energy

120

6

200

5

14

10 x 10 EMA−SIC 8 Cell−AS Li et al.

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

(d) β = 4

160

13

3

120

140

Network scale

(b) β = 2

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

13

1

120

(a) β = 1

6

0 100

140

Network scale

13

(c) β = 3 Latencyα−1 times energy

200

10 x 10 EMA−SIC 8 Cell−AS Li et al.

13

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

180

12

Latencyα−1 times energy

Latencyα−1 times energy

12

120

160

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

2

Latencyα−1 times energy

120

5

Latencyα−1 times energy

1

12

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

0 100

12

11

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

11

5 x 10 EMA−SIC 4 Cell−AS Li et al.

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

Fig. 573: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 200 nodes.

180

13

120

(a) β = 1

0.5

140

160

Network scale

180

200

120

140

160

Network scale

180

200

160

180

200

120

140

160

Network scale

(e) β = 5

180

200

1 140

160

Network scale

180

200

120

(a) β = 1 14

160

180

200

0 100

120

160

180

200

160

(e) β = 5

180

200

4 2

0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 575: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 400 nodes.

200

180

200

1

0 100

120

140

160

Network scale

(d) β = 4 15

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

14

1 0.5 0 100

120

140

160

Network scale

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

1 0.5

120

140

160

Network scale

180

200

180

200

180

200

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(d) β = 4 16

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

160

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

(b) β = 2

15

8 x 10 EMA−SIC Cell−AS 6 Li et al.

180

2

(c) β = 3 Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

15

1

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(d) β = 4

15

180

15

1

200

3

(a) β = 1 Latencyα−1 times energy

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

200

0.5

(c) β = 3

0 100

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

120

160

15

4

0 100

140

Network scale

180

Fig. 576: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 500 nodes.

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

160

14

1

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(e) β = 5

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

2

120

140

Network scale

15

4

0 100

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

120

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

14

13

120

120

(f) β = 6

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

(c) β = 3

1

0 100

200

15

3 x 10 EMA−SIC Cell−AS Li et al. 2

Fig. 574: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 300 nodes.

0 100

0 100

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

5

180

5

15

14

160

14

2

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

(a) β = 1

4

(c) β = 3 Latencyα−1 times energy

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

14

14

120

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

200

1

Latencyα−1 times energy

180

0 100

2

Latencyα−1 times energy

160

0 100

2

3

Latencyα−1 times energy

140

Network scale

2

4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latencyα−1 times energy

120

4

Latencyα−1 times energy

0 100

6

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

13

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 577: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 600 nodes.

181

14

120

(a) β = 1

140

160

Network scale

180

200

4 2 0 100

120

Network scale

140

160

Network scale

180

200

200

1

120

140

160

Network scale

(e) β = 5

180

200

1

140

160

Network scale

180

200

1 0.5 0 100

120

(a) β = 1

1

140

160

Network scale

180

200

0 100

160

(e) β = 5

180

200

120

140

160

Network scale

180

200

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 579: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 800 nodes.

200

180

200

180

200

1 0.5 0 100

120

140

160

Network scale

(d) β = 4

140

160

Network scale

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

15

3 2 1

0 100

120

140

160

Network scale

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(d) β = 4

2

120

180

16

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

160

16

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

2

180

16

16

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

180

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

120

160

Fig. 580: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 900 nodes.

16

1

160

15

(c) β = 3

2

140

Network scale

2

0 100

140

Network scale

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(a) β = 1

5

16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

200

15

2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

15

120

160

120

16

6 x 10 EMA−SIC Cell−AS Li et al. 4

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

0 100

(b) β = 2

14

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

14

1 0.5

(e) β = 5

Fig. 578: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 700 nodes.

120

120

(f) β = 6

3 x 10 EMA−SIC Cell−AS Li et al. 2

200

1.5

(c) β = 3

2

0 100

180

16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latencyα−1 times energy

180

160

2

0 100

Latencyα−1 times energy

160

Latencyα−1 times energy

Latencyα−1 times energy

0.5

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(d) β = 4

1

140

120

15

16

1.5

0 100

0 100

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1

6

(c) β = 3 16

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

120

200

15

1

0 100

180

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

15

120

160

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

1

Latencyα−1 times energy

160

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

5

2

Latencyα−1 times energy

0.5

15

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

Latencyα−1 times energy

Latencyα−1 times energy

1.5

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

14

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 581: Impact of network scale on the latency-energy tradeoff under α = 4 and uniform distribution with 1000 nodes.

182

140

160

Network scale

180

200

1

0 100

120

(a) β = 1

200

120

140

160

Network scale

120

180

200

1

120

(c) β = 3

160

Network scale

180

200

140

160

Network scale

180

200

140

160

Network scale

(e) β = 5

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 582: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 100 nodes.

160

180

200

180

200

180

200

1

120

160

19

5

120

140

Network scale

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

120

120

(b) β = 2

18

2

2 0 100

18

(d) β = 4

4

4

(c) β = 3

17

8 x 10 EMA−SIC Cell−AS 6 Li et al.

6

(a) β = 1

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

17

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

4

17

Latencyα−1 times energy

180

0 100

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

16

16

0 100

140

Network scale

5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

2

17

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

120

Latencyα−1 times energy

1

0 100

16

16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

15

3 x 10 EMA−SIC Cell−AS Li et al. 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 583: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 200 nodes.

183

2

140

160

Network scale

180

200

1 0.5 0 100

120

(c) β = 3

160

180

200

140

160

Network scale

180

200

120

(e) β = 5

140

160

Network scale

180

200

140

160

Network scale

180

200

4 2 120

(a) β = 1

0.5

120

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

200

140

160

180

200

0.5

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

1

0 100

120

Latencyα−1 times energy

0 100

140

160

Network scale

120

180

200

(f) β = 6

Fig. 585: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 400 nodes.

140

160

Network scale

140

160

Network scale

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

19

2

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

5

120

140

160

Network scale

180

200

200

180

200

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(d) β = 4

6 4 2 120

180

20

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

160

20

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

(b) β = 2

(c) β = 3

2

200

0.5

(d) β = 4

20

4 x 10 EMA−SIC Cell−AS 3 Li et al.

180

1

20

19

Network scale

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

(a) β = 1

2

120

200

4

0 100

20

20

Latencyα−1 times energy

180

4

0 100

180

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3

0 100

140

Network scale

180

Fig. 586: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 500 nodes.

19

19

160

1

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

2

120

160

20

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

Latencyα−1 times energy

120

120

140

Network scale

(b) β = 2

18

6

0 100

120

(e) β = 5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

0.5 0 100

0 100

18

1

0 100

(c) β = 3

Fig. 584: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 300 nodes.

18

200

2

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

180

20

5

0 100

160

6 x 10 EMA−SIC Cell−AS Li et al. 4

(d) β = 4 15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

19

2

140

Network scale

19

1.5

19

6 x 10 EMA−SIC Cell−AS Li et al. 4

120

120

(a) β = 1

19

4

0 100

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

18

120

180

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

Latencyα−1 times energy

(a) β = 1

140

Network scale

0.5

Latencyα−1 times energy

120

1

1

Latencyα−1 times energy

200

2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

180

Latencyα−1 times energy

160

3

0 100

Latencyα−1 times energy

140

Network scale

0 100

Latencyα−1 times energy

120

1

Latencyα−1 times energy

0 100

2

19

5 x 10 EMA−SIC 4 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

18

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Latencyα−1 times energy

18

Latencyα−1 times energy

Latencyα−1 times energy

17

6 x 10 EMA−SIC Cell−AS Li et al. 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 587: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 600 nodes.

184

0.5

140

160

180

200

120

(c) β = 3

140

160

Network scale

180

200

160

180

200

times energy α−1

0 100

120

140

160

Network scale

180

200

140

160

Network scale

180

200

5

0 100

120

(a) β = 1

140

160

Network scale

180

200

0 100

120

140

160

Network scale

0.5 140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

180

200

1 140

160

Network scale

180

200

(f) β = 6

Fig. 589: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 800 nodes.

160

180

200

180

200

180

200

4 2

0 100

120

140

160

Network scale

(f) β = 6

1

120

140

160

Network scale

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

120

140

160

Network scale

180

200

200

180

200

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(d) β = 4

3 2 1 120

180

21

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

(b) β = 2

2

0 100

140

Network scale

21

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

2

120

160

21

3

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

120

20

(d) β = 4

1.5

140

Network scale

2

0 100

21

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

(d) β = 4

20

5

(c) β = 3 21

Latencyα−1 times energy

200

20

1

120

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

20

0 100

160

200

21

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

140

Network scale

Latencyα−1 times energy

120

200

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

0.5 0 100

120

180

Fig. 590: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 900 nodes.

19

1

180

(e) β = 5

Fig. 588: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 700 nodes.

19

160

1

(f) β = 6

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

160

5

(c) β = 3

1

(e) β = 5

0 100

1 120

140

Network scale

20

2

Latency

140

Network scale

120

15 x 10 EMA−SIC Cell−AS Li et al. 10

21

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latencyα−1 times energy

120

0 100

(b) β = 2

3

0 100

Latencyα−1 times energy

0 100

Latencyα−1 times energy

5

200

20

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

5 x 10 EMA−SIC 4 Cell−AS Li et al.

21

20

160

(a) β = 1

2

0 100

140

Network scale

0.5

Latencyα−1 times energy

1

Network scale

120

Latencyα−1 times energy

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latencyα−1 times energy

Latencyα−1 times energy

180

20

20

Latencyα−1 times energy

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

140

Network scale

0 100

1

Latencyα−1 times energy

(a) β = 1

0 100

times energy α−1

120

1

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

200

20

Latencyα−1 times energy

180

0 100

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

160

2

Latency

140

Network scale

4

Latencyα−1 times energy

120

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

5

0 100

19

19

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 591: Impact of network scale on the latency-energy tradeoff under α = 5 and uniform distribution with 1000 nodes.

185

140

160

Network scale

180

200

0 100

120

(a) β = 1

1

140

160

Network scale

180

200

2

120

160

(e) β = 5

180

200

140

160

Network scale

180

200

160

180

200

2 1

140

160

Network scale

2

0 100

120

1.5 1 0.5 120

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 592: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 100 nodes.

200

180

200

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

10

5

120

180

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

9

120

4

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

9

4

0 100

Latencyα−1 times energy

Latencyα−1 times energy

140

120

(d) β = 4

0.5

Network scale

0 100

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

9

1

120

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3 9

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

180

8

Latencyα−1 times energy

Latencyα−1 times energy

8

120

160

0.5

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

1

Latencyα−1 times energy

1

5

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

8

8

Latencyα−1 times energy

7

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

7

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 593: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 200 nodes.

186

120

(a) β = 1

140

160

Network scale

180

200

120

160

180

200

140

160

Network scale

180

(e) β = 5

140

160

180

2 140

160

Network scale

180

200

2 1

0 100

120

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

5

2

140

160

Network scale

(e) β = 5

120

180

200

200

0 100

120

140

160

Network scale

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 595: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 400 nodes.

160

180

200

180

200

(d) β = 4

140

160

Network scale

180

200

1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

9

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

2 1

120

140

160

Network scale

180

200

200

180

200

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(d) β = 4

1 0.5

120

180

11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

160

10

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

5

140

Network scale

11

11

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

120

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

10

1

0 100

180

2

(a) β = 1

(d) β = 4

4

160

1

0 100

10

Latencyα−1 times energy

Latencyα−1 times energy

200

2

(c) β = 3

120

180

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

10

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

10

9

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

140

Network scale

200

Fig. 596: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 500 nodes.

times energy

3

180

6 x 10 EMA−SIC Cell−AS Li et al. 4

9

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(a) β = 1

0 100

120

160

10

5

0 100

140

Network scale

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

α−1

Latencyα−1 times energy

Latencyα−1 times energy

4

120

(e) β = 5

9

6

120

120

(f) β = 6

8

0 100

0 100

200

Fig. 594: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 300 nodes.

10 x 10 EMA−SIC 8 Cell−AS Li et al.

2

(c) β = 3

2

Network scale

4

0 100

10

4

120

200

1

200

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

180

0.5

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

1

160

1.5

(d) β = 4

2

140

Network scale

10

10

140

120

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

0 100

(c) β = 3

Network scale

0 100

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(a) β = 1

5

10

4 x 10 EMA−SIC Cell−AS 3 Li et al.

120

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

180

9

4

120

160

(b) β = 2

9

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

Latencyα−1 times energy

160

Latencyα−1 times energy

140

Network scale

Latency

120

0.5

Latencyα−1 times energy

0 100

1

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

1

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

2

9

9

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latencyα−1 times energy

Latencyα−1 times energy

3

Latencyα−1 times energy

9

8

5 x 10 EMA−SIC 4 Cell−AS Li et al.

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 597: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 600 nodes.

187

120

(a) β = 1

1 140

160

Network scale

180

200

0 100

120

140

160

Network scale

180

200

160

180

200

2

120

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

120

1.5 1 0.5 0 100

120

(a) β = 1

140

160

Network scale

180

200

120

(c) β = 3

1

140

160

Network scale

(e) β = 5

180

200

200

140

160

180

200

0 100

140

160

Network scale

180

200

(f) β = 6

Fig. 599: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 800 nodes.

140

160

Network scale

200

140

160

Network scale

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

10

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

6 4 2 120

140

160

Network scale

180

200

180

200

180

200

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

160

(d) β = 4

4 2

120

140

Network scale

11

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

11

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

2

120

120

(d) β = 4

11

4

180

1 0.5

11

10

Network scale

200

1.5

(a) β = 1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

180

1

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

200

11

2

120

180

0.5

11

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

1

0 100

160

2

120

180

Fig. 600: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 900 nodes.

11

2

120

140

Network scale

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

10

140

Network scale

3

0 100

160

(b) β = 2

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(b) β = 2

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

120

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

9

10

2

120

(e) β = 5

Fig. 598: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 700 nodes.

9

0 100

11

2

(f) β = 6

6 x 10 EMA−SIC Cell−AS Li et al. 4

1

(c) β = 3

Latencyα−1 times energy

140

200

11

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

180

4

0 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

1.5

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

11

Network scale

120

10

5

(c) β = 3 11

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

0 100

Latencyα−1 times energy

2

120

200

10

3

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

10

120

160

(b) β = 2

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

2

Latencyα−1 times energy

160

4

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

0.5

Latencyα−1 times energy

0 100

1

8 x 10 EMA−SIC Cell−AS 6 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

Latencyα−1 times energy

2

10

9

10

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

9

4 x 10 EMA−SIC Cell−AS 3 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 601: Impact of network scale on the latency-energy tradeoff under α = 3 and poisson distribution with 1000 nodes.

