Implementing Top-k Query in Duty-Cycled Wireless Sensor Networks Chunsheng Zhu∗ , Laurence T. Yang∗ , Lei Shu† , Takahiro Hara† , Shojiro Nishio† ∗ Department

of Computer Science, St. Francis Xavier University, Antigonish, Canada B2G 2W5 Email: {chunsheng.tom.zhu, ltyang}@gmail.com † Department of Multimedia Engineering, Osaka University, Osaka, Japan 565-0871 Email: {lei.shu, hara, nishio}@ist.osaka-u.ac.jp

Abstract—Top-k query is a very useful and important query in wireless sensor networks (WSNs), aiming to find the k nodes with highest readings among the sensor nodes. In WSNs, there are generally two kinds of networks: always-on WSNs (AO-WSNs) in which sensors always keep awake and duty-cycled WSNs (DCWSNs) where sensors dynamically sleep and wake. To the best of our knowledge, there are a lot of work about top-k query in AOWSNs but little research has been done regarding top-k query in DC-WSNs. However, DC-WSN is a very practical network model in which energy consumption can be greatly saved. In this paper, we analyze the research issues when implementing top-k query in DC-WSNs and propose the DCDC-WSNs (DCWSNs with data replication (DR) and connected k-neighborhood (CKN)) to implement top-k query. Further theoretical analysis and simulation results show that implementing top-k query in DCDC-WSNs can achieve the best tradeoff with respect to query data accessibility and query cost (total energy consumption, query response time), compared with AO-WSNs and DCC-WSNs (DC-WSNs with only CKN). Keywords—Top-k query; WSNs; Always-on WSNs; Dutycycled WSNs; Data replication; Connected k-neighborhood

I. I NTRODUCTION Wireless sensor networks (WSNs) draw a lot of attention because of their potential applications in military, industrial and civilian areas (e.g., battle damage assessment, industrial process control and environmental monitoring). Among all kinds of queries in WSNs, top-k query1 is a very popular and important query in WSNs as it tries to find the k nodes with the highest readings among all sensor nodes which has many wonderful application scenarios (e.g., when forest fire happens, we may issue a top-k query to request the vicinity images of k sensors with the highest temperature readings in the forest). In WSNs, there are generally two types of networks. The first one is the always-on WSNs (AO-WSNs) and the other one is the duty-cycled WSNs (DC-WSNs). In AO-WSNs, sensors always be awake and perform various tasks, e.g., sensing and communication, but with the cost of high energy consumption. In DC-WSNs, sensors dynamically 1 There are two types of top-k queries. Assuming each object has a ranking score which is an aggregated value from a number of attribute values stored at distributed sources, one top-k query is the distributed top-k query aiming to find the k highest ranked objects. Assuming that each node has a reading, another top-k query is to find the k nodes with the highest readings in a sensor network. The top-k query this paper researches belongs to the second type.

c 2011 IEEE 978-1-4244-9538-2/11/$26.00

be awake and asleep and sporadically perform sensing and communication. As only a subset of nodes are awake in each epoch which places the remaining nodes in a minimal power sleep state, DC-WSNs have a natural advantage of saving energy consumption, which is a fundamental and practical design factor in WSNs due to the fact that most sensors are usually equipped with non-rechargeable batteries with limited energy. To the best of our knowledge, current research work about top-k query mainly focus on AO-WSNs [1] [2] [3] [4] [5]. In addition, little research work has been done to explore the research issues when implementing top-k query in DCWSNs. In this paper, we analyze detailed research problems about top-k query in DC-WSNs and put forward the DCDCWSNs (DC-WSNs with data replication (DR) and connected k-neighborhood (CKN)) to implement top-k query. Further theoretical analysis and simulations show that DCDC-WSNs can obtain the best tradeoff regarding query data accessibility and query cost (total energy consumption, response time) compared with AO-WSNs and DCC-WSNs (DC-WSNs with only CKN) when implementing top-k query. The rest parts of this paper are organized as follows. Section II introduces the research problems when implementing top-k query in DC-WSNs and proposes the DCDC-WSNs. Section III describes the preliminaries and section IV analyzes the CKN algorithm. Section V and section VI present the theoretical analysis about DCDC-WSNs, DCC-WSNs and AO-WSNs. Evaluations about the three networks are shown in section VII. Section VIII concludes this paper. II. R ESEARCH PROBLEMS AND PROPOSED METHOD A. Research problems When implementing top-k query in the DC-WSNs, the first problem should be considered is the data accessibility problem in DC-WSNs, due to that sensors may be asleep and data cannot be achieved from the asleep nodes as shown in Fig. 1 (a). The second problem should be solved is the uncertainty of network connectivity in DC-WSNs. Specially, the dynamically awake and asleep duty cycle will make the link highly unstable and query cannot get data if the network is not connected, although awake sensors own the data which is shown in Fig. 1 (b).

