optimal power management in wireless sensor networks for enhanced ...

Volume 3, No. 4, April 2012 Journal of Global Research in Computer Science RESEARCH PAPER Available Online at www.jgrcs.info

OPTIMAL POWER MANAGEMENT IN WIRELESS SENSOR NETWORKS FOR ENHANCED LIFE TIME Durga Prasad Bavirisetti*1, Nagendra Prasad Mandru2 and Sibaram Khara3 School of Electronics Engineering, VIT University, Vellore, Tamilnadu-632014 {bdps19891,nagendra.mandru2}@gmail.com, [email protected] Abstract: wireless sensor networks find great applications in radiation levels control, noise pollution control, atmospheric pollution control, structural health monitoring and smart vehicle parking. All sensors present in wireless sensor network are battery operated devices which have limited battery power. After the deployment of sensor devices it is impossible to replace each and every battery present in the network. So energy conservation must be taken. In this paper we proposed an energy efficient dynamic power management technique which can reduce power consumed by each sensor node by shutting down some components of sensors according to our algorithm which yields better savings and enhanced life time.

INTRODUCTION Wireless sensor network made-up of a large no. of lowpower sensors. Now a days, wireless sensor networks find wide-range of applications such as radiation level control, ,battlefield , noise pollution control, machine failure diagnosis, atmospheric pollution control, biological detection, structural health monitoring, home security, smart vehicle parking, inventory tracking, etc. [2-3]. A wireless sensor network consists of small sensing devices, deployed in an interested region. Each device has processor and wireless communication devices like transceiver and sensors, which enable it to gather information from the environment and to generate and deliver report messages to the remote base station i.e.., remote user. After gathering information, the base station analyzes the report messages decides whether there is an unusual event or normal event occurrence in that particular deployed area. In order to gather accurate information a large no. of wireless sensors deployed in a high density and which improves system reliability. In WSN, the main source of energy is usually battery power. Sensors are often intended to be deployed in areas such as a battlefield or radiation plants; once deployment of sensor network it is impossible to recharge or replace the batteries of all the sensors. But, long system lifetime is needed for any monitoring application. The system lifetime, which is measured as the time until all nodes have been drained out of their energy (battery power).Important challenge to the design of a large wireless sensor network is energy efficient problem. So the proposed design should extend the system lifetime without sacrificing system reliability. A sensor network consists of one or more sensors, processing circuits, memory, and a wireless transceiver. The Main goal of WSN is to prolong the life of the network and preventing connectivity degradation through aggressive energy management as most of these sensor network devices have limited battery life and it is impossible to replace batteries of up to tens of thousands of sensors in most of the applications. Many researchers are doing their research to reduce power consumption in various aspects of hardware design, data processing circuits design, network protocols and © JGRCS 2010, All Rights Reserved

operating systems [3, 4, 5]. Once the system has been designed, extra energy savings can be done by using dynamic power management (DPM), which shuts down the sensor node when there is no work [7]. The basic idea is to shut down sensor devices when not needed and wake them up when necessary which yields good energy savings. So we need to carefully implement Dynamic Power Management (DPM) to get the maximum life of sensor node. If we blindly turn the radio ON and OFF during each idle slot over a period of time, we might end up expending more energy than if the radio had been left on. So, smarter schemes needed to turn the nodes on/off. Means, operation in a power-saving mode is energy-efficient only if the time spent in that mode is greater than a certain threshold value of time. Depending on the number of the states of the microprocessor, memory, A/D converter, and transceiver There are number of such useful modes of operation for a wireless sensor node,. It is also important to consider the state of computation when system turns components on/off to reduce energy. The state of the computation in each period of time represents the state of the application and its restrictions in an instant of time, which can have a direct influence on decisions taken by a power manager. LITERATURE SURVEY The lifetime of a sensor network depends highly on the power consumed at each sensor node. A more efficient power management will provide a longer network lifetime. Several methodologies have been proposed at many levels. Mainly at hardware and system levels, to design energy efficient communication process, sensor node operating system and sensor node circuits. In addition, a variety of DPM techniques have been Proposed to reduce the power consumption in sensor nodes and in general battery-powered embedded systems [2, 9, 10, 11, 12] by selectively shutting down the components. Much work has been done exploiting sleep state and active power management [2,13], Dynamic Voltage Scaling (DVS) [2,11,14] and Dynamic Voltage and Frequency scaling [10], sentry-based power management [15], an application-driven approach [16], software and operating system power management and battery state awareness power management. Depending on the approach 73

