Master’s Project Proposal

Power Aware Wireless Sensor Networks Comparison of Clustered WSNs employing Distance-based Sleep Scheduling and WSNs employing the Multi-hop Backbone Formation protocol

Prithviraj Deshmane [email protected]

Department of Computer Science Rochester Institute of Technology July 2009

Chair

-

Reader

-

Observer

-

Professor James Minseok Kwon

1

Contents 1. Abstract…………………………………………………………………………………………………………………………………3 2. Introduction..………………………………………………………………………………………………………………………...3 3. Making WSNs Energy Efficient……………………………………………………………………………………………….5 3.1 Sleep Scheduling………………………………………………………………………………………………………5 3.2 Topology Control………………………………………………………………………………………………………5 3.3 Clustering…………………………………………………………………………………………………………………6 3.4 Distance-based Sleep Scheduling……………………………………………………………………………..6 3.5Backbone Formation…………………………………………………………………………………………………6 4. Project System Overview……………………………………………………………………………………………………….7 4.1 Scheme 1: Clustered Distance-based Sleep Scheduling Scheme……………………………….7 4.2 Scheme 2: Multi-hop Data Forwarding Scheme using Principles of SPAN…………………8 5. Project Deliverables……………………………………………………………………………………………………………..10 6. System Evaluation………………………………………………………………………………………………………………..10 7. Project Timeline……………………………………………………………………………………………………………………11 8. References……………………………………………………………………………………………………………………………12

2

1. Abstract We look at the issue of improving the network lifetime of wireless sensor networks by making the sensor nodes in the network stay ‘alive’ as long as possible. Wireless nodes in a sensor network are small devices that have very limited resources in the areas of memory, computation strength and electric power. When a node runs out of battery power it is said to have died, and on the death of even a single node, the entire wireless sensor network is said to have perished owing to the hole in coverage and functionality the dead node has left in the network. Hence, it is of primary concern that this issue be handled to strengthen wireless networks and to extend their lifetime. There has been research directed towards a variety of ways which help the wireless sensor network extend its lifetime. Topology Control and Sleep Scheduling have emerged as the frontrunners towards achieving this goal. We will pitch a 2-hop clustered distance based sleep scheduling approach against a multi-hop backbone formation technique and will compare their performances. The results would provide for a bigger picture with respect to the suitability of these approaches to tackling the issue of energy efficiency in wireless sensor networks.

2. Introduction Wireless technology has been recognized as one of the fastest growing technologies across the world. Wireless applications have ballooned and grown rapidly over the past few years to the extent that now it is virtually impossible to imagine life without using this technology in one form or the other. Mobile technology has become affordable and has found myriad applications in the global domain. Wireless networks are primarily classified as Ad-hoc networks and Sensor networks. Adhoc networks follow the 802.11 standards and find applications in networks of wireless computer networking, cellular phones, PDAs etc. Wireless Sensor Networks, on the other hand, deal with a distributed domain which involves sensing of phenomena such as temperature, pressure, motion, etc. using small wireless devices known as sensors. These sensors lack computational prowess as such and function primarily as wireless sensing entities. But what makes these tiny devices noteworthy is their ability to be deployed in remote locations (such as extremely remote terrain, underground, in volcanic craters or even underground) and the ability to function faithfully without human intervention.

3

The tiny size of these sensors facilitates their deployment in large numbers in the remote location in a random manner. It is, then, also easy to form a network of these deployed nodes. Such a network is called a Wireless Sensor Network, or simply a WSN. Yet, the small size and simple build of these sensor nodes is also a bane. The sensor nodes are not equipped with a constant power source and hence function on a small battery. It is only logical that they are also not equipped with computational components since these consume a lot of power and would drain out the small battery in a small period of time. A sensor node is primarily equipped with components which can sense its surroundings and those that can receive and transmit data, such as a radio. They are usually not equipped with other heavy energy consuming components. A sensor node is said to stay alive till the time its battery lasts. Each node senses its environment and transmits data to the Base Station (BS). The base station is the computational powerhouse of the network. All the data sensed in the network is sent to the BS to be made sense of. The BS is responsible for collecting all this data and processing it. The BS is always supplied with continuous power and does not exhaust its power supply like the nodes. Once the battery in a node is drained it is pronounced dead. When any node in the network dies, it leaves a hole in the network with respect to coverage and network functionality. When a sufficient number of nodes die, generally decided as a threshold, leaving behind many such holes, the entire network is said to have died. Since the entire functionality of the application utilizing the WSN depends on the network being alive, it directly depends on the rate at which energy is consumed in the network. To allow for longer running time for the application, it is imperative that the network be allowed to stay up for as long as possible. There have been suggestions of random redeployment of more sensors owing to their inexpensive nature, but the true solution lies in energy conservation, not just for a single node, but for the collective network. Much research has been devoted to making WSNs energy efficient and to facilitate optimal energy consumption. Notable amongst these are routing techniques, transmission scheduling, sleep scheduling, topology Control, and node distribution. Of these, I find that sleep scheduling techniques stand out, since the techniques directly facilitates energy conservation and is used in conjunction with almost all other techniques in practice. Let us look at sleep scheduling and other wireless networking techniques that are of significance to this project.