B. Energy model A

B

C

D

B

C

I

J

S

E

F

S

G

H

F

(a) Data accessibility problem

G

(b) Network connectivity problem

Fig. 1. Top-k query problems in DC-WSNs: gray nodes represent the nodes that are awake and black nodes represent the nodes that are asleep, node S is the base station. In Fig. 1 (a), because nodes A, D, E, H are asleep, data owned by them cannot achieved by S. In Fig. 1 (b), although nodes B, C, F, G, I, J are all awake and they have the data, because the whole network is not connected, only data owned by nodes B, C, F, G can be achieved by S and data from nodes I, J cannot be obtained by S.

We take the first order model in [7] as our energy model. Each packet consists of a packet header and a packet body containing readings. The consumed energy to transmit and receive a packet represented by m bytes over distance d are E T , ER : ET = te · m + ae · m · d2

(1)

E R = re · m

(2)

And the energy consumption of nodes in the idle radio mode is approximately the same with that in the receiving radio mode [8]. E I = ER

B. Proposed method With respect to the data accessibility problem in DCWSNs, we first employ data replication (DR) into DC-WSNs. Specially, we replicate each node’s data to some neighboring nodes within 1-hop. Regarding the connectivity problem in DC-WSNs, we try to use a sleep scheduling algorithm named connected k-neighborhood (CKN) proposed in [6] which can change the sleep/wake state of every node in a DC-WSN to make the whole network always connected by the awake nodes and every node always has at least some certain awake neighboring nodes. III. P RELIMINARIES A. System model A multihop WSN with N sensor nodes is modeled by a graph G = (U, L), where U = {u1 , u2 , · · · , uN } is the set of sensor nodes, L is the set of links. There is a base station s with unlimited energy supply which serves as a gateway between the sensor network and users. Any two sensors u and v are neighbors if they are within the transmission range of each other and the link l(u,v) ∈ L performs bidirectional direct communication. Any two sensors u and v are 2-hop neighbors if l(u,v) ∈ / L and there exists another node w satisfying l(u,w) ∈ L, l(w,v) ∈ L or l(v,w) ∈ L, l(w,u) ∈ L. Each node has the same transmission radius tr without interference. The energy consumption of a sensor by transmitting, receiving one byte and transmitting amplifier are te mJ, re mJ and ae mJ/m2 , respectively. Sensor nodes are deployed with a uniformly random distribution in a two dimension plane A. Time is divided into epoches and each epoch is T . All packets have the same size. A tree is established for top-k query from the base station s (as root), by broadcasting a tree-constructing message in the WSN. When forwarding the tree-constructing message, the 1-hop neighbors of every node will become its children nodes in the tree. Top-k query is issued at s, broadcasted to the entire network and every node sends its top-k results to its parent node.

(3)

Assuming wun represents the probability that a node u has packets to send at the n’th epoch. We can formulate the the energy consumption of a node u in the n’th epoch T is E(wun , T ) = T · wun · (ET + ER ) + T · (1 − wun ) · EI

(4)