Durga Prasad Bavirisetti et al, Journal of Global Research in Computer Science, 3 (4), April 2012, 73-78

that is used, DPM policies are classified as predictive or stochastic policies. A widely used predictive technique consists in turning OFF the system components if an idle time greater than or equal to a timeout threshold value T is detected. This approach is based on the assumption that if the idle time is greater than the threshold T, the system is likely to remain idle for a time period long enough to save energy. A more accurate method is proposed in [20], where the upcoming idle time is predicted by using an exponentialaverage approach. If the predicted idle time is sufficiently long, the system component is switched OFF at once. In [2], the authors proposed an OS-directed power management technique to improve the energy efficiency of sensor nodes. The node would update the probability of even generation. It is event-based power management policy for a single node, but not an effective policy for the whole system. First, it may cause event-missed situation due to the operation system isolates the node in the deepest sleep state and it is awakened until a specific sleeping interval goes by. Second, the authors only considered that an event occur can wake up the sleeping node. The third, it is inefficient if the density of nodes is over dense in some regions [17]. In addition, predictive techniques have a few limitations: they cannot provide an accurate tradeoff between energy saving and performance degradation, and they do not deal with a generic system architecture where service requests can be queued. A stochastic policy has been proposed in [21] to overcome these limitations. The considered system is composed of a service provider, a service requester, a power manager, and a request queue. The service provider and requester are represented as Markov processes, and the power manager determines the device state of operation by issuing commands to the service provider. In this case, the optimal policy strictly depends on how the system is modeled and on the abstractions that have been made. Moreover, the amount of energy consume by the power manager remains to be accounted. MOTIVATION OF THE WORK The main goal of power management is to prolong the lifetime of WSN and preventing connectivity degradation through aggressive power management as the most of the devices have limited battery life. Once the sensor network has been deployed It is impossible to replace each and every battery present in WSN if the battery is emptied. So we should follow power conservation technique in order to save the energy. If blindly turn the radio OFF during each idle slot, over a period of time to save the energy, we might end up expending more energy than if radio has been left on continuously .So, Power saving is energy-efficient only if the energy conserved is greater than energy expended in order to transit to sleep state. PROBLEM SPECIFICATION The most important constraint in wireless sensor network is the energy efficiency problem. Power conservation and power management must be taken into account at all levels of the sensor network system hierarchy. For optimal power management in wireless sensor networks an effective algorithm should be proposed. Proposed

© JGRCS 2010, All Rights Reserved

algorithm should yield better results in terms of network lifetime and reduced energy consumption METHODOLOGY In Dynamic power management (DPM) technique, the basic idea is to shut down devices when not needed and wake them up when necessary which yields good savings. Once the system has been designed additional energy savings can be done using Dynamic Power Management. An energy-efficient algorithm Dynamic power management with modified sleep states has been proposed for enhanced lifetime in this paper. DPM technique can be discussed through two policies. 1. Power aware sensor node model 2. Sleep state transition policy A. Power aware sensor node model Every sensor node consists different components like processor to process the incoming data, memory in order store the data, sensor component to sense the data from environment and radio in order to transmit or receive or for both transmission and reception purpose. Depending on the different states of the components there exist different states as shown in table (1). Processor has 3 states, active, idle and sleep states. Memory has 2 states, only active and sleeps states. Sensor component can be turned into either on or off states. And finally radio can set into either transmission or reception mode. TABLE 1 Different state of components

State S0 S1 S2 S3 S4

Processor Active Idle Sleep Sleep Sleep

Memory Active Sleep Sleep Sleep Sleep

Sensor On On On On Off

Radio TX/RX RX RX Off Off

If all the components are in active state i.e., processor is active state, memory is in active state, sensor is in on state and radio is turned on either for transmission or for reception. This state is called as active state which is represented as S0. In order to reduce the power consumption some of the components should be turned off. If some of components turned off then sensor node consumes less power. This state is also termed as sleep state so. Depending on the power levels there exist different sleep states. If processor is idle, memory is in sleep state, sensor component is turned on and also receiver set into reception mode. It is called as first sleep state which is represented as S1. In second sleep state S2, both processor and memory are in sleep state, sensor component is turned on and radio is set into reception mode. Sleep state S3 is defined as except sensor component all other components turned off. Final state which is nothing but sleep state S4 in which all sensor node components kept into off state. So this state is called as deepest sleep state. Deepest sleep state take very less power compare with all other sleep states because all components are in off state in this. B. Sleep state transition policy