188

12

200

0 100

120

(a) β = 1

2

140

160

Network scale

180

200

2 1 120

0.5

140

160

(e) β = 5

180

200

140

160

Network scale

180

200

0 100

2 1 140

160

Network scale

120

180

200

(f) β = 6

1 140

160

Network scale

180

200

Fig. 602: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 100 nodes.

200

180

200

180

200

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

14

1

120

180

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

160

13

2

120

140

Network scale

(b) β = 2

14

3

120

200

(c) β = 3

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

180

3

0 100

α−1

Latencyα−1 times energy

1

160

13

3

0 100

140

Network scale

5 x 10 EMA−SIC 4 Cell−AS Li et al.

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Network scale

120

(d) β = 4

13

Latencyα−1 times energy

0 100

5

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

4

0 100

180

12

Latencyα−1 times energy

Latencyα−1 times energy

12

120

160

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

180

1

Latencyα−1 times energy

160

2

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

0.5

3

times energy

120

1

12

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latency

0 100

Latencyα−1 times energy

Latencyα−1 times energy

5

12

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

10

15 x 10 EMA−SIC Cell−AS Li et al. 10

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 603: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 200 nodes.

189

(a) β = 1

140

160

Network scale

180

200

1

120

160

180

200

2 140

160

180

200

0.5

120

(e) β = 5

140

160

Network scale

180

200

140

160

Network scale

180

200

1 0.5 0 100

120

(a) β = 1

1

120

140

160

Network scale

180

200

120

(c) β = 3

1 0.5 140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1.5

120

0 100

120

180

200

160

Network scale

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 605: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 400 nodes.

180

200

180

200

1.5 1 0.5 0 100

120

140

160

Network scale

160

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 606: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 500 nodes.

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

180

200

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

120

140

160

Network scale

180

200

200

180

200

180

200

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

140

160

Network scale

(d) β = 4

6 4 2 120

180

16

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

160

(b) β = 2

5

0 100

140

Network scale

15

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

2

0 100

140

Network scale

15

6 x 10 EMA−SIC Cell−AS Li et al. 4

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1

180

180

15

14

140

160

(d) β = 4

1

15

15

200

2

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

160

5

0 100

180

3

14

2

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

14

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

140

Network scale

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

120

120

140

Network scale

(b) β = 2

2

14

2

120

(e) β = 5

Fig. 604: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 300 nodes.

13

0 100

15

4

(f) β = 6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

200

15

1

0 100

180

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

Latencyα−1 times energy

4

160

1

(c) β = 3

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

15

6

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

2

0 100

14

Network scale

120

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

0 100

14

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

14

14

120

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

2

Latencyα−1 times energy

180

4

Latencyα−1 times energy

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

2

Latencyα−1 times energy

0 100

4

Latencyα−1 times energy

0.5

14

13

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

Latencyα−1 times energy

13

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 607: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 600 nodes.

190

160

180

200

2

120

140

160

Network scale

180

200

140

160

Network scale

180

200

1

(e) β = 5

140

160

Network scale

180

200

0.5 0 100

120

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1

5

0 100

120

(a) β = 1

120

140

160

Network scale

180

200

0 100

120

200

1

140

160

Network scale

(e) β = 5

180

200

140

160

180

200

1 140

160

Network scale

180

200

(f) β = 6

Fig. 609: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 800 nodes.

160

Latencyα−1 times energy

180

200

200

180

200

2

0 100

120

140

160

Network scale

(f) β = 6

15

120

140

160

Network scale

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

2

120

140

160

Network scale

180

200

200

180

200

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(d) β = 4

3 2 1 120

180

16

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

16

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

(b) β = 2

(c) β = 3

2

120

180

16

3

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

1

0 100

16

Latencyα−1 times energy

Latencyα−1 times energy

16

140

Network scale

15

Network scale

120

16

2

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

180

5

(c) β = 3

0 100

160

15

1

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

15

0 100

(d) β = 4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(b) β = 2

3 x 10 EMA−SIC Cell−AS Li et al. 2

200

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

200

Fig. 610: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 900 nodes.

14

1.5

180

1

120

180

5

(e) β = 5

Fig. 608: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 700 nodes.

14

160

2

(f) β = 6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

160

(b) β = 2

1 120

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

16

2

120

120

(c) β = 3

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

0 100

15

2

0 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

1

200

3

16

16

180

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

160

(a) β = 1

4

0 100

140

Network scale

Latencyα−1 times energy

140

Network scale

120

0.5

15

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(c) β = 3 Latencyα−1 times energy

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

120

180

15

15

0 100

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

140

Network scale

0 100

1

Latencyα−1 times energy

(a) β = 1

0 100

times energy α−1

120

1

Latencyα−1 times energy

200

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

180

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

160

0 100

Latency

140

Network scale

2

Latencyα−1 times energy

120

4

Latencyα−1 times energy

0 100

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1 0.5

Latencyα−1 times energy

14

14

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 611: Impact of network scale on the latency-energy tradeoff under α = 4 and poisson distribution with 1000 nodes.

191

160

180

200

0 100

120

(a) β = 1

0.5

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1

120

2 1 140

160

(e) β = 5

180

200

Latencyα−1 times energy

3

140

160

Network scale

180

200

180

200

140

160

Network scale

1

0 100

120

4 2

120

140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 612: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 100 nodes.

200

180

200

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

19

4 2

120

180

(d) β = 4

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

18

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

(b) β = 2

18

2

120

160

2

(c) β = 3

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

17

17

Network scale

120

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1

2

0 100

17

Latencyα−1 times energy

0 100

(d) β = 4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

0 100

180

16

16

120

160

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

2

5

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

4

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

2

17

16

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

15

Latencyα−1 times energy

Latencyα−1 times energy

15

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 613: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 200 nodes.

192

(a) β = 1

140

160

Network scale

180

200

120

160

Network scale

180

200

140

160

180

200

times energy α−1

120

140

160

Network scale

180

200

160

180

200

2 1

0 100

120

(a) β = 1

0.5

120

140

160

Network scale

180

200

200

0 100

1 120

140

160

Network scale

180

200

2 140

160

Network scale

(e) β = 5

180

200

120

140

160

Network scale

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 615: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 400 nodes.

140

160

Network scale

200

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

19

140

160

Network scale

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

160

180

200

180

200

180

200

20

4 2

120

140

160

Network scale

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

160

(d) β = 4

2

120

140

Network scale

20

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

(b) β = 2

20

1

0 100

120

20

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

Latencyα−1 times energy

4

Latencyα−1 times energy

6

0 100

(c) β = 3

3 x 10 EMA−SIC Cell−AS Li et al. 2

180

2

(d) β = 4

19

2

0 100

200

4

(a) β = 1

20

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

120

(d) β = 4

19

Latencyα−1 times energy

180

3

(c) β = 3

0 100

160

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

180

2

19

19

160

6 x 10 EMA−SIC Cell−AS Li et al. 4

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

Latencyα−1 times energy

5

3

200

6

18

5 x 10 EMA−SIC 4 Cell−AS Li et al.

180

Fig. 616: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 500 nodes.

18

17

140

Network scale

1

120

160

19

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(e) β = 5

Fig. 614: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 300 nodes.

120

120

(f) β = 6

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

(b) β = 2

20

2

0 100

0 100

(c) β = 3

4

(e) β = 5

0 100

0 100

Latency

1

200

1

19

2

180

Latencyα−1 times energy

times energy times energy 140

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

19

160

19

α−1

0 100

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(d) β = 4

4 x 10 EMA−SIC Cell−AS 3 Li et al.

Network scale

120

(a) β = 1

5

(c) β = 3

120

200

18

2

0 100

180

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

18

120

160

(b) β = 2

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

Latencyα−1 times energy

120

0.5

Latencyα−1 times energy

200

1

Latencyα−1 times energy

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

160

19

1

0 100

Latency

140

Network scale

0 100

Latencyα−1 times energy

120

0.5

3 x 10 EMA−SIC Cell−AS Li et al. 2

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

1

α−1

2

Latency

3

0 100

18

18

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

17

5 x 10 EMA−SIC 4 Cell−AS Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 617: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 600 nodes.

193

120

(a) β = 1

160

180

200

1

120

160

180

140

160

180

200

0.5

(e) β = 5

140

160

Network scale

180

160

180

200

120

(a) β = 1

160

180

200

0 100

120

times energy 140

160

180

200

160

(e) β = 5

180

200

120

160

180

200

1 120

140

160

Network scale

180

200

(f) β = 6

Fig. 619: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 800 nodes.

Latencyα−1 times energy

200

0 100

120

140

160

Network scale

180

200

(f) β = 6

19

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

1

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

21

1

120

180

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

160

20

2

120

140

Network scale

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

α−1

Latencyα−1 times energy

140

Network scale

2

160

2

21

3

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

140

Network scale

20

Network scale

120

21

1

0 100

21

5

0 100

(a) β = 1

2

180

2

(d) β = 4

3 x 10 EMA−SIC Cell−AS Li et al. 2

(d) β = 4

20

Latencyα−1 times energy

200

Latencyα−1 times energy

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

180

6 x 10 EMA−SIC Cell−AS Li et al. 4

(c) β = 3

0 100

160

times energy

120

Latencyα−1 times energy

Latencyα−1 times energy

5

0 100

140

Network scale

200

Fig. 620: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 900 nodes.

20

19

200

1

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

0.5

α−1

5

0 100

160

1.5

120

180

4

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

140

Network scale

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(e) β = 5

19

5

120

120

(f) β = 6

18

0 100

0 100

140

Network scale

20

0.5

200

Fig. 618: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 700 nodes.

15 x 10 EMA−SIC Cell−AS Li et al. 10

120

(b) β = 2

21

1

120

0 100

(c) β = 3

1.5

0 100

200

1

Latencyα−1 times energy

2

180

1.5

(d) β = 4

4

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

21

6

140

Network scale

20

2

(c) β = 3

Network scale

120

(a) β = 1

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

20

120

0 100

200

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

180

(b) β = 2

5

120

160

20

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Latencyα−1 times energy

0 100

0.5

Latencyα−1 times energy

200

1

Latencyα−1 times energy

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

160

0.5

Latency

140

Network scale

2

1

Latency

120

4

20

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

4

Latencyα−1 times energy

Latencyα−1 times energy

6

0 100

19

19

18

10 x 10 EMA−SIC 8 Cell−AS Li et al.

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 621: Impact of network scale on the latency-energy tradeoff under α = 5 and poisson distribution with 1000 nodes.

194

180

200

0 100

120

(a) β = 1

200

0 100

120

140

160

Network scale

180

200

1

0 100

120

140

160

Network scale

180

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

200

1 0.5

140

160

Network scale

1

0 100

120

2 120

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 622: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 100 nodes.

200

180

200

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

9

3 2 1 120

180

(d) β = 4

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

9

4

0 100

140

Network scale

(b) β = 2

9

120

2

(c) β = 3

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

180

6

(d) β = 4

4

160

10 x 10 EMA−SIC 8 Cell−AS Li et al.

200

9

6

140

Network scale

8

3 x 10 EMA−SIC Cell−AS Li et al. 2

(c) β = 3 10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

120

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(a) β = 1 Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

8

Latencyα−1 times energy

180

8

8

0 100

160

2

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

4

Latencyα−1 times energy

160

6

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

2

Latencyα−1 times energy

120

4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

0 100

Latencyα−1 times energy

Latencyα−1 times energy

5

8

7

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

7

6

15 x 10 EMA−SIC Cell−AS Li et al. 10

8 x 10 EMA−SIC Cell−AS 6 Li et al. 4 2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 623: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 200 nodes.

195

120

(a) β = 1

0.5 140

160

Network scale

180

200

2 1 120

160

180

200

160

180

200

1 0.5 120

(e) β = 5

140

160

Network scale

180

200

140

160

Network scale

180

200

1 0.5 0 100

120

(a) β = 1

2 1 120

140

160

Network scale

180

200

120

(c) β = 3

1 0.5 140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1.5

120

180

200

140

160

200

1

120

140

160

Network scale

180

200

(f) β = 6

Fig. 625: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 400 nodes.

200

180

200

1.5 1 0.5 0 100

120

140

160

Network scale

160

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

9

6 4 2 0 100

120

140

160

Network scale

180

200

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

6 4 2 120

140

160

Network scale

180

200

200

180

200

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(d) β = 4

4 2

120

180

10

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

10

10 x 10 EMA−SIC 8 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

2

0 100

140

Network scale

10

4 x 10 EMA−SIC Cell−AS 3 Li et al.

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1

180

200

10

9

Network scale

180

(d) β = 4

10 x 10 EMA−SIC 8 Cell−AS Li et al.

10

10

200

1

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

160

5

0 100

180

2

120

160

Fig. 626: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 500 nodes.

9

3

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

9

160

4 x 10 EMA−SIC Cell−AS 3 Li et al.

(b) β = 2

5 x 10 EMA−SIC 4 Cell−AS Li et al.

140

Network scale

8

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

120

120

140

Network scale

(b) β = 2

2

0 100

9

2

120

(e) β = 5

Fig. 624: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 300 nodes.

8

0 100

10

4

(f) β = 6

4 x 10 EMA−SIC Cell−AS 3 Li et al.

200

10

1.5

0 100

180

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

Latencyα−1 times energy

140

160

1

(c) β = 3

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

10

5

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

3

0 100

9

Network scale

120

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

0 100

9

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

1

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0 100

180

9

9

120

160

(b) β = 2

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

Latencyα−1 times energy

160

2

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

2

6 x 10 EMA−SIC Cell−AS Li et al. 4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

0.5

4

Latencyα−1 times energy

1

9

8

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1.5

Latencyα−1 times energy

8

8

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 627: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 600 nodes.

196

180

200

120

(a) β = 1

160

180

200

3 2 1 120

140

160

Network scale

180

160

180

200

200

(e) β = 5

140

160

Network scale

180

200

0.5 0 100

120

140

160

Network scale

180

200

0 100

0.5

120

140

160

Network scale

180

200

160

180

200

120

(a) β = 1

0.5

140

160

Network scale

180

200

120

(c) β = 3

0.5

times energy 160

Network scale

180

200

140

160

(e) β = 5

180

200

1 0.5 140

160

Network scale

120

180

200

(f) β = 6

Fig. 629: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 800 nodes.