4

3. Making WSNs energy efficient 3.1 Sleep scheduling A wireless sensor node consumes some energy sensing its surroundings, but the energy it consumes while transmitting is significantly higher than while it does for any other purpose. Moreover, the energy consumed for transmission also depends not only on the size of the data, but also on the distance over which transmission takes place. The longer the distance, the higher is the energy consumed. Generally, a wireless node is either sensing its surroundings, transmitting/receiving data from other nodes, or is in an idle state waiting for an incoming signal. This signal could be a data signal or a control signal. But, since idling involves continuous sensing of the channel, the node would end up consuming the same amount of energy as it would for transmission. Also, when compared to the total time for which a node may be active, the time for which it would benefit from idling in this manner is significantly small. Hence, idling is an absolute waste of valuable energy resources for a wireless node. To avoid this unnecessary waste of energy, the node is allowed to sleep. When a node sleeps, it basically switches off all its components barring a timer which consumes negligible energy. This timer allows the node to wake up, switch on its components and continue working till it needs to sleep again. The topology of the network, which is the way in which a network is organized, determines which components in the network decide when and for how long a particular node is going to be allowed to sleep. We shall see what we mean by this statement in the following sections.

3.2 Topology Control As mentioned earlier, the topology of a network is the way in which the components of the wireless network are organized. Topology control is the way in which communication within the network is realized. It involves the logical (not physical) arrangement of the nodes in the wireless network in such a way that it facilitates energy conservation and easy data propagation. There are two ways in which topology can be controlled – backbone formation and clustering. Backbone formation is the technique in which a subset of the nodes is elected to function as the backbone of the network. This backbone is responsible for bringing about data transfer in the network, from a node to any other node in the network following a multi-hop model. Clustering, on the other 5

hand, is the organization of nodes into clusters. A Cluster Head (CH) is elected periodically for each cluster which is responsible for transmitting data from its cluster to the base station. We shall look at these in detail in the following sections.

3.3 Clustering Clustering is the process in which nodes in a WSN are grouped into clusters. Clustering is primarily a geographical process and which cluster a node belongs to depends mostly on the node’s geographical location. The main advantages of clustering are localization of control and data aggregation. Localization of control is desired for the optimization of the distributed approach and data aggregation facilitates limiting of data being transmitted over the wireless network and thus promoting energy conservation. Periodically, a cluster head (CH) is elected. The CH is usually a node near the center of the cluster, and one of the main criteria while CH election is the remnant energy of the node. The nodes in each cluster will transmit their data to the CH which performs data aggregation and then transmits the data directly to the base station. Hence this is primarily a 2-hop process. Periodic elections ensure that no single node exhausts itself performing CH duties.

3.4 Distance-based Sleep Scheduling This sleep scheduling technique has been recently developed and it follows a unique but very effective sleep scheduling cycle. As mentioned earlier, the energy consumed during communication depends on the distance over which communication takes place, other than the size of the data. The longer the distance, the higher is the energy consumed. Hence, this technique allows a cluster node to sleep in proportion to its distance from the cluster head. The farther a node is from the CH, the longer it is allowed to sleep. This allows for higher conservation of energy in the cluster, as well as a balanced energy consumption model across the cluster.