74


In previous section we defined different sleep states according to power consumption. But in this section we are going to discuss how and when these sleep states will be selected to conserve the energy. And also we will see depending on what threshold value transitioning from one state to another sleep state happens. Figure.1 shows the state transition delay and power. Let us assume an event occurred at time t1 and another event occurred at time t2. In between there exist an idle time which is represented as ti as shown in figure (1). Then t2 can be given as t2 = t1 + ti. At time t1 node decide to transit to sleep state Si from active state S0.For different values of i, i.e., i= 1, 2, 3 it represents first, second and third sleep states respectively. Last sleep state in general taken as Sj. As we discussed in power sensor node model these sleep states S0, Si and Sj have different power levels depending on the no. of components that are in ON condition. We also assume that the process of transitioning from one state to another state is not direct, it is gradual. if a node want to transit from active state S0 to sleep state S3, it can’t directly go to sleep state S3 from active state. First it should go to sleep state S1 from S0, then to S2 and then finally to S3. In order to transit from active state S0 to sleep state Si , if it takes transit time which is also called as forward transit time or forward delay. In the same way in order to transit from S0 from Sj it takes a forward transit time of . To save energy the node must be in idle state for threshold value of time. Depending on the threshold value of time and next event occurrence it should select the sleep state. If any event occurred, node should get back to active state from sleep state; to get back to active state S0 from sleep state Si also it takes some transit time which is termed as reverse transit time or reverse delay. In the same way is the reverse transit time to get to back to active state S0 from sleep state Sj.

consumption needed for awakening the sensor node back to active state S0 and should be able to foresee how long remains idle in a particular state. For reduced power consumption complexity, we assumed that the active state is directly transited to sleep state. The energy saving because of state transition to a sleep state and back is given as The transition from one state to another is useful when energy saved must be greater than the energy consumed. It is clear that the saved energy compensates the expended transition energy only when the node should be in sleep state for that amount of idle time. So the energy saving must be greater than the energy expended by the sensor node for certain threshold value of time.

Tth ,i

1 2

0,i

P0 P0

Pi Pi

i ,0

Deepest sleeping period according to battery status parameter µ can be given as Where

denotes the standard working voltage. denotes the present working voltage of the

battery. The standard working voltage of Li-battery is 3.6V and present working voltage is in between 2.8V and 4.2V. So we can define the deepest sleep state period T using for any µ as a time counter for state . C. Proposed algorithm As shown in figure (2) we proposed an algorithm which can conserve the energy efficiently using dynamic power management technique. Calculating deepest sleep start time TD : If t1 is current time and t2 is next event start time then idle time ti can be given as Deepest sleep start time can be calculated as

Figure 1. State transition delay and power

The conditions for node sleep states for any i < j given as

Every transition from sleep state Si to Sj has cost in terms of delay overhead i , j but for practical point 0, j 0 ,i of view the power consumption associated with transitions from state Si to Sj (i < j) is much lower than the cost associated with reverse transition. So, proposed Dynamic power management technique must see the extra energy © JGRCS 2010, All Rights Reserved

Where, T is deepest sleeping period according to battery status (B). Calculating different sleep state start times: First sleep state start time is nothing but current time i.e.., TS1=t1 Fourth sleep state start time is nothing but deepest sleep state start time. i.e.., TS4 = TD Time gap between deepest sleep state start time and first sleep state start time given as TG = TS4 - TS1 Time interval among each and every state can be calculated as Ti =TG/3 So, the second sleep state start time is nothing but addition of first sleep state start time and time interval. TS2 = TS1 + Ti And third sleep state start time is also given in the same way as the addition of second sleep state start time and time interval among each state. 75


TS3 = TS2 + Ti Selecting different states: As discussed in power aware sensor node model, depending on the status of components and power levels different states defined. So, these states must be select according to current time. If current time is in between first sleep state start time and second sleep state start times then select current state as first sleep state S1 as shown in figure(2). No, if current time is in between second sleep state start time and third sleep state start times then select current state as second sleep state S2. And if current time is fall under the range between third sleep state start time and fourth sleep state start times then select current state third sleep state S3. Otherwise select fourth state i.e.., deepest sleep state as current state.

operation. In this, Base station send hello message to each neighbor and update Id information for each neighbor and then verify which node has least Id. Set that node as cluster head. Cluster head send cluster announcement message to all of its neighbors. After receiving reply to cluster announcement message, cluster head will check the status of each and every neighbor node whether it is cluster member or cluster head.