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

0 100

120

140

160

Network scale

180

200

180

200

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(d) β = 4 11

1.5 1 0.5 120

160

10

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

1.5

120

160

11

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

140

Network scale

1

(d) β = 4

1

Network scale

120

3 x 10 EMA−SIC Cell−AS Li et al. 2

11

Latencyα−1 times energy

Latencyα−1 times energy

140

200

1

10

2

180

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

0 100

200

α−1

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

10

10

120

160

200

9

1

(b) β = 2

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latencyα−1 times energy

times energy α−1

0 100

160

(f) β = 6

Latencyα−1 times energy

140

Network scale

4

Latency

120

6

2

140

Network scale

Fig. 630: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 900 nodes.

Latencyα−1 times energy

0 100

120

0 100

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

0.5

180

2

11

1

Latency

1

Latencyα−1 times energy

Latencyα−1 times energy

9

200

4

(e) β = 5

10 x 10 EMA−SIC 8 Cell−AS Li et al.

180

8 x 10 EMA−SIC Cell−AS 6 Li et al.

9

9

160

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

0 100

140

Network scale

10

1

(f) β = 6

Fig. 628: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 700 nodes.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

(b) β = 2

11

0.5

120

0 100

(c) β = 3

1

0 100

Latencyα−1 times energy

180

1.5

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

140

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(d) β = 4

5

140

Network scale

5

(a) β = 1

11

Network scale

120

10

(c) β = 3 Latencyα−1 times energy

0 100

200

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

10

120

180

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

1 0.5

(b) β = 2

5

120

140

Network scale

10

9

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

Latencyα−1 times energy

0 100

1.5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

160

2

9

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latencyα−1 times energy

120

Latencyα−1 times energy

Latencyα−1 times energy

5

0 100

9

9

8

15 x 10 EMA−SIC Cell−AS Li et al. 10

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 631: Impact of network scale on the latency-energy tradeoff under α = 3 and cluster distribution with 1000 nodes.

197

140

160

Network scale

180

200

1

0 100

120

(a) β = 1

120

140

160

Network scale

180

200

200

0 100

120

1

120

(c) β = 3

160

180

200

140

160

Network scale

180

4 2 0 100

120

140

160

Network scale

180

200

140

160

Network scale

(e) β = 5

180

200

0 100

120

140

160

Network scale

180

200

(f) β = 6

Fig. 632: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 100 nodes.

4 2

0 100

120

200

180

200

180

200

1.5 1 0.5 0 100

120

140

160

Network scale

13

8 x 10 EMA−SIC Cell−AS 6 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

120

5

180

(d) β = 4

13

1

160

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

140

Network scale

13

6

200

12

2

120

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(d) β = 4

12

4 x 10 EMA−SIC Cell−AS 3 Li et al.

2

0 100

12

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(a) β = 1 Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

5

Latencyα−1 times energy

180

12

11

0 100

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

1 0.5

Latencyα−1 times energy

120

2

Latencyα−1 times energy

0 100

3

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

12

12

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latencyα−1 times energy

11

11

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

Fig. 633: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 200 nodes.

198

120

(a) β = 1 13

160

180

200

0 100

120

(c) β = 3

180

200

140

160

Network scale

180

200

2 1

(e) β = 5

140

160

Network scale

180

200

2 140

160

Network scale

180

200

3 2 1

0 100

120

(a) β = 1

2 140

160

Network scale

180

200

2

140

160

Network scale

(e) β = 5

120

140

160

Network scale

180

200

180

200

120

140

160

Network scale

180

200

140

160

Network scale

120

180

200

(f) β = 6

Fig. 635: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 400 nodes.

140

160

Network scale

200

180

200

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

14

1.5 1 0.5 0 100

120

140

160

Network scale

180

200

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

2 1

120

140

160

Network scale

180

200

200

180

200

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

15

1.5 1 0.5 120

180

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

160

14

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

(b) β = 2

15

5

0 100

0 100

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

180

2

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(d) β = 4

4

200

14

1

120

180

4

(a) β = 1

14

Latencyα−1 times energy

Latencyα−1 times energy

200

2

(c) β = 3

120

180

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

14

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

160

200

Fig. 636: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 500 nodes.

Latencyα−1 times energy

4

Latencyα−1 times energy

Latencyα−1 times energy

6

120

0 100

14

13

0 100

140

Network scale

180

15

(b) β = 2

10 x 10 EMA−SIC 8 Cell−AS Li et al.

160

5

Latencyα−1 times energy

4

Latencyα−1 times energy

Latencyα−1 times energy

6

140

Network scale

13

5 x 10 EMA−SIC 4 Cell−AS Li et al.

160

14

15 x 10 EMA−SIC Cell−AS Li et al. 10

13

12

120

120

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(e) β = 5

Fig. 634: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 300 nodes.

0 100

0 100

(f) β = 6

10 x 10 EMA−SIC 8 Cell−AS Li et al.

120

(b) β = 2

14

3

120

0 100

(c) β = 3

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

200

0.5

Latencyα−1 times energy

160

180

1

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

140

160

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

14

1

140

Network scale

14

5

14

Network scale

120

(a) β = 1 Latencyα−1 times energy

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

120

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

13

2

120

160

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

200

2

Latencyα−1 times energy

180

0 100

4

Latencyα−1 times energy

160

0.5

6

Latencyα−1 times energy

140

Network scale

0 100

Latencyα−1 times energy

120

0.5

1

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

1

1

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

0 100

13

13

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

12

5 x 10 EMA−SIC 4 Cell−AS Li et al.

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 637: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 600 nodes.

199

180

200

0 100

120

(a) β = 1

160

180

200

0 100

120

140

160

Network scale

180

200

160

180

200

2

120

(e) β = 5

140

160

Network scale

180

200

160

180

200

180

200

120

140

160

Network scale

180

200

2

140

160

Network scale

180

200

2

0 100

140

160

Network scale

180

200

0 100

120

(c) β = 3

5

0 100

160

180

200

2 1 140

160

(e) β = 5

180

200

Latencyα−1 times energy

3

5

120

140

160

Network scale

180

200

(f) β = 6

Fig. 639: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 800 nodes.

140

160

Network scale

200

160

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

120

140

160

Network scale

180

200

200

180

200

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

(d) β = 4

5

120

180

16

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

(b) β = 2

5

0 100

140

Network scale

15

15 x 10 EMA−SIC Cell−AS Li et al. 10

(c) β = 3

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

15

15

Network scale

120

(d) β = 4

15

5 x 10 EMA−SIC 4 Cell−AS Li et al.

120

140

Network scale

120

(a) β = 1

1

180

5

14

14

Latencyα−1 times energy

120

Latencyα−1 times energy

4

200

(f) β = 6

15 x 10 EMA−SIC Cell−AS Li et al. 10

15

3 x 10 EMA−SIC Cell−AS Li et al. 2

180

Fig. 640: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 900 nodes.

(b) β = 2

8 x 10 EMA−SIC Cell−AS 6 Li et al.

160

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

160

140

Network scale

15

13

1

0 100

120

(d) β = 4

4

120

200

1

0 100

Latencyα−1 times energy

140

Network scale

14

Latencyα−1 times energy

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

180

2

(e) β = 5

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

Latencyα−1 times energy

120

160

(b) β = 2

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

120

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

0 100

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

14

13

6 x 10 EMA−SIC Cell−AS Li et al. 4

120

15

2

(f) β = 6

Fig. 638: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 700 nodes.

0 100

0 100

15

4

0 100

200

1

(c) β = 3

8 x 10 EMA−SIC Cell−AS 6 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

180

4

15

1

160

6

(d) β = 4

2

140

Network scale

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

0.5

15

Network scale

120

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

140

Network scale

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

0 100

14

1

(c) β = 3

120

200

15

2

0 100

180

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

14

120

160

(b) β = 2

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

140

Network scale

Latencyα−1 times energy

160

Latencyα−1 times energy

140

Network scale

5

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

1

Latencyα−1 times energy

1

14

13

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

2

3 x 10 EMA−SIC Cell−AS Li et al. 2

Latencyα−1 times energy

Latencyα−1 times energy

3

Latencyα−1 times energy

14

13

5 x 10 EMA−SIC 4 Cell−AS Li et al.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 641: Impact of network scale on the latency-energy tradeoff under α = 4 and cluster distribution with 1000 nodes.

200

180

200

120

(a) β = 1

160

180

200

1

120

4 2

160

(e) β = 5

180

200

160

180

200

160

180

200

4 2

140

160

Network scale

120

120

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 642: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 100 nodes.

200

180

200

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

140

160

Network scale

18

5

120

180

(d) β = 4

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

160

(b) β = 2

5

0 100

140

Network scale

17

17

120

0 100

(c) β = 3

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

16

140

120

5

(a) β = 1

2

(c) β = 3

Network scale

0 100

16

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

16

120

200

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

180

16

1

120

160

(b) β = 2

15

3 x 10 EMA−SIC Cell−AS Li et al. 2

0 100

140

Network scale

1 0.5

Latencyα−1 times energy

0 100

1.5

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

160

16

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

2

Latencyα−1 times energy

120

6 x 10 EMA−SIC Cell−AS Li et al. 4

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

16

15

15

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

2 x 10 EMA−SIC Cell−AS 1.5 Li et al. 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 643: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 200 nodes.

201

120

(a) β = 1

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

5

1

120

140

160

Network scale

180

200

160

180

200

2

(e) β = 5

140

160

Network scale

180

200

160

180

200

5

0 100

120

(a) β = 1

140

160

Network scale

180

200

120

(c) β = 3

140

160

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

0.5

200

140

160

180

200

1 140

160

Network scale

180

200

(f) β = 6

Fig. 645: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 400 nodes.

180

200

180

200

1 0.5 0 100

120

140

160

Network scale

(d) β = 4 19

140

160

Network scale

180

200

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

18

6 4 2 0 100

120

140

160

Network scale

180

200

6 x 10 EMA−SIC Cell−AS Li et al. 4

2

0 100

120

5

120

140

160

Network scale

180

200

180

200

180

200

180

200

5 x 10 EMA−SIC 4 Cell−AS Li et al. 3 2 1

0 100

120

140

160

Network scale

(d) β = 4 20

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

160

19

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

(b) β = 2

19

2

200

1.5

(c) β = 3

3

120

200

18

Network scale

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

(a) β = 1

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

180

10 x 10 EMA−SIC 8 Cell−AS Li et al.

(d) β = 4

1

Network scale

120

160

Fig. 646: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 500 nodes.

19

19

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

120

180

5

0 100

160

2

0 100

Latencyα−1 times energy

1

0 100

160

18

2

120

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

18

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

120

140

Network scale

(b) β = 2

2

0 100

Latencyα−1 times energy

140

Network scale

15 x 10 EMA−SIC Cell−AS Li et al. 10

Latencyα−1 times energy

Latencyα−1 times energy

120

120

19

17

17

1

0 100

(e) β = 5

Fig. 644: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 300 nodes.

17

200

4

(f) β = 6

3 x 10 EMA−SIC Cell−AS Li et al. 2

180

1

(c) β = 3

4

120

160

19

6

0 100

140

Network scale

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4 10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

120

3 x 10 EMA−SIC Cell−AS Li et al. 2

(a) β = 1

18

2

0 100

0 100

18

2

(c) β = 3

120

200

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

18

6 x 10 EMA−SIC Cell−AS Li et al. 4

0 100

180

18

17

120

160

(b) β = 2

15 x 10 EMA−SIC Cell−AS Li et al. 10

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

Latencyα−1 times energy

160

2

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

1

Latencyα−1 times energy

0 100

2

6 x 10 EMA−SIC Cell−AS Li et al. 4

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

3

Latencyα−1 times energy

4

18

17

5 x 10 EMA−SIC 4 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

6

Latencyα−1 times energy

17

16

10 x 10 EMA−SIC 8 Cell−AS Li et al.

3 x 10 EMA−SIC Cell−AS Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 647: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 600 nodes.

202

120

(a) β = 1

0.5 140

160

Network scale

180

200

4 2 0 100

120

140

160

Network scale

180

200

160

Network scale

180

200

2 1 120

140

160

Network scale

180

200

0.5 140

160

180

200

Latencyα−1 times energy

5

0 100

120

(a) β = 1

1

140

160

Network scale

180

200

120

1 0.5 140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1.5

140

160

180

200

2

140

160

Network scale

120

180

200

(f) β = 6

Fig. 649: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 800 nodes.

140

160

Network scale

180

200

180

200

(d) β = 4

140

160

Network scale

180

200

10 x 10 EMA−SIC 8 Cell−AS Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

19

2

120

140

160

Network scale

180

200

2.5 x 10 EMA−SIC 2 Cell−AS Li et al. 1.5 1 0.5 0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

4 x 10 EMA−SIC Cell−AS 3 Li et al. 2 1

0 100

120

140

160

Network scale

(d) β = 4

3 2 1 120

180

20

5 x 10 EMA−SIC 4 Cell−AS Li et al.

0 100

160

20

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

140

Network scale

(b) β = 2

(c) β = 3

4

120

0 100

20

20

8 x 10 EMA−SIC Cell−AS 6 Li et al.

0 100

200

4

0 100

20

20

180

1 0.5

19

Network scale

200

1.5

(a) β = 1

1

0 100

160

8 x 10 EMA−SIC Cell−AS 6 Li et al.

(d) β = 4

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

120

200

0.5

(c) β = 3

0 100

180

20

2

120

160

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

19

140

Network scale

1

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

Latencyα−1 times energy

1

180

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

18

15 x 10 EMA−SIC Cell−AS Li et al. 10

160

20

2

120

140

Network scale

Fig. 650: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 900 nodes.

18

1.5

120

(b) β = 2

4 x 10 EMA−SIC Cell−AS 3 Li et al.

0 100

Latencyα−1 times energy

18

0 100

(e) β = 5

Fig. 648: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 700 nodes.

Network scale

120

(f) β = 6

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

200

(c) β = 3

3

0 100

180

20

5 x 10 EMA−SIC 4 Cell−AS Li et al.

(e) β = 5

120

0 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

160

2

20

20

140

Network scale

6 x 10 EMA−SIC Cell−AS Li et al. 4

(d) β = 4

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

140

120

19

6

(c) β = 3

0 100

0 100

(a) β = 1

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

120

200

19

1.5

0 100

180

Latencyα−1 times energy

19

120

160

(b) β = 2

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

1 0.5

Latencyα−1 times energy

200

1

1.5

Latencyα−1 times energy

180

2

Latencyα−1 times energy

160

2

3

2.5 x 10 EMA−SIC 2 Cell−AS Li et al.