IV. CKN ANALYSIS In CKN [6], as each node u picks a random rank ranku and nodes dynamically sleep and wake, we use the following ′ notations. Nu is the set of u’s neighbors and Nu is the set of ′ ′ u’s 2-hop neighbors. Cu , Cu and U are the subsets of Nu , ′ ′ ′ Nu and U whose ranks ≤ ranku . |Nu |, |Nu |, |Cu |, |Cu | and ′ ′ ′ |U | are the number of the elements in Nu , Nu , Cu , Cu and ′ U , respectively. The process of CKN is as follows. At first, each node u picks a random rank ranku , broadcasts its ranku and collects its neighbors Nu ’s ranks in a set Ru . After that, u broadcasts Ru and receives Rv from each v ∈ Nu . If u or any node in Nu has less than k neighbors, u will remain awake. Otherwise, u goes to sleep if it satisfies the following two conditions: 1) “any two nodes in Cu are connected either directly themselves or indirectly through nodes within u’s 2-hop neighborhood that have rank less than ranku ” and 2) “any node in Nu has at least k neighbors from Cu ”. After analysis, we can get that the aforementioned two conditions can be interpreted into the following corresponding conditions: (1) graph GCu +Cu′ is connected and (2) graph GNu is k-connected by nodes in Cu . We calculate these two probabilities. From [9], we can get that for a uniformly random distribution in two dimensions, ρ = N A is the node density and the probability that a random node has n neighbors is (ρπt2r )n −ρπtr 2 (5) ·e n! We can further deduce that the probability that the communication graph G is k-connected can be calculated as Pneighbor (n) =

the probability that there exists at least k different paths connecting any two different vertices in the graph which is

P (Gk−connected ) = (1 −

k−1 X

(ρπt2r )n −ρπtr 2 N ) ·e n! n=0

(6)



With ρ = |UA | , we can get that the probability that graph GCu +Cu′ is connected is ′



P rob1 = (1 − e−ρ



πtr 2 (|Cu |+|Cu |)

)

(7)

And the probability that graph GNu is k-connected by nodes in Cu is k−1 X

ECKN (u) = 2ET + 2|Nu |(ET + ER )

(8)

EDCDC = (9)

(15)

The total energy consumption of DCDC-WSNs during each epoch is

Thus, we can get that the asleep probability and awake probability of u is Pasleep (u) = P rob1 · P rob2

(14)

During top-k query, base station s will broadcast the query through the awake nodes after CKN and each awake node will receive one query packet from its parent node and send one response packet to its parent node. We assume the awake nodes number is Na and leaf nodes number is Nl , there will be 2Na − Nl packets in the network and the query energy consumption is Etopk (Na , Nl ) = (2Na − Nl ) · (ET + ER )

n



′ 2 (ρ πtr 2 ) P rob2 = (1 − · e−ρ πtr )|Cu | n! n=0

In CKN [6], each node must send 2 packets to broadcast ranku , Ru and receive 2|Nu | packets during each epoch and the energy consumption of CKN is

i=N X

Pawake (i) · E(win , T ) + EDR (i) + ECKN (i)

i=1

+ Etopk (NaDCDC , NlDCDC ) (16)

Pawake (u) = 1 − P rob1 · P rob2

(10)

V. DCDC-WSN S In DCDC-WSNs, during each epoch, each node u first replicates data items to randomly selected replication-k 1-hop ′ ′ ′ ′ neighbors u = {u1 , u2 , · · · , ureplication−k }, runs CKN and then performs communication. A. Data accessibility As the data cannot be achieved if u is asleep or u’s replication-k neighbors are asleep, the data accessibility for data owned by u is

C. Response time With a routing tree when implementing the top-k query and the root of the tree is the base station s, from [10], we can assume that the base station s has ds children which are u1 , u2 , · · · , uds and they are indexed in increasing order of their transmission time delays, i.e., t(u1 ) ≤ t(u2 ) ≤ · · · ≤ t(uds ). tu is the transmission time delay at node u and tr(uj , Mj ) is the transmission time used to transmit Mj packets from uj to s with each packet containing no more than k readings, 1 ≤ j ≤ ds . We can get that the response time rt to answer a top-k query at base station s can be defined as the earliest time that s received all top-k readings from its children node which is



DADCDC (u) =1 − Pasleep (u) · Pasleep (u1 ) · · · ′

Pasleep (ureplication−k )

(11) rt = max1≤i≤ds (t(ui ) +

And the data accessibility of the total network during each epoch is

DADCDC =

Pi=N i=1

DADCDC (ui ) N

tr(uj , Mj ))

(17)

j=i

As for DCDC-WSNs, we can assume its response time is RTDCDC =max1≤i≤dsDCDC (t(ui )