Figure 3 comparison of average energy of AODV and EEDPM techniques

Figure 4 comparison of network life time of both AODV and EEDPM

Figure 2. Optimal power management algorithm

PERFORMANCE We used NS-2 tool for our simulations. Clustering based protocol used, in which the base station is positioned at the middle of the field. Didn’t considered multi hope © JGRCS 2010, All Rights Reserved

If the node is not a cluster head add those nodes as members to that particular node. Those cluster members store sender Id as cluster head Id. If a node comes under the vision of two or more cluster heads then it will add as a member to the cluster head which is having greatest sender Id. After clustering formation, at the time of receiving packets at the time of receiving packets at each and every sensor node we applied our proposed algorithm of energy efficient DPM which can save energy very efficiently. Assume all nodes are fixed and having initial energy of 3(joules). In this, we considered AODV protocol as routing protocol. Without applying our DPM technique to this routing protocol we calculated average energy consumptions for different number of nodes. And also calculate lifetime of the wireless sensor network. For the same AODV routing protocol we 76


applied our DPM technique and calculated average energy consumption and also calculated network lifetime. Figure (3) shows the comparison of average energy consumption of both AODV and EEDPM i.e.., energy efficient dynamic power management technique after applying to the same AODV protocol. From figure (3) what we can observe is, by applying EEDPM technique to the sensor network then average energy consumption will reduce. Figure (4) shows the comparison of lifetime of the sensor network with and without applying energy efficient dynamic power management. Because of all components in ON state AODV consumes much energy. Because of applying EEDPM technique some of the components are in OFF state consumes less energy. So, the EEDPM have more lifetimes compare with AODV technique. As the no. of nodes increases average energy consumption increase and lifetime of the sensor network will reduce. . CONCLUSION In this paper we proposed an energy efficient dynamic power management and applied it to AODV routing protocol. This algorithm yields reduced power consumption and satisfactory lifetime. In this algorithm we didn’t consider missed event when sensor node is in deepest sleep state. This problem is subjected to our future research. REFERENCES Chuan Lin, Yan-Xiang He, Naixue Xiong, ―An Energy Efficient Dynamic power management in wireless sensor networks‖, Proceedings of the fifth international symposium on parallel and distributed computing (ISPDC’06), pp. 148154, 2006. [2] A.Sinha, A.Chandrakasan, ―Dynamic Power Management in Wireless Sensor Networks,‖ IEEE Design and Test of Computers, Vol. 18, Issue 2, pp. 62-74, March-April, 2001. [3] B. H. Calhoun, D. C. Daly, N. Verma, D. Finchelstein, D. D. Wentzloff, A. Wang, S.-H. Cho, and A. P. Chandrakasan, "Design Considerations for Ultra-low Energy Wireless Microsensor Nodes," IEEE Transactions on Computers, June, 2005. [4] K. Sohrabi, J. Gao, V. Ailawadhi, G.J. Pottie, ―Protocols for Self-organization of a Wireless Sensor Network,‖ IEEE Personal Communications, Vol.7, Issue 5, pp. 16 – 27, October, 2000. [5] V. Raghunathan, C. Schurgers, Sung Park, M.B. Srivastava, ―Energy-Aware Wireless Microsensor Networks,‖ IEEE Signal Processing Magazine, Vol. 19, Issue 2, pp. 40 – 50, March, 2002. [6] J. Hill,R. Szewczyk, A. Woo, S. Hollar, D.E. Culler, and K. Pister, ―System Architecture Directions for Networked Sensors,‖ Architectural Support for Programming Languages and Operating Systems, pp. 93-104, 2000. [7] L. Benini, G. D. Micheli, Dynamic Power Management: Design Techniques and CAD Tools, Kluwer Academic Pub_NY, NY, 1997. [7] I.F. Akyildiz, Weilian Su, Y. Sankarasubramaniam, [8] Cayirci, ―A Survey on Sensor Networks,‖ IEEE Communications Magazine, Vol.40, Issue 8, pp. 102 –114, August, 2002. [9] E.Y.Chung, L.Benini,and G.D.Micheli, ―Dynamic Power Management Using Adaptive Learning Tree,‖ International Conference on Computer-Aided Design (ICCAD), pp. 274 – 279, 7-11 November, 1999. [10] A.L.A.P. Zuquim, L.F.M. Vieira, M.A. Vieira, A.B. Vieira, H.S. Carvalho, J.A. Nacif, C.N. Jr. Coelho, D.C. Jr. da Silva, A.O. Fernandes, A.A.F. Loureiro, ―Efficient Power Management in Real-time Embedded Systems,‖ IEEE [1]


[11] [12]

[13] [14]

[15]

[16]

[17] [18] [19]

[20]

[21]

[22] [23]

[24]

[25]

[26]