3.5 Backbone Formation Backbone formation protocols such as SPAN, are unique in their own way. They are mainly suited for ad-hoc networks, but the principle of multi-hop communication they present is of particular interest for the project. SPAN periodically elects backbone nodes which are known as Coordinators. This follows the Coordinator Eligibility Rule in which a node is elected as a coordinator if it needs to be awake to serve as a conduit between two of its neighbor nodes. If a node’s help is needed to bring about communication between two nodes, then it is elected as a coordinator. The backbone that is formed this way is responsible for transferring data from one node to any other node in the 6

network. The non-coordinator nodes continue sleeping till they satisfy the coordinator eligibility rule, which is when their services are required as coordinators. Similar to the eligibility rule, SPAN also has a Coordinator Withdrawal Rule in which a node can withdraw as a coordinator if its services are no longer required, or once its remaining energy renders it unable to serve as a coordinator. The entire purpose of the backbone is to facilitate communication in a multi-hop manner while the other nodes sleep and conserve energy. Periodically all sleeping nodes awake to test their eligibility to be coordinators. If not, they continue to sleep. If a node withdraws as a coordinator, it goes back to sleep and its associated non-coordinator nodes find other coordinator nodes to associate with. This protocol allows the nodes in the network to optimize their sleep cycles and conserve high amounts of energy.

4. Project System Overview Clustering combined with the Distance-based Sleep (DS) scheduling scheme is a 2-hop approach, wherein cluster nodes transmit data to the cluster head (CH) and the CH transfers data to the base station (BS). Multi-hop forwarding follows the logic that many small hops may be better than one or two long hops. Yet, clustering as an approach has enjoyed a lot of success with respect to energy conservation and being energy efficient. SPAN has emerged as one of the very few successful techniques for multi-hop forwarding. So the main objective of the project is to compare the two schemes and the respective sleep scheduling techniques employed by them.

4.1 Scheme 1 - Clustered Distance-based Sleep Scheduling Scheme The first protocol that I will be designing and simulating will be a clustered wireless sensor network employing the distance-based sleep scheduling technique. Employing the clustered approach will localize control to the cluster heads and their periodic election ensures that no single cluster head is overused. The DS scheme ensures that the nodes which are farther away from the CH conserve their energy by sleeping longer than the ones which are closer to the CH. This not only prevents from holes being created around the CH by exhausting the border nodes, but also acts as a factor that balances the overall energy consumption within the cluster, and consequently within the network. Once the Clustering and Cluster Head election algorithms are run on the network, the network will be divided into clusters. The cluster nodes will send all their data to their respective CHs. The CHs will perform data aggregation to reduce the traffic of data directed towards the BS. This further facilitates energy conservation. 7

Scheme 1 Legend Sleeping nodes Active nodes Cluster Head Base Station Communication between active cluster nodes and Cluster Head Communication between Cluster Heads and Base Station

The above diagram shows how the network for scheme 1 will be arranged. As can be seen, the network is divided into clusters. The cluster nodes that are awake transmit data to their respective cluster heads. The CHs in turn, collect all the data and transmit it to the base station. The nodes in a cluster will be allowed to sleep based on the DS scheme.

4.2 Scheme 2 - Multi-hop scheme using principles of SPAN I will be designing this scheme based on the principles of SPAN. The main objective is to form a backbone of nodes which will transfer data from a network node to the base station. This will resemble a star topology with the backbones converging at the BS. A node will be chosen as a backbone node by evaluating its utility factor which depends on its usefulness as a conduit node for communication as well as its remnant energy. 8

Periodically every non-coordinator node will test itself to satisfy the Coordinator Eligibility Rule, and will become a coordinator if its services as one are required. Similarly, every coordinator node will periodically test itself for Coordinator Withdrawal, and will withdraw as coordinator if it can no longer function as one. Whenever a new coordinator is elected, or a coordinator withdraws, all nodes update routing information according to the change. Regular updates will bring about minimum delay in communication.

Scheme 2 Legend Sleeping nodes Active (awake) nodes Coordinator nodes Base Station Communication between active nodes and their respective Coordinators Coordinators forming a multi-hop route to the Base Station

The above diagram shows the schemata for Scheme 2. As can be seen there are no clusters here. The coordinator nodes form a backbone within the network and take responsibility for transmitting data sensed by non cluster nodes to the base station. Non-cluster nodes will sleep till the time they need to assume the responsibility of a 9

coordinator. A coordinator that relinquishes its role will go to sleep. One can see that there are numerous small hops from a node to the BS, which may allow for energy conservation.