(12)

dsDCDC

+

B. Total energy consumption

X

tr(uj , Mj DCDC ))

(18)

j=i

In data replication, every node replicates its data items to randomly selected replication-k neighbors within 1-hop, thus it should broadcast 1 packet and make replication-k nodes receive packets in each epoch and the energy consumption of DR is EDR (u) = ET + replication−k · ER

ds X

(13)

VI. DCC-WSN S AND AO-WSN S For DCC-WSNs, in each epoch, each node u first runs CKN and then communicates with each other. From the above analysis, we can easily get that the data accessibility of the data owned by u, the data accessibility of DCC-WSNs and the total energy consumption of DCC-WSNs during each epoch are

DADCC (u) = 1 − Pasleep (u)

DADCC =

EDCC =

i=N X

Pi=N i=1

DADCC (ui ) N

Pawake (i) · E(win , T ) + ECKN (i)

i=1

(19)

(20)

(21)

+ Etopk (NaDCC , NlDCC ) Regarding AO-WSNs, its data accessibility is almost 100% as sensors are always awake and query cannot miss any data unless sensor dies. Thus, DAAO (u) = 1 − 0 = 100%

(22)

nodes number of DCC-WSNs, DCDC-WSNs and AO-WSNs will be the same given the same prerequisite, i.e., NaDCC = NaDCDC ≤ N and NlDCC = NlDCDC = NlAO . Due to awake sensors in DCC-WSNs generally have less data than the awake sensors in DCDC-WSNs but sensors can transmit multiple data in one packet, so Etopk (NaDCC , NlDCC ) ≃ Etopk (NaDCDC , NlDCDC ) ≤ Etopk (N, NlAO ). From equation (27) (28) (24), we can get EDCC ≈ EDCDC ≤ EAO . About response time, because sensors are always awake in AO-WSNs and sensors are dynamically asleep and awake in DCDC-WSNs and DCC-WSNs, there will be generally more awake children of base station and more packets for them to transmit in AO-WSNs than that in DCDC-WSNs and DCC-WSNs, provided the same prerequisite, i.e., dsDCC ≃ dsDCDC ≤ dsAO and Mj DCC ≃ Mj DCDC ≤ Mj AO , from equation (25) (18) (26), we can get RTDCC ≃ RTDCDC ≤ RTAO . VII. E VALUATION

DAAO ≃ 100%

(23)

And the total energy consumption of AO-WSNs is EAO =

i=N X

E(win , T ) + Etopk (N, NlAO )

(24)

i=1

Regarding the response time for DCC-WSNs and AOWSNs, they are also related to the time that base station s gets all top-k readings from its children. The response times for DCC-WSNs and AO-WSNs are expressed as dsDCC

RTDCC = max1≤i≤dsDCC (t(ui ) +

X

tr(uj , Mj DCC ))

j=i

(25) dsAO

RTAO = max1≤i≤dsAO (t(ui ) +

X

tr(uj , Mj AO )) (26)

j=i

From above analysis, we can clearly get that DADCC ≤ DADCDC ≤ DAAO as DADCC (u) ≤ DADCDC (u) ≤ DAAO (u) with equation (19) (11) (22) (20) (12) (23). Due to the fact that sensors generally transmit quite a lot packets if they are awake, we can assume that EDR (u), ECKN (u) only take a small portion of E(win , T ), thus EDCC ≃

i=N X

Pawake (i) ·

E(win , T )

i=1

(27)

+ Etopk (NaDCC , NlDCC ) EDCDC ≈

i=N X

Pawake (i) · E(win , T )

i=1

(28)

+ Etopk (NaDCDC , NlDCDC ) As both networks have the same CKN, the awake nodes of DCC-WSNs, DCDC-WSNs will be the same and the leaf