[27]

International Conference on Emerging Technologies and Factory Automation-RTFA’03, vol.1, pp. 496 – 505, 16-19 September, 2003. IBM and MontaVista Software, ―Dynamic Power Management for Embedded System,‖ Ver.1.1, 19-November, 2002. C.F Chiasserini, R.R. Rao, ―Improving Energy Saving in Wireless Systems by Using Dynamic Power Management‖, IEEE Transactions on Wireless Communications, Vol.2, Issue 5, pp. 1090-1100, September, 2003. B. Brock, K. Rajamani, ―Dynamic Power Management for Embedded System,‖ IEEE International System-On-Chip (SOC) Conference, pp. 416 – 419, 17-20 September, 2003. B.Calhoun, A. P. Chandrakasan, "Standby Power Reduction Using Dynamic Voltage Scaling and Canary Flip-Flop Structures," IEEE Journal of Solid-State Circuits, Vol. 39, No. 9, September, 2004. J. Hui, Z. Ren, B.H. Krogh, ―Sentry-based Power Management in Wireless Sensor Networks,‖ Second International Workshop on Information Processing in Sensor Networks, pp. 458-472, April, 2003. Rodrigo M. Passos, Claudionor J. N. Coelho Jr., Antonio A. F. Loureiro, Raquel A. F. Mini, ―Dynamic Power Management in Wireless Sensor Networks: An ApplicationDriven Approach,‖ Second Annual Conference on Wireless On-demand Network Systems and Services (WONS'05), pp. 109-118, January, 2005. Ren C. Luo, Liang Chao Tu, Ogst Chen, ―An Efficient Dynamic Power Management Policy on Sensor Network,‖ Proceeding of the 19th International Conference on Advanced Information Networking and Applications (AINA’05), vol. 2, pp. 341 – 344, 28-30 March, 2005. Hewlett-Packard, Intel, Microsoft, Phoenix, and Toshiba, Advanced Configuration & Power Interface (ACPI): an open industry Specification - Revision 3.0a. Available at: http://www.acpi.info/. T.A. Henzinger, ―The Theory of Hybrid Automata,‖ Eleventh Annual IEEE Symposium on Logic in Computer Science (LICS), pp. 278-292, July, 1996. C.-H. Hwang, A. C.-H. Wu, ―A Predictive System Shutdown Method for Energy Saving of Event-driven Computation,‖ IEEE/ACM International Conference on Computer-Aided Design, pp. 28–32, San Jose, CA, Nov. 1997. L. Benini, A. Bogliolo, G. A. Paleologo, and G. De Micheli, ―Policy Optimization for Dynamic Power Management,‖ IEEE Transactions on Computer-Aided Design, vol. 18, pp. 813–833, June, 1999. X. Li, P. Wan, and O. Frieder, ―Coverage in Wireless Ad-hoc Sensor Networks,‖ IEEE Transactions on Computers, Vol. 52, pp. 753- 763, June, 2003. H. Zhang and J. C. Hou, ―Maintaining Sensing Coverage and Connectivity in Large Sensor Networks,‖ In NSF International Workshop on Theoretical and Algorithmic Aspects of Sensor, Ad Hoc Wireless, and Peer-to-Peer Networks, 2004. Ye, G. Zhong, S. Lu, and L. Zhang. ―Peas: A Robust Energy Conserving Protocol for Long-lived Sensor Networks,‖ Proceeding of the 23rd International Conference on Distributed Computing Systems (ICDCS), 2003. D. Estrin, R. Govindan, J. S. Heidemann, and S. Kumar, ―Next century challenges: Scalable Coordination in Sensor Networks,‖ In Proceeding of ACM MobiCom’99, Washington, August, 1999. S. Slijepcevic, M. Potkonjak, ―Power efficient Organization of Wireless Sensor Networks,‖ Proceedings of IEEE International Conference on Communications, pp. 472—476, Finland, June 2001. A.E. Cerpa, D. Estrin, ―ASCENT: Adaptive Self-Configuring sEnsor Networks Topologies,‖ IEEE Transactions on Mobile Computing 3 (3), pp. 272-285, 2004.

77


[28] Ye, G. Zhong, S. Lu, and L. Zhang, ―Energy Efficient Robust Sensing Coverage in Large Sensor Networks,‖ Technical report, UCLA, 2002. [29] D.Tian, N. D. Georganas, ―A Coverage-preserving Node Scheduling Scheme for Large Wireless Sensor Networks,‖ In First ACM International Workshop on Wireless Sensor Networks and Applications, Georgia, GA, 2002.


78