Latencyα−1 times energy

140

Network scale

4

Latencyα−1 times energy

120

6

19

5 x 10 EMA−SIC 4 Cell−AS Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

18

18

10 x 10 EMA−SIC 8 Cell−AS Li et al.

Latencyα−1 times energy

18

2 x 10 EMA−SIC Cell−AS 1.5 Li et al.

15 x 10 EMA−SIC Cell−AS Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 651: Impact of network scale on the latency-energy tradeoff under α = 5 and cluster distribution with 1000 nodes.

203

120

140

160

Network scale

180

200

1.5 1 100

120

(a) β = 1

2.5 100

200

120

140

160

Network scale

180

200

1

0.5 100

140

160

Network scale

180

200

0 100

120

(c) β = 3

160

180

200

1 160

(e) β = 5

180

200

Latencyα−1 times energy

2

140

2

120

(d) β = 4

3

Network scale

3

1 100

4 2

0 100

120

140

160

Network scale

180

200

140

160

Network scale

180

200

0 100

120

(f) β = 6

Fig. 652: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 100 nodes.

140

160

Network scale

200

180

200

8

2 1

120

180

5

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

0 100

200

10

8

8 x 10 EMA−SIC Li et al. 6

180

(b) β = 2

(c) β = 3

7

5 x 10 EMA−SIC 4 Li et al.

120

140

Network scale

160

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

0.5

Latencyα−1 times energy

120

1

140

Network scale

7

5 x 10 EMA−SIC Li et al. 4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

Latencyα−1 times energy

4

120

(a) β = 1

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

3

1.5

7

7

Latencyα−1 times energy

180

7

8 x 10 EMA−SIC Li et al. 6

0 100

160

3.5

(b) β = 2

6

0 100

140

Network scale

4

2 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

5

2

7

5 x 10 EMA−SIC 4.5 Li et al.

Latencyα−1 times energy

6

4 100

6

6

3 x 10 EMA−SIC Li et al. 2.5

Latencyα−1 times energy

Latencyα−1 times energy

5

7 x 10 EMA−SIC Li et al.

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 653: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 200 nodes.

204

(a) β = 1

140

160

Network scale

180

200

2 1 120

160

180

200

2 140

160

180

200

1

120

(e) β = 5

140

160

Network scale

180

200

160

180

200

10 8 6 4 100

120

(a) β = 1

1.5 120

140

160

Network scale

180

200

120

1 0.5 140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1.5

140

160

Network scale

180

200

1

120

180

200

0 100

140

160

Network scale

120

180

200

(f) β = 6

Fig. 655: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 400 nodes.

140

160

Network scale

180

200

8 x 10 EMA−SIC Li et al. 6 4 2

0 100

120

140

160

Network scale

(f) β = 6

8

0 100

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Li et al. 3

2

1 100

120

6 4

120

140

160

Network scale

180

200

200

180

200

180

200

3 x 10 EMA−SIC 2.5 Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

9

4 2

120

180

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

160

9

10 x 10 EMA−SIC Li et al. 8

2 100

140

Network scale

(b) β = 2

9

2

200

1 0.5

(c) β = 3

4 x 10 EMA−SIC Li et al. 3

0 100

160

8

2

180

2 x 10 EMA−SIC Li et al. 1.5

(a) β = 1

4

0 100

140

Network scale

5

9

9

Latencyα−1 times energy

200

200

9

1

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

120

180

6

(c) β = 3

0 100

160

180

(d) β = 4

10

Latencyα−1 times energy

2

200

15 x 10 EMA−SIC Li et al.

8

2.5

1 100

140

Network scale

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

8

180

2

(b) β = 2

3.5 x 10 EMA−SIC 3 Li et al.

160

4 x 10 EMA−SIC Li et al. 3

120

160

Fig. 656: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 500 nodes.

Latencyα−1 times energy

140

Network scale

140

Network scale

7

14 x 10 EMA−SIC 12 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

120

2 120

140

Network scale

9

3

0 100

7

3

120

(e) β = 5

Fig. 654: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 300 nodes.

7

0.5 100

(b) β = 2

4

(f) β = 6

4 x 10 EMA−SIC Li et al.

200

9

0.5 0 100

180

6 x 10 EMA−SIC 5 Li et al.

1 100

Latencyα−1 times energy

4

160

(c) β = 3

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

9

6

140

Network scale

8

3

0 100

8

Network scale

120

(d) β = 4

10 x 10 EMA−SIC 8 Li et al.

1 100

2 100

(a) β = 1

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

8

8

120

160

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

0 100

140

Network scale

Latencyα−1 times energy

120

1

Latencyα−1 times energy

2 100

1.5

Latencyα−1 times energy

200

4

2.5 x 10 EMA−SIC Li et al. 2

Latencyα−1 times energy

180

3

6

Latencyα−1 times energy

160

4

8

8 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

5

Latencyα−1 times energy

120

7 x 10 EMA−SIC 6 Li et al.

Latencyα−1 times energy

1

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0.5 100

7

7

7

2 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 657: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 600 nodes.

205

8

(a) β = 1

140

160

Network scale

180

200

1 100

120

140

160

Network scale

180

200

160

180

200

0.5

120

(e) β = 5

140

160

Network scale

180

200

2 1

140

160

180

200

6 4 2 100

120

(a) β = 1

0.5 100

120

140

160

Network scale

180

200

180

200

5

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

10

120

160

180

200

140

160

Network scale

180

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 659: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 800 nodes.

180

200

180

200

1

0 100

120

140

160

Network scale

(f) β = 6

9

3 2 120

140

160

Network scale

180

200

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100

120

4 x 10 EMA−SIC Li et al.

2

120

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

10

1.5 1 0.5 120

180

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

0 100

160

9

3

1 100

140

Network scale

(b) β = 2

10

1

160

2

(c) β = 3

2

140

Network scale

(d) β = 4

4

times energy 140

Network scale

120

4 x 10 EMA−SIC Li et al. 3

9

3 x 10 EMA−SIC Li et al.

0 100

2 100

10

6 x 10 EMA−SIC 5 Li et al.

1 100

α−1

0 100

200

4

(a) β = 1

2

200

Fig. 660: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 900 nodes.

10

15 x 10 EMA−SIC Li et al.

120

120

(d) β = 4

9

Latencyα−1 times energy

160

4

(c) β = 3

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

Latencyα−1 times energy

1

180

0.5

(b) β = 2

1.5

160

1

9

9

2.5 x 10 EMA−SIC Li et al. 2

140

Network scale

2 x 10 EMA−SIC Li et al. 1.5

0 100

180

6

8

8

10 x 10 EMA−SIC Li et al. 8

Latencyα−1 times energy

Latencyα−1 times energy

8

Network scale

120

160

10 x 10 EMA−SIC Li et al. 8

(e) β = 5

Fig. 658: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 700 nodes.

120

0.5 100

140

Network scale

9

1

(f) β = 6

4 x 10 EMA−SIC Li et al. 3

120

(c) β = 3

1

0 100

4 100

(b) β = 2

2

Latencyα−1 times energy

140

200

10

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

2

180

1.5

(d) β = 4

4

160

3 x 10 EMA−SIC 2.5 Li et al.

10

6

140

Network scale

9

2

(c) β = 3

Network scale

120

(a) β = 1

3

9

10 x 10 EMA−SIC 8 Li et al.

0 100

0 100

Latencyα−1 times energy

0.5

120

200

5 x 10 EMA−SIC Li et al. 4

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

9

9

120

160

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

0 100

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

2 100

Latencyα−1 times energy

200

6

Latencyα−1 times energy

180

8

Latencyα−1 times energy

160

1

12 x 10 EMA−SIC Li et al. 10

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

2

Latencyα−1 times energy

120

3

Latency

0.5 100

3

8

5 x 10 EMA−SIC 4 Li et al.

Latencyα−1 times energy

1

4

Latencyα−1 times energy

Latencyα−1 times energy

1.5

8

6 x 10 EMA−SIC Li et al. 5

Latencyα−1 times energy

8

2.5 x 10 EMA−SIC Li et al. 2

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 661: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and uniform distribution with 1000 nodes.

206

160

180

200

0 100

120

(a) β = 1

10

8

140

160

Network scale

180

200

6 4

140

160

Network scale

(e) β = 5

120

3

2 120

140

160

Network scale

180

200

180

200

200

0.5

140

160

Network scale

5

0 100

120

1.5 1

120

140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 662: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 100 nodes.

200

180

200

180

200

5.5 x 10 EMA−SIC 5 Li et al. 4.5 4 3.5 3 100

120

140

160

Network scale

12

10 8 6 120

180

(d) β = 4

14 x 10 EMA−SIC 12 Li et al.

4 100

160

11

2.5 x 10 EMA−SIC Li et al. 2

0.5 100

140

Network scale

(b) β = 2

11

1

120

180

10

(c) β = 3

2 x 10 EMA−SIC Li et al. 1.5

0 100

160

11

2.5

1.5 100

140

Network scale

15 x 10 EMA−SIC Li et al.

(a) β = 1

11

Latencyα−1 times energy

10

Latencyα−1 times energy

0 100

(d) β = 4

10 x 10 EMA−SIC Li et al. 8

120

200

4 x 10 EMA−SIC 3.5 Li et al.

(c) β = 3

2 100

180

10

Latencyα−1 times energy

Latencyα−1 times energy

9

120

160

(b) β = 2

12 x 10 EMA−SIC Li et al.

6 100

140

Network scale

2

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

2

4

Latencyα−1 times energy

1

4

10

8 x 10 EMA−SIC Li et al. 6

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

10

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

9

Latencyα−1 times energy

Latencyα−1 times energy

9

4 x 10 EMA−SIC Li et al. 3

3 x 10 EMA−SIC 2.5 Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

(f) β = 6

Fig. 663: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 200 nodes.

207

120

(a) β = 1

160

180

200

120

140

160

Network scale

180

200

3 140

160

Network scale

180

200

times energy α−1

5

0 100

120

140

160

Network scale

180

200

5

140

160

Network scale

180

200

1 0.5 120

(a) β = 1

2 1.5 140

160

Network scale

180

200

120

0 100

1

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1.5

140

160

Network scale

180

200

2

120

200

0.6 100

140

160

Network scale

120

180

200

(f) β = 6

Fig. 665: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 400 nodes.

140

160

Network scale

180

200

8 x 10 EMA−SIC Li et al. 6

4

2 100

120

140

160

Network scale

(f) β = 6

12

120

140

160

Network scale

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

0.5

120

140

160

Network scale

180

200

200

180

200

180

200

3.5 x 10 EMA−SIC 3 Li et al. 2.5 2 1.5 1 100

120

140

160

Network scale

14

6 5 4 120

180

(d) β = 4

8 x 10 EMA−SIC 7 Li et al.

3 100

160

13

1

0 100

140

Network scale

(b) β = 2

13

3

200

1 0.8

(c) β = 3

4 x 10 EMA−SIC Li et al.

1 100

180

2 x 10 EMA−SIC Li et al. 1.5

13

2 x 10 EMA−SIC Li et al.

160

13

4

180

1.2

(a) β = 1

5

3 100

140

Network scale

2

(d) β = 4

13

Latencyα−1 times energy

200

6

(c) β = 3

0.5 100

180

8 x 10 EMA−SIC 7 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2.5

120

160

200

13

4

12

12

1 100

140

Network scale

180

(d) β = 4

6 x 10 EMA−SIC Li et al.

(b) β = 2

3.5 x 10 EMA−SIC 3 Li et al.

200

12

1.5

0 100

180

2

120

160

Fig. 666: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 500 nodes.

Latencyα−1 times energy

10

160

3

12

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

11

140

Network scale

5 x 10 EMA−SIC Li et al. 4

1 100

140

Network scale

1.6 x 10 EMA−SIC 1.4 Li et al.

(e) β = 5

Fig. 664: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 300 nodes.

120

120

(f) β = 6

15 x 10 EMA−SIC Li et al.

120

(b) β = 2

13

10

(e) β = 5

0 100

2 100

Latency

4

1

13

4

12

5

200

2

(c) β = 3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

12

180

6

(d) β = 4

7 x 10 EMA−SIC 6 Li et al.

160

3

(a) β = 1

1.5 1 100

140

Network scale

8 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

140

Network scale

120

12

2

(c) β = 3

120

200

3 x 10 EMA−SIC Li et al. 2.5

Latencyα−1 times energy

Latencyα−1 times energy

4

2 100

180

12

6

120

160

(b) β = 2

11

10 x 10 EMA−SIC Li et al. 8

2 100

140

Network scale

Latencyα−1 times energy

200

Latencyα−1 times energy

180

5 x 10 EMA−SIC 4 Li et al.

0 100

Latencyα−1 times energy

160

0 100

Latencyα−1 times energy

140

Network scale

0.5

Latencyα−1 times energy

120

0 100

1

Latencyα−1 times energy

0 100

2

1.5

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

4

Latencyα−1 times energy

2

12

12

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

11

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

Latencyα−1 times energy

11

4 x 10 EMA−SIC Li et al. 3

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 667: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 600 nodes.

208

160

180

200

120

(a) β = 1

1 140

160

Network scale

180

200

2 100

120

140

160

Network scale

180

200

Network scale

180

200

140

160

Network scale

180

200

180

200

200

180

200

180

200

10 8 6 4 100

120

140

160

Network scale

(d) β = 4 14

2

140

160

Network scale

180

200

7 x 10 EMA−SIC 6 Li et al. 5 4 3

2 100

120

(e) β = 5

140

160

Network scale

(f) β = 6

Fig. 670: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 900 nodes.

120

140

160

Network scale

180

200

2 1

0 100

2

140

160

180

200

10 x 10 EMA−SIC Li et al.

6

120

(c) β = 3

140

160

Network scale

180

200

140

160

Network scale

(e) β = 5

180

200

0 100

2

140

160

Network scale

120

4 2 100

120

140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 669: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 800 nodes.

180

200

180

200

180

200

2.5 x 10 EMA−SIC Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

14

4 3

120

160

(d) β = 4

6 x 10 EMA−SIC Li et al. 5

2 100

140

Network scale

14

14

3

120

2

(c) β = 3

5 x 10 EMA−SIC Li et al. 4

1 100

4

(b) β = 2

6

Latencyα−1 times energy

1

Latencyα−1 times energy

1.5

200

8

14

14

180

12 x 10 EMA−SIC 10 Li et al.