In this section, we compare the top-k query performance of DCDC-WSNs with that of DCC-WSNs and AO-WSNs in terms of query data accessibility and query cost (total energy consumption, query response time). A. Evaluation setup We use the simulator NetTopo [11] to perform the simulation runs. The network size is 800 × 600 m2 . The number of total sensor nodes N ranges from 100 to 1000 (each time with an increment of 100 nodes). Both the value of k in CKN and replication-k vary from 1 to 10 (each time increased by 1). For every k, N and replication-k, 100 randomly network topologies are generated and one base station is deployed at the center of the network region. The transmission radius of each node is 100 m. And the energy consumption of a sensor by transmitting, receiving one byte and transmitting amplifier are 0.0144 mJ [10], 0.00576 mJ [10] and 0.0288 nJ/m2 respectively. The time delay for transmitting one byte is 0.4 ms [10]. Each packet is represented by 12 bytes [10]. For simplicity, we assume each sensor has one reading initially and transmits 1000 packets during their transmission time without transmission interference. The k in top-k equals 15 which means the simulation tries to find the 15 nodes with the highest readings among all sensor nodes. B. Performance evaluation The performance of DCDC-WSNs, DCC-WSNs and AOWSNs in terms of query data accessibility, total energy consumption and query response time when replication-k equals 1 while k and N change are shown in Fig. 2. The impact of replication-k on their performances are shown in Fig. 3. 1) DCDC-WSNs VS. DCC-WSNs: From Fig. 2(a) and Fig. 2(b), we can clearly see that the data accessibility of DCDCWSNs is almost always higher than that of DCC-WSNs, especially when k is small. At many points, the data accessibility achieved by the former is nearly 200% of that achieved by the latter. Moreover, the stability of data accessibility of the

#Replication-k=1

#Replication-k=1

70 60 50 40 30 20 10

70 60 50 40 30 20

60 50 40 30 20 10

#N

#N

(b)

(c) #Replication-k=1 320000 300000 280000 260000 240000 220000 200000 180000 160000 140000 120000 100000 80000 60000 40000 20000 0 100 200 300 400 500 600 700 800 900 1000

(d)

(e)

(f)

#Replication-k=1

#Replication-k=1

1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 100 200 300 400 500 600 700 800 900 1000

2400

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

2160 1920 1680 1440 1200 960 720 480 240 0 100 200 300 400 500 600 700 800 900 1000

#N

#N

#N

(g)

(h)

(i)

Performance of DCDC-WSNs (a)(d)(g), DCC-WSNs (b)(e)(h) and AO-WSNs (c)(f)(i) (replication-k=1)

#N=1000, k=1

#N=1000, k=1

90 80

#Total Energy Consumption (mJ)

DCDC-WSNs DCC-WSNs AO-WSNs

70 60 50 40 30 20 10 0 5

6

7

8

9

10

#Replication-k

320000 300000 280000 260000 240000 220000 200000 180000 160000 140000 120000 100000 80000 60000 40000 20000 0

#N=1000, k=1 2400

DCDC-WSNs DCC-WSNs AO-WSNs

1920 1680 1440 1200 960 720 480 240 0

1

2

3

4

5

6

7

8

9

10

#Replication-k

(a)

(b) Fig. 3.

DCDC-WSNs DCC-WSNs AO-WSNs

2160 #Query Response Time (ms)

100

4

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

#Query Response Time (ms)

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

#N

#Replication-k=1

Fig. 2.

3

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

#Total Energy Consumption (mJ)

#Replication-k=1 200000 190000 180000 170000 160000 150000 140000 130000 120000 110000 100000 90000 80000 70000 60000 50000 40000 30000 20000 10000 0 100 200 300 400 500 600 700 800 900 1000

#Total Energy Consumption (mJ)

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

#N

1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 100 200 300 400 500 600 700 800 900 1000

2

0 100 200 300 400 500 600 700 800 900 1000

#N

#Query Response Time (ms)

#Query Response Time (ms)

70

0 100 200 300 400 500 600 700 800 900 1000

#N

1

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

80

(a) #Replication-k=1 200000 190000 180000 170000 160000 150000 140000 130000 120000 110000 100000 90000 80000 70000 60000 50000 40000 30000 20000 10000 0 100 200 300 400 500 600 700 800 900 1000

#Total Energy Consumption (mJ)

80

90

10

0 100 200 300 400 500 600 700 800 900 1000

#Query Data Accessibility (%)

100

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

90 #Query Data Accessibility (%)

#Query Data Accessibility (%)

80

#Replication-k=1

100

k=1 k=2 k=3 k=4 k=5 k=6 k=7 k=8 k=9 k=10

90

#Query Data Accessibility (%)