5. Project Deliverables One of the most important deliverable will be the project report. It will contain a detailed overview of the issue of energy efficiency in wireless sensor networks. The report will explain in detail the techniques of clustering, cluster head election, distancebased sleep scheduling and the principles of SPAN that I intend to use to bring about multi-hop communication in the WSN. I will provide algorithms and/or pseudo code and illustrations wherever needed for the systems I design. I will also enlist the tests I perform and the results I achieve from these tests. All source code, documentations and test cases developed will also be provided as part of the deliverables. Also, a detailed user manual will be provided to enable easy setup and use of the systems.

6. System Evaluation This section briefly explains how I will evaluate the two schemes. The two schemes will be designed primarily for static networks, i.e. immobile nodes, and the clusters that scheme 1 employs are static, i.e. a node belongs to only one cluster in its lifetime. The performance of the 2-hop clustering distance based sleep scheduling scheme will be matched against that of the multi-hop scheme for communication using the principles of SPAN. The primary metric that I will be evaluating for will be network lifetime. When a node in a wireless sensor node dies, it leaves behind a hole in the network with respect to coverage, routing and transmission. It also results in smaller sleep cycles for the rest of the nodes in the network. This defeats the entire purpose of the network, and in the strict sense a network is said to have died when the very first node in the network runs out of its battery power supply; thus the network lifetime is only as long as its fastest energy consuming node can stay alive. Practically, however, there are thresholds which determine the network lifetime. Depending on the application being designed, a designer may consider his network dead when 5%, 10%, 20% or even 40% of the nodes die. I will be examining how fast the nodes in my two schemes die, which would be done by examining the rates of energy consumption in the respective schemes. I will plot, 10

examine and compare how faithfully the two schemes are able to lengthen their respective network lifetimes. To bring about this evaluation I will take the help of simulations and emulations of the wireless networks. I will use the java emulator as a tool to design my systems. These java programs will take care of the functionality aspect of the networks. To visualize and to simulate the performance of the networks, I will employ Matlab. Performances of the systems will be compared when 5%, 10%, 20% and 40% of the nodes die. This will also give an overview of the rates of energy consumption in both the networks, and also, a measure of how well the schemes handle deaths of nodes. It will be interesting to note the changes in performance of the systems as nodes continue to die. Scalability is a very crucial virtue to be considered while designing and evaluating any system. Performance of the systems will also be evaluated under different network sizes, densities and traffic loads. This allows for determining how scalable the two approaches are.

7. Project Timelines Pre-Proposal

June 30, 2009

Proposal

August 11, 2009

Project Deliverables

November 25, 2009

11

8. References [1] Java SunSPOT Developer’s Guide http://sunspotworld.com/docs/Purple/spot-developers-guide.pdf [2] Xiaobing Wu, Guihai Chen and Sajal K. Das ‘On the Energy Hole Problem of NonUniform Node Distribution in Wireless Networks,’ IEEE 2006 http://ieeexplore.ieee.org/iel5/4053889/4053890/04053901.pdf?tp=&isnumber=&arnu mber=4053901 [3] Deng, Han, Heinzelman, Varshney, ‘Scheduling Sleeping Nodes in High Density Cluster-based Sensor Networks’, ACM 2005 http://portal.acm.org/citation.cfm?id=1160128 [4] Yuanyuan Zhou Muralidhar Medidi, ‘Sleep-based Topology Control for Wakeup Scheduling in Wireless Sensor Networks’, Secon 2007 http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4292842 [5] Benjie Chen, Kyle Jamieson, Hari Balakrishnan, Robert Morris, ‘Span: an energyefficient coordination algorithm for topology maintenance in ad hoc wireless networks’, ISSN:1022-0038, Year of Publication: 2002 http://portal.acm.org/citation.cfm?id=582461 [6] Paolo Santi, Topology control in wireless ad hoc and sensor networks, Year of Publication: 2005, ISSN:0360-0300 http://portal.acm.org/citation.cfm?id=1089736 [7] J. Deng, Y. Han, W. Heinzelman and P. Varshney, "Balanced-energy Sleep Scheduling in High Density Cluster-based Sensor Networks," Elsevier's Computer Communications Journal, Vol. 28, 2005, pp. 1631-1642. http://www.ece.rochester.edu/research/wcng/papers/journal/deng05_ccj.pdf [8] S. Soro and W. Heinzelman, "Cluster Head Election Techniques for Coverage Preservation in Wireless Sensor Networks," Elsevier Ad Hoc Networks Journal, Vol. 7, No. 5, July 2009, pp. 955-972. http://www.ece.rochester.edu/research/wcng/papers/journal/Soro0209.pdf

12