(d) β = 4

2.5 x 10 EMA−SIC Li et al. 2

160

13

8

4 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

13

Network scale

120

(b) β = 2

3

13

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

200

0 100

Latencyα−1 times energy

Latencyα−1 times energy

180

Latencyα−1 times energy

180

1

13

Latencyα−1 times energy

160

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

160

5 x 10 EMA−SIC Li et al. 4

120

160

14 x 10 EMA−SIC 12 Li et al.

13

2

(a) β = 1

0.5 100

140

Network scale

3

120

140

Network scale

13

4 x 10 EMA−SIC Li et al.

1 100

Latencyα−1 times energy

140

Network scale

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

Latencyα−1 times energy

1 0.5

120

120

13

13

2 x 10 EMA−SIC Li et al. 1.5

1 100

120

(f) β = 6

Fig. 668: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 700 nodes.

120

2 100

α−1

1

(e) β = 5

0 100

(b) β = 2

4

times energy

2

120

2

14

1.5

0.5 100

200

6

Latency

160

180

4

(c) β = 3

3 x 10 EMA−SIC 2.5 Li et al.

Latencyα−1 times energy

0.8

160

8 x 10 EMA−SIC Li et al.

14

1

140

Network scale

13

3

14

140

120

(d) β = 4

1.4 x 10 EMA−SIC Li et al. 1.2

0 100

0 100

6 x 10 EMA−SIC Li et al.

(a) β = 1

4

(c) β = 3 Latencyα−1 times energy

200

Latencyα−1 times energy

1.5

120

180

6 x 10 EMA−SIC Li et al. 5

Latencyα−1 times energy

Latencyα−1 times energy

2

0.6 100

160

13

13

120

140

Network scale

1

(b) β = 2

3 x 10 EMA−SIC 2.5 Li et al.

0.5 100

times energy α−1

0 100

2

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

0.5

13

3 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

2

1

Latency

4

Latencyα−1 times energy

Latencyα−1 times energy

6

0 100

13

13

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

12

10 x 10 EMA−SIC 8 Li et al.

12 x 10 EMA−SIC 10 Li et al. 8 6 4 2 100

120

140

160

Network scale

(f) β = 6

Fig. 671: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and uniform distribution with 1000 nodes.

209

120

140

160

Network scale

180

200

4

2

0 100

120

(a) β = 1

2 0 100

200

120

140

160

Network scale

180

200

1 0.5 0 100

140

160

Network scale

180

200

120

(c) β = 3

140

160

Network scale

180

200

(d) β = 4

1

140

160

(e) β = 5

180

200

Latencyα−1 times energy

1.5

Network scale

3 2

120

160

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 672: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 100 nodes.

200

180

200

1 120

140

160

Network scale

15

4 2

120

180

2

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

200

3

0 100

15

5 x 10 EMA−SIC Li et al. 4

1 100

140

Network scale

Latencyα−1 times energy

1

120

180

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

14

2.5 x 10 EMA−SIC Li et al. 2

120

times energy α−1

0 100

160

(b) β = 2

2

0 100

Latencyα−1 times energy

120

5

140

Network scale

15

3 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

10

Latency

4

120

(a) β = 1

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

4

2 x 10 EMA−SIC Li et al. 1.5

15

14

Latencyα−1 times energy

180

13

8 x 10 EMA−SIC Li et al. 6

0.5 100

160

6

(b) β = 2

13

0 100

140

Network scale

15

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

1

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

2

0 100

14

13

Latencyα−1 times energy

Latencyα−1 times energy

13

3 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 673: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 200 nodes.

210

120

(a) β = 1

140

160

Network scale

180

200

1

120

16

160

180

200

140

160

180

200

6 4 2 100

120

140

160

Network scale

180

160

180

200

2

120

2 120

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

4

0 100

0 100

120

140

160

Network scale

180

200

1.5 1 0.5 140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

2 1.5 120

140

160

Network scale

180

200

(f) β = 6

Fig. 675: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 400 nodes.

140

160

Network scale

200

12 x 10 EMA−SIC 10 Li et al. 8 6 4 2 100

120

140

160

Network scale

(f) β = 6

17

1 0.5 0 100

120

140

160

Network scale

180

200

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

18

1 0.5

120

180

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

0 100

160

17

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

(b) β = 2

18

2.5

1 100

120

(c) β = 3

3.5 x 10 EMA−SIC 3 Li et al.

180

1

0 100

17

5

200

2

(a) β = 1

10

180

3

17

1.5

200

17

Latencyα−1 times energy

17

Latencyα−1 times energy

180

200

(d) β = 4

2

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

120

160

15 x 10 EMA−SIC Li et al.

(c) β = 3

0 100

140

Network scale

16

6

200

4

(b) β = 2

16

180

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

4

(a) β = 1 10 x 10 EMA−SIC 8 Li et al.

160

17

6 x 10 EMA−SIC Li et al.

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

120

180

Fig. 676: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 500 nodes.

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

Latencyα−1 times energy

120

120

160

5 x 10 EMA−SIC 4 Li et al.

(e) β = 5

16

1

0 100

0 100

140

Network scale

(b) β = 2

0.5

0 100

200

Fig. 674: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 300 nodes.

2

120

17

1

(f) β = 6

16

0 100

17

(e) β = 5

3 x 10 EMA−SIC Li et al.

200

1.5

Latencyα−1 times energy

2

180

(c) β = 3

10 x 10 EMA−SIC Li et al. 8

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

16

3

160

(a) β = 1

2

0 100

140

Network scale

2.5 x 10 EMA−SIC 2 Li et al.

(d) β = 4

5 x 10 EMA−SIC Li et al. 4

Network scale

120

17

4 x 10 EMA−SIC Li et al. 3

(c) β = 3

120

0 100

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

1 100

180

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

120

160

16

16

0 100

140

Network scale

Latencyα−1 times energy

0 100

0.5

Latencyα−1 times energy

200

2

1

Latencyα−1 times energy

180

4

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

160

6

Latencyα−1 times energy

140

Network scale

5

Latencyα−1 times energy

120

10

17

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

4

0 100

16

15

15 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

15

8 x 10 EMA−SIC Li et al. 6

2.5 x 10 EMA−SIC Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

(f) β = 6

Fig. 677: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 600 nodes.

211

200

0 100

120

(a) β = 1

140

160

Network scale

180

200

1 0.5 0 100

120

1

140

160

Network scale

180

200

140

160

180

200

2

140

160

Network scale

(e) β = 5

180

200

140

160

180

200

120

2

1

120

140

160

Network scale

180

200

200

1 140

160

Network scale

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

180

200

4

120

160

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 679: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 800 nodes.

160

200

180

200

1 0.5 0 100

120

140

160

Network scale

(f) β = 6

18

120

140

160

Network scale

180

200

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

19

1 0.5

120

180

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

0 100

160

18

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

(b) β = 2

19

6

140

Network scale

2 x 10 EMA−SIC Li et al. 1.5

(c) β = 3

10 x 10 EMA−SIC Li et al. 8

2 100

140

Network scale

1

0 100

18

4

120

19

18

2

120

0 100

(a) β = 1

3

0 100

18

Latencyα−1 times energy

180

180

2

(d) β = 4

2

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

120

160

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

0 100

140

Network scale

18

Latencyα−1 times energy

Latencyα−1 times energy

18

200

3 x 10 EMA−SIC Li et al.

(b) β = 2

3 x 10 EMA−SIC Li et al.

0 100

120

200

Fig. 680: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 900 nodes.

times energy

0.5

(a) β = 1

180

4

α−1

1

0 100

160

6

Latency

2

140

Network scale

8

2 100

180

4

18

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

4

Network scale

120

160

18

12 x 10 EMA−SIC 10 Li et al.

18

6

120

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

(e) β = 5

Fig. 678: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 700 nodes.

17

120

(b) β = 2

1

(f) β = 6

10 x 10 EMA−SIC 8 Li et al.

1

0 100

18

3

120

200

2

(c) β = 3

5 x 10 EMA−SIC Li et al. 4

1 100

180

2

Latencyα−1 times energy

2

160

3

18

Latencyα−1 times energy

Latencyα−1 times energy

18

140

Network scale

5 x 10 EMA−SIC 4 Li et al.

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

Network scale

120

18

1.5

(c) β = 3

0 100

0 100

4 x 10 EMA−SIC Li et al. 3

(a) β = 1

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

5

120

200

18

10

0 100

180

Latencyα−1 times energy

17

120

160

(b) β = 2

15 x 10 EMA−SIC Li et al.

0 100

140

Network scale

Latencyα−1 times energy

180

Latencyα−1 times energy

160

0.5

Latencyα−1 times energy

140

Network scale

1

Latencyα−1 times energy

120

2

Latencyα−1 times energy

0 100

4

Latencyα−1 times energy

1

6

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

18

18

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

17

10 x 10 EMA−SIC 8 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

17

5 x 10 EMA−SIC 4 Li et al.

3 x 10 EMA−SIC 2.5 Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

(f) β = 6

Fig. 681: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and uniform distribution with 1000 nodes.

212

160

180

200

120

(a) β = 1

140

160

Network scale

180

200

0.5 0 100

120

2 1 160

(e) β = 5

180

200

160

180

200

180

200

0.5 100

3 2 120

4 2

140

160

Network scale

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 682: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 100 nodes.

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

8

2 1

120

180

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

0 100

160

7

4

1 100

140

Network scale

(b) β = 2

8

120

120

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

0 100

160

6 x 10 EMA−SIC 5 Li et al.

(d) β = 4 Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

7

3

140

Network scale

1

7

(c) β = 3

140

120

1.5

(a) β = 1

1

7

5 x 10 EMA−SIC 4 Li et al.

Network scale

1 100

Latencyα−1 times energy

2

120

200

7

4

0 100

180

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

6

120

160

(b) β = 2

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

2

Latencyα−1 times energy

0 100

3

2.5 x 10 EMA−SIC Li et al. 2

Latencyα−1 times energy

140

Network scale

1

Latencyα−1 times energy

120

2

7

5 x 10 EMA−SIC Li et al. 4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

5

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

6

4 100

6

6

Latencyα−1 times energy

Latencyα−1 times energy

5

8 x 10 EMA−SIC Li et al. 7

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

Fig. 683: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 200 nodes.

213

(a) β = 1

140

160

Network scale

180

200

2 1 120

160

180

200

140

160

180

200

0.5

120

(e) β = 5

140

160

Network scale

180

200

160

180

200

1

0.5 100

120

(a) β = 1

1.5 120

140

160

Network scale

180

200

0 100

120

1 0.5 140

160

(e) β = 5

180

200

140

160

Network scale

180

200

1

120

160

180

200

0 100

140

160

Network scale

120

180

200

(f) β = 6

Fig. 685: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 400 nodes.

140

160

Network scale

180

200

8 x 10 EMA−SIC Li et al. 6 4 2

0 100

120

140

160

Network scale

(f) β = 6

8

6 4 100

120

140

160

Network scale

180

200

5 x 10 EMA−SIC Li et al. 4 3 2

1 100

120

8 6 4 120

140

160

Network scale

180

200

200

180

200

180

200

3 x 10 EMA−SIC 2.5 Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

9

4 2

120

180

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

160

9

12 x 10 EMA−SIC 10 Li et al.

2 100

140

Network scale

(b) β = 2

9

2

200

1 0.5

(c) β = 3

4 x 10 EMA−SIC Li et al. 3

0 100

140

Network scale

8

2

180

2 x 10 EMA−SIC Li et al. 1.5

(a) β = 1

4

Latencyα−1 times energy

1.5

Network scale

1

8

9

9

Latencyα−1 times energy

200

200

9

2

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

120

180

6

(c) β = 3

0 100

160

180

(d) β = 4

10

Latencyα−1 times energy

2

200

14 x 10 EMA−SIC 12 Li et al.

8

2.5

1 100

140

Network scale

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

8

180

3

(b) β = 2

3.5 x 10 EMA−SIC 3 Li et al.

160

5 x 10 EMA−SIC 4 Li et al.

120

160

Fig. 686: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 500 nodes.

Latencyα−1 times energy

140

Network scale

140

Network scale

7

1.5 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

120

3 120

140

Network scale

9

4

0 100

8

3

120

(e) β = 5

Fig. 684: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 300 nodes.

7

1 100

(b) β = 2

5

(f) β = 6

4 x 10 EMA−SIC Li et al.

200

9

1

0 100

180

7 x 10 EMA−SIC 6 Li et al.

2 100

Latencyα−1 times energy

5

160

(c) β = 3

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

9

10

140

Network scale

8

3

0 100

8

Network scale

120

(d) β = 4

15 x 10 EMA−SIC Li et al.

1 100

2 100

(a) β = 1

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

180

8

8

120

160

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

0 100

140

Network scale

Latencyα−1 times energy

120

1.5

Latencyα−1 times energy

2 100

2

Latencyα−1 times energy

200

4

3 x 10 EMA−SIC Li et al. 2.5

Latencyα−1 times energy

180

3

6

Latencyα−1 times energy

160

4

8

8 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

5

Latencyα−1 times energy

120

7 x 10 EMA−SIC 6 Li et al.

Latencyα−1 times energy

1

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0.5 100

7

7

7

2 x 10 EMA−SIC Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 687: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 600 nodes.

214

8

(a) β = 1

140

160

Network scale

180

200

1 100

120

140

160

Network scale

180

200

180

200

0.5

120

(e) β = 5

140

160

Network scale

180

200

2 1

140

160

Network scale

180

200

6 4 2 100

120

(a) β = 1

120

140

160

Network scale

180

200

180

200

0 100

120

160

Network scale

180

200

120

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

120

140

160

Network scale

180

200

(f) β = 6

Fig. 689: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 800 nodes.

180

200

180

200

1

0 100

120

140

160

Network scale

(f) β = 6

9

3 2 120

140

160

Network scale

180

200

2 x 10 EMA−SIC Li et al. 1.5

1

0.5 100

120

2

120

140

160

Network scale

180

200

200

180

200

180

200

12 x 10 EMA−SIC 10 Li et al. 8 6 4 2 100

120

140

160

Network scale

(d) β = 4

2

1

120

180

10

3 x 10 EMA−SIC Li et al.

0 100

160

9

3

1 100

140

Network scale

(b) β = 2

5 x 10 EMA−SIC Li et al. 4

times energy

1

160

2

(c) β = 3

2

140

Network scale

4 x 10 EMA−SIC Li et al. 3

10

3 x 10 EMA−SIC Li et al.