100

Impact of replication-k on performance (N =1000, k=1)

1

2

3

4

5

6

#Replication-k

(c)

7

8

9

10

former is also better than that of DCC-WSNs as decreasing rate of data accessibility of DCDC-WSNs is slower than that of DCC-WSNs, although the data accessibility of DCDC-WSNs and DCC-WSNs both decrease when N increases. Fig. 2(d) 2(g) and Fig. 2(e) 2(h) present the total energy consumption and query response time of the two networks. The total energy consumption of DCDC-WSNs is slightly higher than that of DCC-WSNs due to the a little more data replication energy consumption and top-k query energy consumption in DCDC-WSNs. The query response time of both the former and the latter are almost the same as data can only be obtained from the awake nodes after running the scheduling CKN algorithm and each sensor can transmit multiple data in one packet. 2) DCDC-WSNs VS. AO-WSNs: Fig. 2(a) and Fig. 2(c) show the query data accessibility of DCDC-WSNs and AOWSNs. The query data accessibility performance of AO-WSNs is very good as sensors are always awake in AO-WSNs. But the total energy consumption of AO-WSNs shown in Fig. 2(f) almost always outweighs that of DCDC-WSNs a lot shown in Fig. 2(d). In terms of same N value, when k increases around 10, the total energy consumption and query data accessibility of the former and the latter are almost the same, but when k decreases around 1, the total energy consumption of AOWSNs is about at least 160% (N =100, k=1) and at most 950% (N =1000, k=1) of that of DCDC-WSNs while the query data accessibility of AO-WSNs is about 120% and 500% of that of DCDC-WSNs for the same cases. Furthermore, the query response time of AO-WSNs shown in Fig. 2(i) is always higher than that of DCDC-WSNs shown in Fig. 2(g), which further shortens the performance of AO-WSNs, particularly in some real-time applications (e.g., forest fire monitoring). 3) Impact of replication-k on performance: The impact of replication-k on top-k query performance is shown in Fig. 3 in which N =1000 and k=1. We take this case as N =1000 and k=1 can be taken as an ideal application model in which there are a lot of sensor nodes but less awake neighboring nodes are needed to make the network connected. When replication-k increases, we can clearly see that there is outstanding improvement of query data accessibility for DCDC-WSNs (about total 250% improvement, final 70% accessibility) while keeping up almost the same obvious query cost advantage compared with AO-WSNs (about 90% less total energy consumption, 90% less query response time). From above analysis, we can get that DCDC-WSNs have good advantage about query data accessibility at almost the same query cost of DCC-WSNs. And compared with AOWSNs, the query cost advantage of DCDC-WSNs outperforms the query data accessibility advantage of AO-WSNs, especially during some response-time constrained applications and when replication-k increases which will improve query data accessibility. We conclude that the DCDC-WSNs can achieve the best tradeoff in terms of query data accessibility and query cost.