0 100

120

(d) β = 4

4

1 100

α−1

5

2 100

10

6 x 10 EMA−SIC 5 Li et al.

(d) β = 4

10

200

4

9

140

200

Fig. 690: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 900 nodes.

10

15 x 10 EMA−SIC Li et al.

180

0.5 120

180

6

(a) β = 1

2

Latencyα−1 times energy

Latencyα−1 times energy

160

4

(c) β = 3 9

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

Latencyα−1 times energy

1

160

1

(b) β = 2

1.5

140

Network scale

1.5

0 100

160

9

2.5 x 10 EMA−SIC 2 Li et al.

9

9

2.5 x 10 EMA−SIC Li et al. 2

0.5 100

120

140

Network scale

10 x 10 EMA−SIC Li et al. 8

8

8

10 x 10 EMA−SIC Li et al. 8

Latencyα−1 times energy

Latencyα−1 times energy

8

120

(e) β = 5

Fig. 688: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 700 nodes.

120

1 100

(f) β = 6

4 x 10 EMA−SIC Li et al. 3

4 100

(c) β = 3

1

0 100

6

(b) β = 2

1.5

Latencyα−1 times energy

160

200

10

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

180

2

(d) β = 4

5

160

3 x 10 EMA−SIC Li et al. 2.5

10

10

140

9

2

(c) β = 3

Network scale

120

(a) β = 1

3

9

15 x 10 EMA−SIC Li et al.

0 100

1 100

Latencyα−1 times energy

1

120

200

5 x 10 EMA−SIC Li et al. 4

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0 100

Network scale

180

9

9

120

160

(b) β = 2

2 x 10 EMA−SIC Li et al.

0.5 100

140

Latencyα−1 times energy

120

Latencyα−1 times energy

2 100

Latencyα−1 times energy

200

8

Latencyα−1 times energy

Network scale

180

10

Latencyα−1 times energy

160

14 x 10 EMA−SIC 12 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

2

Latencyα−1 times energy

120

3

3

Latencyα−1 times energy

0.5 100

4

Latencyα−1 times energy

1

5

8

5 x 10 EMA−SIC Li et al. 4

Latency

Latencyα−1 times energy

1.5

8

7 x 10 EMA−SIC 6 Li et al.

Latencyα−1 times energy

8

2.5 x 10 EMA−SIC Li et al. 2

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 691: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and poisson distribution with 1000 nodes.

215

160

180

200

120

(a) β = 1

1

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1.2

2

6 4 140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

140

160

Network scale

180

200

1.5 1 0.5 120

200

140

160

Network scale

6 4 100

120

1.8 1.6 1.4 120

140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 692: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 100 nodes.

200

180

200

180

200

8 x 10 EMA−SIC 7 Li et al. 6 5 4

3 100

120

140

160

Network scale

12

1 0.5

120

180

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

0 100

160

11

2.2 x 10 EMA−SIC 2 Li et al.

1.2 100

140

Network scale

(b) β = 2

12

2.5 x 10 EMA−SIC 2 Li et al.

0 100

180

8

(c) β = 3

11

8

160

11

3

120

140

Network scale

10

(a) β = 1

4

1 100

10

(e) β = 5

120

(d) β = 4

12 x 10 EMA−SIC 10 Li et al.

120

200

6 x 10 EMA−SIC 5 Li et al.

(c) β = 3

2 100

180

0 100

10

10

120

160

2

(b) β = 2

1.6 x 10 EMA−SIC Li et al. 1.4

0.8 100

140

Network scale

4

Latencyα−1 times energy

140

Network scale

3 100

Latencyα−1 times energy

120

4

Latencyα−1 times energy

0 100

5

14 x 10 EMA−SIC 12 Li et al.

Latencyα−1 times energy

1

6

10

8 x 10 EMA−SIC Li et al. 6

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

10

8 x 10 EMA−SIC 7 Li et al.

Latencyα−1 times energy

9

Latencyα−1 times energy

Latencyα−1 times energy

9

4 x 10 EMA−SIC Li et al. 3

3 x 10 EMA−SIC 2.5 Li et al. 2 1.5 1 0.5 100

120

140

160

Network scale

(f) β = 6

Fig. 693: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 200 nodes.

216

2 100

120

(a) β = 1

6 140

160

Network scale

180

200

2 1.5 100

120

140

160

Network scale

180

200

140

160

180

200

10

5

0 100

120

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

1.5 1 0.5 100

120

(a) β = 1

2 1.5 1 100

120

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

2.5

1.5 1

140

160

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

Network scale

180

200

120

140

160

Network scale

180

200

2

120

140

160

Network scale

180

200

(f) β = 6

Fig. 695: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 400 nodes.

200

180

200

180

200

1 0.5 100

120

140

160

Network scale

(d) β = 4 13

140

160

Network scale

180

200

10 x 10 EMA−SIC Li et al. 8 6 4 2 100

120

140

160

Network scale

(f) β = 6

12

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

180

200

12 x 10 EMA−SIC 10 Li et al. 8 6 4 2 100

120

1.2 1

120

140

160

Network scale

180

200

200

180

200

180

200

4 x 10 EMA−SIC 3.5 Li et al. 3 2.5 2 1.5 100

120

140

160

Network scale

14

6 4

120

180

(d) β = 4

10 x 10 EMA−SIC Li et al. 8

2 100

160

13

1.6 x 10 EMA−SIC Li et al. 1.4

0.8 100

140

Network scale

(b) β = 2

13

3

180

1.5

(c) β = 3

5 x 10 EMA−SIC Li et al. 4

1 100

200

13

5

160

2.5 x 10 EMA−SIC Li et al. 2

(a) β = 1

6

140

Network scale

Fig. 696: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 500 nodes.

13

13

Network scale

120

(d) β = 4

2.5 x 10 EMA−SIC Li et al. 2

120

160

7

(c) β = 3

0.5 100

140

9 x 10 EMA−SIC 8 Li et al.

4 100

180

2

12

12

160

3

(b) β = 2

3.5 x 10 EMA−SIC 3 Li et al.

140

Network scale

5 x 10 EMA−SIC Li et al. 4

1 100

Latencyα−1 times energy

120

120

12

2.5 x 10 EMA−SIC Li et al. 2

Latencyα−1 times energy

Latencyα−1 times energy

5

120

13

4

12

10

1 100

(e) β = 5

Fig. 694: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 300 nodes.

11

2

(b) β = 2

5

(f) β = 6

15 x 10 EMA−SIC Li et al.

200

13

Latencyα−1 times energy

4

180

6

3 100

12

6

160

3

(c) β = 3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

12

140

Network scale

8 x 10 EMA−SIC 7 Li et al.

(d) β = 4

8 x 10 EMA−SIC Li et al.

Network scale

120

12

2.5

(c) β = 3

0 100

0 100

5 x 10 EMA−SIC Li et al. 4

(a) β = 1

3.5 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

Latencyα−1 times energy

7

120

200

12

8

2 100

180

Latencyα−1 times energy

11

120

160

(b) β = 2

10 x 10 EMA−SIC 9 Li et al.

5 100

140

Network scale

Latencyα−1 times energy

200

0.5

Latencyα−1 times energy

180

1

Latencyα−1 times energy

160

1.5

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

3

Latencyα−1 times energy

0 100

4

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

5

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

2

12

12

7 x 10 EMA−SIC 6 Li et al.

Latencyα−1 times energy

11

Latencyα−1 times energy

Latencyα−1 times energy

11

4 x 10 EMA−SIC Li et al. 3

2 x 10 EMA−SIC Li et al. 1.5

1

0.5 100

120

140

160

Network scale

(f) β = 6

Fig. 697: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 600 nodes.

217

13

120

(a) β = 1

1.5

140

160

Network scale

180

200

times energy

4 120

140

160

Network scale

180

200

140

160

Network scale

180

200

times energy 140

160

Network scale

180

200

140

160

Network scale

180

200

4 x 10 EMA−SIC Li et al. 3

1

120

(a) β = 1

140

160

Network scale

180

200

6 100

120

2 1.5

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

140

160

Network scale

180

200

3 2 120

140

160

Network scale

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 699: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 800 nodes.

160

180

200

180

200

6 4 2 100

120

140

160

Network scale

(f) β = 6

13

1

120

140

160

Network scale

180

200

8 x 10 EMA−SIC Li et al. 6 4 2

0 100

120

8 6

120

140

160

Network scale

180

200

200

180

200

180

200

3 x 10 EMA−SIC Li et al. 2.5 2 1.5 1 100

120

140

160

Network scale

14

5 4 3 120

180

(d) β = 4

7 x 10 EMA−SIC 6 Li et al.

2 100

160

14

12 x 10 EMA−SIC Li et al. 10

4 100

140

Network scale

(b) β = 2

14

4

140

Network scale

10 x 10 EMA−SIC Li et al. 8

(c) β = 3

6 x 10 EMA−SIC 5 Li et al.

1 100

120

14

2

0 100

14

14

0.5 100

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

(d) β = 4

3 x 10 EMA−SIC Li et al. 2.5

200

13

8

200

1

(a) β = 1

10

(c) β = 3

120

200

Latencyα−1 times energy

2

1 100

180

13

3

120

160

12 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

13

1 100

140

Network scale

180

Fig. 700: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 900 nodes.

(b) β = 2

5 x 10 EMA−SIC Li et al. 4

180

2

120

160

1.5

13

2

0 100

160

3

1 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

140

Network scale

5 x 10 EMA−SIC Li et al. 4

13

120

120

140

Network scale

2 x 10 EMA−SIC Li et al.

(e) β = 5

Fig. 698: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 700 nodes.

13

120

14

4

(f) β = 6

2 x 10 EMA−SIC Li et al. 1.5

1 100

(b) β = 2

5

3 100

α−1

1

120

2

14

2

(e) β = 5

200

3

(c) β = 3

4 x 10 EMA−SIC Li et al. 3

0 100

180

6

Latency

1

Latencyα−1 times energy

1.5

160

4

13

5

3 100

140

Network scale

8 x 10 EMA−SIC 7 Li et al.

14

14

120

120

(d) β = 4

2 x 10 EMA−SIC Li et al.

0 100

0 100

6 x 10 EMA−SIC 5 Li et al.

(a) β = 1

6

(c) β = 3 Latencyα−1 times energy

200

8 x 10 EMA−SIC 7 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

0.5 100

Network scale

180

13

13

120

160

(b) β = 2

3 x 10 EMA−SIC Li et al. 2.5

1 100

140

Latencyα−1 times energy

0.5 100

Latencyα−1 times energy

200

Latencyα−1 times energy

Network scale

180

1

Latencyα−1 times energy

160

2

Latencyα−1 times energy

140

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

1

13

3 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

1.5

α−1

4

Latencyα−1 times energy

Latencyα−1 times energy

6

13

2 x 10 EMA−SIC Li et al.

Latency

12

10 x 10 EMA−SIC 8 Li et al.

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 701: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and poisson distribution with 1000 nodes.

218

120

140

160

Network scale

180

200

4

2

0 100

120

(a) β = 1

2 0 100

200

120

140

160

Network scale

180

200

1 0.5 0 100

140

160

Network scale

180

200

120

(c) β = 3

140

160

Network scale

180

140

160

(e) β = 5

180

200

Latencyα−1 times energy

1.5

4 3

2 100

160

180

200

Latencyα−1 times energy

140

Network scale

120

140

160

Network scale

180

200

120

(f) β = 6

Fig. 702: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 100 nodes.

140

160

Network scale

200

180

200

16

6 4

120

180

2

(d) β = 4

10 x 10 EMA−SIC Li et al. 8

2 100

200

3

1 100

15

6 x 10 EMA−SIC Li et al. 5

180

5 x 10 EMA−SIC Li et al. 4

(c) β = 3

(d) β = 4

2

Network scale

1

120

160

(b) β = 2

2

0 100

200

14

3 x 10 EMA−SIC Li et al. 2.5

120

times energy α−1

0 100

Latencyα−1 times energy

120

5

140

Network scale

15

3 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

10

Latency

4

120

(a) β = 1

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

4

2 x 10 EMA−SIC Li et al. 1.5

15

14

Latencyα−1 times energy

180

13

8 x 10 EMA−SIC Li et al. 6

1 100

160

6

(b) β = 2

13

0 100

140

Network scale

15

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

1

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

2

0 100

14

13

Latencyα−1 times energy

Latencyα−1 times energy

13

3 x 10 EMA−SIC Li et al.

1.6 x 10 EMA−SIC 1.4 Li et al. 1.2 1 0.8 0.6 100

120

140

160

Network scale

(f) β = 6

Fig. 703: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 200 nodes.

219

0 100

120

(a) β = 1

200

160

180

200

1

120

(c) β = 3 16

160

180

140

160

Network scale

180

200

120

140

160

Network scale

(e) β = 5

180

140

160

180

200

4

2

120

(a) β = 1

6 4 2 0 100

120

140

160

Network scale

180

200

0 100

140

160

Network scale

180

200

2

1.5

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

3 2.5 120

140

160

Network scale

180

200

(f) β = 6

Fig. 705: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 400 nodes.

140

160

Network scale

200

12 x 10 EMA−SIC Li et al. 10

8

6 100

120

140

160

Network scale

(f) β = 6

17

1.5 1 0.5 0 100

120

140

160

Network scale

180

200

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

12 x 10 EMA−SIC 10 Li et al. 8 6 4 2 100

120

140

160

Network scale

18

1.2 1 0.8 120

180

(d) β = 4

1.6 x 10 EMA−SIC 1.4 Li et al.

0.6 100

160

17

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

(b) β = 2

18

3.5

2 100

120

(c) β = 3

4.5 x 10 EMA−SIC 4 Li et al.

180

1

0 100

17

120

200

2

(a) β = 1

5

180

3

17

2.5 x 10 EMA−SIC 2 Li et al.

200

10

200

Fig. 706: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 500 nodes.

17

Latencyα−1 times energy

17

Latencyα−1 times energy

180

180

(d) β = 4

3

(d) β = 4

2.5 x 10 EMA−SIC Li et al.

120

160

15 x 10 EMA−SIC Li et al.

(c) β = 3

1 100

140

Network scale

16

Latencyα−1 times energy

Latencyα−1 times energy

16

200

4

(b) β = 2

10 x 10 EMA−SIC 8 Li et al.

180

17

6 x 10 EMA−SIC Li et al.

0 100

160

5

120

160

5 x 10 EMA−SIC 4 Li et al.