VIII. C ONCLUSION Implementing top-k query in duty-cycled WSNs (DCWSNs) is a very valuable and challenging research topic as the dynamically sleep and wake mechanism in DC-WSNs can greatly save energy consumption but there are data accessibility and network connectivity problems in DC-WSNs. In this paper, we advancingly apply data replication (DR) into DC-WSNs and strategically combine it with connectedk neighborhood (CKN) to solve the two problems. Further theoretical analysis and evaluations regarding top-k query in DC-WSNs with DR and CKN (DCDC-WSNs), DC-WSNs with only CKN (DCC-WSNs) and always-on WSNs (AOWSNs) are also presented. We conclude that implementing top-k query in DCDC-WSNs can achieve very high query data accessibility at the cost of low total energy consumption and top-k query response time, especially when more sensors data are replicated to which will increase query data accessibility. We hope our work can offer a valuable and enlightening insight about top-k query in DC-WSNs. ACKNOWLEDGMENT Lei Shu’s research in this paper was supported by Grant-inAid for Scientific Research (S) (21220002) of the Ministry of Education, Culture, Sports, Science and Technology, Japan. R EFERENCES [1] A. Silberstein, R. Braynard, C. Ellis, K. Munagala, and J. Yang, “A sampling-based approach to optimizing top-k queries in sensor networks,” in Proc. ICDE’06, Atlanta, GA, USA, 2006, pp. 68–80. [2] M. Wu, J. Xu, X. Tang, and W.-C. Lee, “Monitoring top-k query in wireless sensor networks,” in Proc. ICDE’06, Atlanta, GA, USA, 2006, pp. 143–145. [3] D. Zeinalipour-Yazti, Z. Vagena, D. Gunopulos, V. Kalogeraki, and V. Tsotras, “The threshold join algorithm for top-k queries in distributed sensor networks,” in Proc. DMSN’05, Trondheim, Norway, 2005, pp. 61–66. [4] Y. Cho, J. Son, and Y. D. Chung, “Pot: an efficient top-k monitoring method for spatially correlated sensor readings,” in Proc. DMSN’08, Auckland, New Zealand, 2008, pp. 8–13. [5] Q. Pan, M. Li, Y. Wu, and W. Shu, “Optimization of accurate top-k query in sensor networks with cached data,” in Proc. WCNC’07, Hong Kong, 2007, pp. 4233–4238. [6] S. Nath and P. B. Gibbons, “Communicating via fireflies: geographic routing on duty-cycled sensors,” in Proc. IPSN’07, Cambridge, Massachusetts, USA, 2007, pp. 440–449. [7] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishman, “Energyefficient communication protocol for wireless microsensor networks,” in Proc. HICSS’00, Hawaii, USA, 2000, pp. 3005–3014. [8] V. Raghunathan, C. Schurgers, S. Park, M. Srivastava, and B. Shaw, “Energy-aware wireless microsensor networks,” IEEE Signal Processing Magazine, vol. 19, no. 2, pp. 40–50, 2002. [9] C. Bettstetter, “On the minimum node degree and connectivity of a wireless multihop network,” in Proc. MobiHoc’02, Lausanne, Switzerland, 2002, pp. 80–91. [10] W. Liang, B. Chen, and J. Y. Xu, “Response time constrained top-k query evaluation in sensor networks,” in Proc. ICPADS’08, Melbourne, Victoria, Australia, 2008, pp. 575–582. [11] L. Shu, M. Hauswirth, H.-C. Chao, M. Chen, and Y. Zhang, “Nettopo: A framework of simulation and visualization for wireless sensor networks,” Ad Hoc Networks, Elservier, 2010.

Implementing Top-k Query in Duty-Cycled Wireless ... - CiteSeerX

s with unlimited energy supply which serves as a gateway between the sensor network and users ..... the center of the network region. The transmission radius of.

155KB Sizes 1 Downloads 197 Views

Recommend Documents

Implementing Top-k Query in Duty-Cycled Wireless ... - CiteSeerX
that are awake and black nodes represent the nodes that are asleep, node S ..... [8] V. Raghunathan, C. Schurgers, S. Park, M. Srivastava, and B. Shaw,.

Implementing Top-k Query in Duty-Cycled Wireless ...
∗Department of Computer Science, St. Francis Xavier University, Antigonish, Canada B2G 2W5 ... Abstract—Top-k query is a very useful and important query in wireless sensor networks ..... 100 200 300 400 500 600 700 800 900 1000.

Intelligent Jamming in Wireless Networks with ... - CiteSeerX
create a denial of service attack. ... Index Terms—Denial of Service, MAC protocol attacks, .... presented in [1] that showed that wireless networks using TCP.

Query By Committee Made Real - CiteSeerX
algorithms such as the k-nearest neighbors can be optimal. However ... labels. QBC works in an online fashion where each instance is considered only once to de- ... ing algorithm [11] and the perceptron based active learning algorithm [12].

A Home-Proxy Based Wireless Internet Framework in ... - CiteSeerX
cellular mobile telecommunications and Internet pene- tration. The natural ... based on the assumption of a single service, best effort, and a static networking ...

Packet Loss Behavior in a Wireless Broadcast Sensor ... - CiteSeerX
A good understanding of the loss behavior in a broadcast setting leads to the design ..... laptop computer; These receivers are very close to nearby objects which ...

Cross Layer Self-Healing Mechanisms in Wireless Networks - CiteSeerX
Abstract1. The use of wireless mobile ad-hoc networks (MANETs) has gained rapid ... neighboring sensor element B may be able to service the request.