(e) β = 5

Latencyα−1 times energy

Latencyα−1 times energy

Latencyα−1 times energy

1

140

Network scale

7 x 10 EMA−SIC 6 Li et al.

2 100

200

16

Network scale

120

(f) β = 6

2

120

0 100

140

Network scale

(b) β = 2

0.5

Latencyα−1 times energy

6 4 100

120

17

17

8

16

0 100

(c) β = 3

Fig. 704: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 300 nodes.

3 x 10 EMA−SIC Li et al.

200

1

200

12 x 10 EMA−SIC Li et al. 10

Latencyα−1 times energy

3

120

140

Network scale

16

4

180

1.5

(d) β = 4

6 x 10 EMA−SIC Li et al. 5

160

(a) β = 1

2

0 100

140

Network scale

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

140

Network scale

120

17

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

120

0 100

0 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

Latencyα−1 times energy

180

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

2 100

160

16

16

0 100

140

Network scale

Latencyα−1 times energy

200

0.5

Latencyα−1 times energy

180

2

1

Latencyα−1 times energy

160

4

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

140

Network scale

5

6

Latencyα−1 times energy

120

10

17

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

4

0 100

16

15

15 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

15

8 x 10 EMA−SIC Li et al. 6

3 x 10 EMA−SIC Li et al. 2.5 2 1.5 1 100

120

140

160

Network scale

(f) β = 6

Fig. 707: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 600 nodes.

220

200

0 100

120

(a) β = 1

140

160

Network scale

180

200

1 0.5 0 100

120

1.5

140

160

Network scale

180

200

140

160

180

200

4 3 120

140

160

Network scale

180

200

140

160

180

200

120

2

1

140

160

Network scale

180

200

18

200

140

160

Network scale

5 4 3 140

160

Network scale

(e) β = 5

180

200

180

200

6

120

140

160

Network scale

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 709: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 800 nodes.

160

200

180

200

1.4 1.2 1 0.8 100

120

140

160

Network scale

(f) β = 6

18

120

140

160

Network scale

180

200

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

4 2

120

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

19

1.5

1

120

180

(d) β = 4

2 x 10 EMA−SIC Li et al.

0.5 100

160

18

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

(b) β = 2

19

8

140

Network scale

1.8 x 10 EMA−SIC 1.6 Li et al.

(c) β = 3

12 x 10 EMA−SIC Li et al. 10

4 100

120

19

18

1 120

0 100

(a) β = 1

2

180

2

(d) β = 4

1

0 100

18

Latencyα−1 times energy

Latencyα−1 times energy

180

3

0 100

200

2

(d) β = 4

7 x 10 EMA−SIC 6 Li et al.

120

160

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

2 100

140

Network scale

18

Latencyα−1 times energy

Latencyα−1 times energy

18

180

3 x 10 EMA−SIC Li et al.

(b) β = 2

3 x 10 EMA−SIC Li et al.

120

times energy

0.5

(a) β = 1

0 100

120

200

Fig. 710: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 900 nodes.

α−1

1

0 100

160

6

Latency

2

140

Network scale

8

4 100

180

4

18

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

4

Network scale

120

160

18

12 x 10 EMA−SIC Li et al. 10

18

6

120

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

(e) β = 5

Fig. 708: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 700 nodes.

17

120

(b) β = 2

1

(f) β = 6

10 x 10 EMA−SIC 8 Li et al.

1

0 100

18

5

(e) β = 5

200

2

(c) β = 3

7 x 10 EMA−SIC 6 Li et al.

2 100

180

2

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

2.5

160

3

18

18

140

Network scale

5 x 10 EMA−SIC 4 Li et al.

(d) β = 4

3.5 x 10 EMA−SIC 3 Li et al.

Network scale

120

18

1.5

(c) β = 3

0 100

0 100

4 x 10 EMA−SIC Li et al. 3

(a) β = 1

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

5

120

200

18

10

1 100

180

Latencyα−1 times energy

17

120

160

(b) β = 2

15 x 10 EMA−SIC Li et al.

0 100

140

Network scale

Latencyα−1 times energy

180

Latencyα−1 times energy

160

0.5

Latencyα−1 times energy

140

Network scale

1

Latencyα−1 times energy

120

2

Latencyα−1 times energy

0 100

4

Latencyα−1 times energy

1

6

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

3

18

18

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

17

10 x 10 EMA−SIC 8 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

17

5 x 10 EMA−SIC 4 Li et al.

3.5 x 10 EMA−SIC Li et al. 3 2.5 2 1.5 100

120

140

160

Network scale

(f) β = 6

Fig. 711: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and poisson distribution with 1000 nodes.

221

200

120

(a) β = 1

2

140

160

Network scale

180

200

0 100

120

140

160

Network scale

180

200

140

160

(e) β = 5

180

200

Latencyα−1 times energy

1

200

0 100

2

140

160

Network scale

120

140

160

Network scale

180

200

180

200

(f) β = 6

Fig. 712: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 100 nodes.

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

8

1.5 1 0.5 120

180

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

0 100

160

7

1

120

140

Network scale

(b) β = 2

8

4

120

180

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

0 100

160

2

0 100

7

2

140

Network scale

4 x 10 EMA−SIC Li et al. 3

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

Network scale

120

7

5

7

Latencyα−1 times energy

0 100

1 0.5

(a) β = 1

10

(c) β = 3

120

200

6

4

0 100

180

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

6

120

160

1

(b) β = 2

6 x 10 EMA−SIC Li et al.

0 100

140

Network scale

2

Latencyα−1 times energy

180

3

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

160

Latencyα−1 times energy

140

Network scale

0.5 100

Latencyα−1 times energy

120

1

7

5 x 10 EMA−SIC 4 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

1.5

Latencyα−1 times energy

4

0 100

6

6

2.5 x 10 EMA−SIC Li et al. 2

Latencyα−1 times energy

Latencyα−1 times energy

5

8 x 10 EMA−SIC Li et al. 6

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 713: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 200 nodes.

222

7

(a) β = 1

140

160

Network scale

180

200

120

8

160

180

200

140

160

180

200

5

120

(e) β = 5

140

160

Network scale

180

200

140

160

Network scale

180

200

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

5

120

(a) β = 1

1

120

140

160

Network scale

180

200

180

200

0 100

120

140

160

180

200

0.5

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

1

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 715: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 400 nodes.

140

160

Network scale

200

180

200

9

140

160

Network scale

180

200

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

(f) β = 6

8

10

5

0 100

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Li et al. 3 2 1

0 100

120

6 4 2 120

140

160

Network scale

180

200

200

180

200

180

200

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

9

3 2 1 120

180

(d) β = 4

5 x 10 EMA−SIC 4 Li et al.

0 100

160

9

10 x 10 EMA−SIC 8 Li et al.

0 100

140

Network scale

(b) β = 2

9

2

0 100

0 100

(c) β = 3

3 x 10 EMA−SIC Li et al.

180

5

(d) β = 4

15 x 10 EMA−SIC Li et al.

9

2 x 10 EMA−SIC Li et al. 1.5

200

8

Network scale

200

10

(a) β = 1

2

180

Fig. 716: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 500 nodes.

(d) β = 4

9

Latencyα−1 times energy

160

4

(c) β = 3

0 100

120

8

2

0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

Latencyα−1 times energy

8

180

1

(b) β = 2

3 x 10 EMA−SIC Li et al.

160

7

10

0 100

140

Network scale

2

0 100

160

8

4 x 10 EMA−SIC Li et al. 3

7

2

120

120

140

Network scale

15 x 10 EMA−SIC Li et al.

(e) β = 5

Fig. 714: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 300 nodes.

7

120

(b) β = 2

2

(f) β = 6

4 x 10 EMA−SIC Li et al. 3

0 100

9

10

0 100

200

4

0 100

Latencyα−1 times energy

2

180

(c) β = 3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

8

4

160

8

1

0 100

140

Network scale

6 x 10 EMA−SIC Li et al.

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

Network scale

120

(a) β = 1

2

(c) β = 3

0 100

0 100

Latencyα−1 times energy

5

120

200

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

Latencyα−1 times energy

10

0 100

180

8

7

120

160

(b) β = 2

15 x 10 EMA−SIC Li et al.

0 100

140

Network scale

Latencyα−1 times energy

120

1 0.5

Latencyα−1 times energy

0 100

1.5

Latencyα−1 times energy

200

2

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

180

4

Latencyα−1 times energy

160

8

8 x 10 EMA−SIC Li et al. 6

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

2

Latencyα−1 times energy

0.5

4

Latencyα−1 times energy

Latencyα−1 times energy

1

7

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

7

2 x 10 EMA−SIC Li et al. 1.5

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

Fig. 717: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 600 nodes.

223

(a) β = 1

140

160

Network scale

180

200

1

120

140

160

Network scale

180

200

160

180

200

5

120

140

160

Network scale

(e) β = 5

180

200

160

180

200

4 2

120

(a) β = 1

120

140

160

Network scale

180

200

200

140

160

Network scale

180

200

10

5

140

160

(e) β = 5

180

200

1

0 100

120

0 100

140

160

Network scale

120

180

200

(f) β = 6

Fig. 719: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 800 nodes.

140

160

Network scale

200

180

200

(d) β = 4

140

160

Network scale

180

200

3 x 10 EMA−SIC Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

8

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

1

120

140

160

Network scale

180

200

200

180

200

180

200

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

10

1 0.5

120

180

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

0 100

160

(b) β = 2

2

0 100

140

Network scale

9

4 x 10 EMA−SIC Li et al. 3

10

0.5

180

2

(c) β = 3

2 x 10 EMA−SIC Li et al. 1.5

200

4

(a) β = 1

1 120

200

9

2

180

Fig. 720: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 900 nodes.

10

Latencyα−1 times energy

Latencyα−1 times energy

180

3

0 100

9

Network scale

120

(d) β = 4

15 x 10 EMA−SIC Li et al.

120

160

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

0 100

0 100

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

0 100

140

Network scale

160

10

(b) β = 2

2 x 10 EMA−SIC Li et al. 1.5

180

5

9

9

160

8

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

10

Latencyα−1 times energy

140

Latencyα−1 times energy

Latencyα−1 times energy

1

Network scale

120

15 x 10 EMA−SIC Li et al.

8

2

120

0.5 0 100

140

Network scale

8 x 10 EMA−SIC Li et al. 6

(e) β = 5

Fig. 718: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 700 nodes.

8

120

(b) β = 2

1

(f) β = 6

4 x 10 EMA−SIC Li et al. 3

0 100

9

9

10

0 100

200

1.5

Latencyα−1 times energy

140

180

(c) β = 3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

160

2.5 x 10 EMA−SIC 2 Li et al.

9

4

140

Network scale

(a) β = 1

2

0 100

9

Network scale

120

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

0 100

9

4 x 10 EMA−SIC Li et al. 3

(c) β = 3

120

200

Latencyα−1 times energy

5

Latencyα−1 times energy

Latencyα−1 times energy

10

0 100

180

(b) β = 2

15 x 10 EMA−SIC Li et al.

120

160

9

8

0 100

140

Network scale

Latencyα−1 times energy

120

5

Latencyα−1 times energy

0 100

1

Latencyα−1 times energy

200

2

10

Latencyα−1 times energy

180

3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

160

2

8

5 x 10 EMA−SIC 4 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

140

Network scale

4

Latencyα−1 times energy

120

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

1 0.5

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0 100

8

8

8

2.5 x 10 EMA−SIC 2 Li et al.

4 x 10 EMA−SIC Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 721: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 3 and cluster distribution with 1000 nodes.

224

9

120

(a) β = 1

140

160

Network scale

180

200

120

140

160

Network scale

180

200

(d) β = 4

140

160

Network scale

(e) β = 5

180

200

0 100

120

1

140

160

Network scale

180

200

(f) β = 6

140

160

Network scale

180

200

Fig. 722: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 100 nodes.

180

200

180

200

180

200

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

140

160

Network scale

12

10

5

120

160

(d) β = 4

15 x 10 EMA−SIC Li et al.

0 100

140

Network scale

11

11

0.5

120

120

(c) β = 3

2 x 10 EMA−SIC Li et al. 1.5

0 100

0 100

(b) β = 2

0.5

Latencyα−1 times energy

2

200

1

140

160

Network scale

180

200

2.5 x 10 EMA−SIC 2 Li et al. 1.5 1 0.5 0 100

120

(e) β = 5

140

160

Network scale

(f) β = 6

Fig. 723: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 200 nodes.

11

Latencyα−1 times energy

4

Latencyα−1 times energy

6

180

2 x 10 EMA−SIC Li et al. 1.5

11

10 x 10 EMA−SIC 8 Li et al.

160

11

2

120

140

Network scale

5

(a) β = 1

3

1 100

10

Latencyα−1 times energy

2

0 100

5 x 10 EMA−SIC Li et al. 4

(c) β = 3

120

200

Latencyα−1 times energy

5

Latencyα−1 times energy

Latencyα−1 times energy

10

0 100

Network scale

180

10

15 x 10 EMA−SIC Li et al.

120

160

(b) β = 2

9

0 100

140

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

10

Latencyα−1 times energy

Network scale

180

4

15 x 10 EMA−SIC Li et al.

11

4 x 10 EMA−SIC Li et al. 3 2 1

0 100

120

140

160

Network scale

180

200

Latencyα−1 times energy

160

10

8 x 10 EMA−SIC Li et al. 6 4 2

0 100

120

(a) β = 1 11

6 4 2 120

140

160

Network scale

180

200

200

180

200

180

200

1.5 1 0.5 100

120

140

160

Network scale

(d) β = 4

12

12

6 x 10 EMA−SIC Li et al. 4

2

120

180

2.5 x 10 EMA−SIC Li et al. 2

(c) β = 3

0 100

160

12

10 x 10 EMA−SIC 8 Li et al.

0 100

140

Network scale

(b) β = 2 Latencyα−1 times energy

140

2

Latencyα−1 times energy

120

4

Latencyα−1 times energy

0 100

10

8 x 10 EMA−SIC Li et al. 6

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

8 x 10 EMA−SIC Li et al. 6

Latencyα−1 times energy

2

Latencyα−1 times energy

Latencyα−1 times energy

9

4 x 10 EMA−SIC Li et al. 3

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

(f) β = 6

Fig. 724: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 300 nodes.

225

200

0 100

120

(a) β = 1

1

140

160

Network scale

180

200

2

120

0.5

140

160

180

200

Latencyα−1 times energy

140

160

Network scale

180

200

2 1

120

140

160

Network scale

(e) β = 5

180

200

140

160

Network scale

180

200

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

(a) β = 1

180

200

180

0 100

140

160

Network scale

180

200

140

160

Network scale

180

200

120

(c) β = 3

4 2 0 100

120

140

160

Network scale

180

200

0 100

160

180

1 140

160

(e) β = 5

180

200

Latencyα−1 times energy

2

Network scale

120

120

140

160

Network scale

180

200

(f) β = 6

Fig. 726: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 500 nodes.

200

1 0.5 0 100

120

160

180

200

160

180

200

180

200

180

200

4

2

0 100

120

140

160

Network scale

(d) β = 4 14

15 x 10 EMA−SIC Li et al. 10

5

0 100

140

Network scale

6 x 10 EMA−SIC Li et al.

(c) β = 3

2

0 100

140

Network scale

13

4

160

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

1

0 100

200

8 x 10 EMA−SIC Li et al. 6

140

Network scale

13

2

(d) β = 4

3

120

(b) β = 2

3 x 10 EMA−SIC Li et al.

13

5 x 10 EMA−SIC 4 Li et al.

120

140

Network scale

180

5

13

6

times energy α−1

0 100

200

10

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

120

160

15 x 10 EMA−SIC Li et al.

(a) β = 1

0.5

140

Network scale

(f) β = 6

10 x 10 EMA−SIC 8 Li et al.

Latency

2

120

Fig. 727: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 600 nodes.

200

1

180

1

13

2 120

200

(d) β = 4

4

0 100

180

2

13

Latencyα−1 times energy

Latencyα−1 times energy

160

2 x 10 EMA−SIC Li et al. 1.5

13

Latencyα−1 times energy

140

Network scale

13

4

0 100

160

6

(b) β = 2

12

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

0.5

Latencyα−1 times energy

Latencyα−1 times energy

1

120

120

160

13

0.5 0 100

140

Network scale

4 x 10 EMA−SIC Li et al. 3

12

12

1.5

120

(e) β = 5

Fig. 725: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 400 nodes.

12

0 100

(b) β = 2

1

(f) β = 6

2.5 x 10 EMA−SIC 2 Li et al.

200

13

4 x 10 EMA−SIC Li et al. 3

0 100

180

(c) β = 3

Latencyα−1 times energy

Latencyα−1 times energy

1

160

13

4

0 100

140

Network scale

2 x 10 EMA−SIC Li et al. 1.5

13

13

Network scale

120

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

0 100

0 100

5

(a) β = 1

8 x 10 EMA−SIC Li et al. 6

(c) β = 3

120

200

Latencyα−1 times energy

2

0 100

180

12

Latencyα−1 times energy

Latencyα−1 times energy

12

120

160

(b) β = 2

4 x 10 EMA−SIC Li et al. 3

0 100

140

Network scale

Latencyα−1 times energy

180

10

Latencyα−1 times energy

160

2

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

140

Network scale

1 0.5

4

Latencyα−1 times energy

120

1.5

12

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

5

Latencyα−1 times energy

Latencyα−1 times energy

10

0 100

12

12

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

11

15 x 10 EMA−SIC Li et al.

3 x 10 EMA−SIC Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 728: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 700 nodes.

226

(a) β = 1

140

160

Network scale

180

200

4 2 0 100

120

1 0.5 160

180

200

Latencyα−1 times energy

Latencyα−1 times energy

1.5

140

160

Network scale

180

1 140

160

Network scale

180

200

0 100

120

140

160

Network scale

180

200

4

2

120

(a) β = 1

160

180

200

200

120

140

160

Network scale

180

200

140

160

(e) β = 5

180

200

Latencyα−1 times energy

1

Latencyα−1 times energy

0 100

2

120

140

160

Network scale

120

180

200

(f) β = 6

Fig. 730: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 900 nodes.

140

160

Network scale

180

200

180

200

(d) β = 4 14

140

160

Network scale

180

200

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

13

2

1

120

140

160

Network scale

180

200

6 x 10 EMA−SIC Li et al. 4

2

0 100

120

2

120

140

160

Network scale

180

200

200

180

200

180

200

15 x 10 EMA−SIC Li et al. 10

5

0 100

120

140

160

Network scale

14

1.5 1 0.5 120

180

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

0 100

160

(b) β = 2

4

0 100

140

Network scale

13

8 x 10 EMA−SIC Li et al. 6

14

4

0 100

200

3 x 10 EMA−SIC Li et al.

0 100

14

2

180

1

(c) β = 3

8 x 10 EMA−SIC Li et al. 6

200

Fig. 731: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 1000 nodes.

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

Network scale

120

180

0.5

(a) β = 1

0.5

14

Latencyα−1 times energy

180

1

0 100

160

2

0 100

160

1.5

13

(c) β = 3

120

160

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

Latencyα−1 times energy

140

Network scale

140

Network scale

4

(b) β = 2

2

0 100

140

Network scale

14

4

120

120

140

Network scale

(b) β = 2

6 x 10 EMA−SIC Li et al.

times energy

6 x 10 EMA−SIC Li et al.

0 100

13

8 x 10 EMA−SIC Li et al. 6

0 100

0 100

α−1

Latencyα−1 times energy

Latencyα−1 times energy

1

120

2.5 x 10 EMA−SIC 2 Li et al.

13

13

2

0 100

(e) β = 5

Fig. 729: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 4 and cluster distribution with 800 nodes.

13

2

14

5

(f) β = 6

3 x 10 EMA−SIC Li et al.

4

(c) β = 3

2

(e) β = 5

200

14

3

120

180

10

200

5 x 10 EMA−SIC 4 Li et al.

0 100

160

15 x 10 EMA−SIC Li et al.

(d) β = 4

2.5 x 10 EMA−SIC 2 Li et al.

140

Network scale

8 x 10 EMA−SIC Li et al. 6

(a) β = 1

14

140

120

13

6

(c) β = 3 14

Network scale

0 100

Latencyα−1 times energy

1

120

200

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

180

13

3

120

160

(b) β = 2

13

5 x 10 EMA−SIC 4 Li et al.

0 100

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

160

1

Latencyα−1 times energy

140

Network scale

2

Latency

120

1

Latencyα−1 times energy

0 100

2

13

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

0.5

13

4 x 10 EMA−SIC Li et al. 3

Latencyα−1 times energy

Latencyα−1 times energy

1

Latencyα−1 times energy

13

13

2 x 10 EMA−SIC Li et al. 1.5

5 x 10 EMA−SIC Li et al. 4 3 2

1 100

120

140

160

Network scale

(f) β = 6

Fig. 732: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 100 nodes.

227

15

120

(a) β = 1

1

140

160

Network scale

180

200

1 140

160

Network scale

180

200

140

160

180

200

5

120

(e) β = 5

140

160

Network scale

180

200

160

180

200

10

5

0 100

120

(a) β = 1

140

160

Network scale

180

200

Latencyα−1 times energy

Latencyα−1 times energy

0.5

16

200

3 2 1

0 100

140

160

Network scale

(e) β = 5

120

140

160

Network scale

180

200

180

200

2 120

140

160

Network scale

180

200

(f) β = 6

Fig. 734: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 300 nodes.

200

0 100

120

140

160

Network scale

180

200

180

200

(d) β = 4 17

140

160

Network scale

180

200

4 x 10 EMA−SIC Li et al. 3 2 1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 735: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 400 nodes.

17

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

180

200

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100

120

1 0.5 120

140

160

Network scale

180

200

200

180

200

180

200

5 x 10 EMA−SIC 4 Li et al. 3 2 1

0 100

120

140

160

Network scale

17

4 2

120

180

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

160

(b) β = 2

1.5

0 100

140

Network scale

17

2.5 x 10 EMA−SIC 2 Li et al.

17

4

180

5

(c) β = 3

6

160

10

(a) β = 1

10 x 10 EMA−SIC 8 Li et al.

0 100

200

17

16

Latencyα−1 times energy

Latencyα−1 times energy

180

1

120

180

0.5

(d) β = 4

5 x 10 EMA−SIC 4 Li et al.

120

160

2

(c) β = 3

0 100

140

Network scale

4 x 10 EMA−SIC Li et al. 3

0 100

160

1

(b) β = 2

1

140

Network scale

1.5

16

16

120

120

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

140

Network scale

2 x 10 EMA−SIC Li et al. 1.5

0 100

0 100

140

Network scale

15 x 10 EMA−SIC Li et al.

16

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

120

2

15

4

120

(e) β = 5

Fig. 733: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 200 nodes.

15

0 100

16

4

(f) β = 6

8 x 10 EMA−SIC Li et al. 6

2

(b) β = 2

17

10

0 100

200

6

Latencyα−1 times energy

2

180

10 x 10 EMA−SIC 8 Li et al.

15

4

160

4

(c) β = 3

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

times energy

2

120

140

Network scale

16

3

0 100

15

Network scale

120

(d) β = 4

8 x 10 EMA−SIC Li et al. 6

0 100

0 100

6 x 10 EMA−SIC Li et al.

(a) β = 1

5 x 10 EMA−SIC 4 Li et al.

(c) β = 3

120

200

Latencyα−1 times energy

2

0 100

180

15

Latencyα−1 times energy

Latencyα−1 times energy

15

120

160

(b) β = 2

3 x 10 EMA−SIC Li et al.

0 100

140

Network scale

Latencyα−1 times energy

0 100

Latencyα−1 times energy

200

Latencyα−1 times energy

180

1

Latencyα−1 times energy

160

2

Latencyα−1 times energy

140

Network scale

Latencyα−1 times energy

120

Latencyα−1 times energy

0 100

0.5

16

3 x 10 EMA−SIC Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

2

1

α−1

4

Latencyα−1 times energy

Latencyα−1 times energy

6

16

2 x 10 EMA−SIC Li et al. 1.5

Latency

14

10 x 10 EMA−SIC 8 Li et al.

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

Fig. 736: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 500 nodes.

228

120

(a) β = 1

140

160

Network scale

180

200

0 100

120

140

160

Network scale

180

200

140

160

180

200

1 0.5 0 100

120

140

160

Network scale

180

200

160

180

200

4 2 0 100

120

(a) β = 1 17

120

140

160

Network scale

180

200

0 100

120

140

160

Network scale

180

200

0 100

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

1

2 1 120

160

180

200

140

160

Network scale

180

200

(f) β = 6

Fig. 738: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 700 nodes.

140

160

Network scale

Latencyα−1 times energy

200

10 x 10 EMA−SIC 8 Li et al. 6 4 2 0 100

120

140

160

Network scale

(f) β = 6

18

120

140

160

Network scale

180

200

4 x 10 EMA−SIC Li et al. 3 2 1

0 100

120

2 1 120

140

160

Network scale

180

200

200

180

200

180

200

8 x 10 EMA−SIC Li et al. 6 4 2

0 100

120

140

160

Network scale

19

10

5

120

180

(d) β = 4

15 x 10 EMA−SIC Li et al.

0 100

160

18

3

0 100

140

Network scale

(b) β = 2

18

3

0 100

140

Network scale

5 x 10 EMA−SIC 4 Li et al.

18

2

120

(c) β = 3

5 x 10 EMA−SIC 4 Li et al.

180

1

0 100

18

0.5

200

2

(a) β = 1

1

18

Latencyα−1 times energy

200

180

3

18

0.5

(d) β = 4

4 x 10 EMA−SIC Li et al. 3

120

180

1.5

(c) β = 3

0 100

160

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

5

0 100

140

Network scale

200

(d) β = 4

1

18

10

200

2

(b) β = 2

15 x 10 EMA−SIC Li et al.

180

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

140

Network scale

6

180

Fig. 739: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 800 nodes.

Latencyα−1 times energy

120

160

18

10 x 10 EMA−SIC 8 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

1

140

Network scale

4

120

160

(b) β = 2

8 x 10 EMA−SIC Li et al. 6

0 100

17

2

0 100

120

140

Network scale

5 x 10 EMA−SIC 4 Li et al.

(e) β = 5

Fig. 737: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 600 nodes.

3

120

18

1

(f) β = 6

17

0 100

18

1.5

(e) β = 5

5 x 10 EMA−SIC 4 Li et al.

200

2

0 100

Latencyα−1 times energy

0.5

180

(c) β = 3

2.5 x 10 EMA−SIC 2 Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

times energy α−1

5

18

1

160

(a) β = 1

10

18

140

Network scale

3 x 10 EMA−SIC Li et al.

(d) β = 4

2 x 10 EMA−SIC Li et al. 1.5

Network scale

120

18

(c) β = 3

120

0 100

200

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

2

0 100

180

17

4

120

160

(b) β = 2

17

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

Latencyα−1 times energy

0 100

0.5

Latencyα−1 times energy

200

2

1

Latencyα−1 times energy

180

4

2 x 10 EMA−SIC Li et al. 1.5

Latencyα−1 times energy

160

1

6

Latency

140

Network scale

2

Latencyα−1 times energy

120

3

18

10 x 10 EMA−SIC 8 Li et al.

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

0.5

5 x 10 EMA−SIC 4 Li et al.

Latencyα−1 times energy

1

Latencyα−1 times energy

Latencyα−1 times energy

1.5

0 100

17

17

17

2.5 x 10 EMA−SIC 2 Li et al.

2 x 10 EMA−SIC Li et al. 1.5 1 0.5 0 100

120

140

160

Network scale

(f) β = 6

Fig. 740: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 900 nodes.

229

18

2

1

0 100

120

140

160

Network scale

180

200

6 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

18

3 x 10 EMA−SIC Li et al.

4

2

0 100

120

(a) β = 1

2

140

160

Network scale

180

200

180

200

180

200

5

0 100

120

140

160

Network scale

(d) β = 4 19

2 x 10 EMA−SIC Li et al. 1.5 1 0.5

120

140

160

Network scale

(e) β = 5

180

200

Latencyα−1 times energy

Latencyα−1 times energy

200

10

(c) β = 3 19

0 100

180

18

4

120

160

15 x 10 EMA−SIC Li et al.

Latencyα−1 times energy

Latencyα−1 times energy

18

8 x 10 EMA−SIC Li et al. 6

0 100

140

Network scale

(b) β = 2

3 x 10 EMA−SIC Li et al. 2

1

0 100

120

140

160

Network scale

(f) β = 6

Fig. 741: Impact of network scale on the latency-energy tradeoff: a separate comparison between EMA-SIC and Li et al.’s algorithm under α = 5 and cluster distribution with 1000 nodes.