Implementing Register Files for High-Performance ... - CiteSeerX
Abstract— 3D integration is a new technology that will greatly increase transistor density ... improvement with a simultaneous energy reduction of 58.5%, while a four-die version ..... Figure 3(d) shows an alternative implementation of a 2-die ...

Implementing a Hidden Markov Model Speech ... - CiteSeerX
School of Electronic and Electrical Engineering, University of Birmingham, Edgbaston, ... describe a system which uses an FPGA for the decoding and a PC for pre- and ... Current systems work best if they are allowed to adapt to a new speaker, the ...

Energy-Efficient Surveillance System Using Wireless ... - CiteSeerX
an application is to alert the military command and control unit in advance to .... to monitor events. ...... lack of appropriate tools for debugging a network of motes.

Epitomized Summarization of Wireless Capsule ... - CiteSeerX
Endoscopic Videos for Efficient Visualization. Xinqi Chu1 .... and quantitative evaluations on real data from the hospital. ... Also, important features with large lo-.

an algorithm for finding effective query expansions ... - CiteSeerX
analysis on word statistical information retrieval, and uses this data to discover high value query expansions. This process uses a medical thesaurus (UMLS) ...

Agglomerative Mean-Shift Clustering via Query Set ... - CiteSeerX
To find the clusters of a data set sampled from a certain unknown distribution is important in many machine learning and data mining applications. Probability.

Implementing Query Expansion for Improvement of Prior Art ... - IJRIT
1PG Student, Department of Computer Engineering, D. Y. Patil COE ... Query expansion has two major classes such as global methods and local methods.

Implementing Query Expansion for Improvement of Prior Art ... - IJRIT
IJRIT International Journal of Research in Information Technology, Volume 2, Issue ... 1PG Student, Department of Computer Engineering, D. Y. Patil COE ... Query expansion has two major classes such as global methods and local methods.

Enabling Schema-Free XQuery with meaningful query focus - CiteSeerX
Nov 29, 2004 - N, where Ai is comprised of nodes of type Ai (i ∈. [1, ..., m]), a .... Only the sec- ond and third MQF have title similar to “HTML” and are the.

ETRI-QM: Reward Oriented Query Model for Wireless ...
One unifying view is to treat them as distributed databases. The simplest .... Imagine a heterogeneous network consisting of many different types of sensors:.

an algorithm for finding effective query expansions ... - CiteSeerX
UMLS is the Metathesaurus, a medical domain specific ontology. A key constituent of the Metathesaurus is a concept, which serves as nexus of terms across the.

Mining Search Engine Query Logs via Suggestion Sampling - CiteSeerX
and notice is given that copying is by permission of the Very Large Data .... The performance of suggestion sampling and mining is measured ...... Estimating the efficiency of backtrack programs. Mathematics of Computation, 29(129):121–136,.

Agglomerative Mean-Shift Clustering via Query Set ... - CiteSeerX
learning and data mining applications. Probability ..... Figure 1: Illustration of iterative query set compression working mechanism on a 2D toy dataset. See text for the ..... MS and LSH-MS, lies in that it is free of parameter tuning, hence is more

Using Web Search Query Data to Monitor Dengue ... - CiteSeerX
May 31, 2011 - and analysis, decision to publish, and preparation of the manuscript. ... Such a tool would be most ... Tools have been developed to help.

Enabling Schema-Free XQuery with meaningful query focus - CiteSeerX
Nov 29, 2004 - default is to use keyword-based search and we are all too ...... able in most XQuery engines. In Sect. 4.2, we ..... Otherwise, we first check if the current stack top ...... Different relaxation and ranking techniques have been pro-.

Implementing an Interactive Visualization System on a ... - CiteSeerX
Department of Computer Science. University of Massachusetts-Lowell. One University Avenue. Lowell, MA 01854. Abstract. The use of a massively parallel ...

Implementing an Interactive Visualization System on a ... - CiteSeerX
formed by connecting line segments at their endpoints according to a set of geometric ... supercomputer version of the color icon and briefly outline the benefits and ..... The memory limitations of the Terasys forces the use of memory to be an impor