Energy Balancing LEACH for Wireless Sensor Networks

Transcription

1 Energy Balancing LEACH for Wireless Sensor Networks Department of Computer Science, Sookmyung Women s University, Korea choejn@sookmyung.ac.kr, axxujun@hotmail.com Abstract This paper suggests an algorithm which increases the energy efficiency based on LEACH (Low Energy Adaptive Clustering Hierarchy)[1] for sensor network routing. Due to the random cluster head election process, LEACH does not guarantee the optimization for the number and position of cluster heads. Also it results in inefficiency in energy consumption. This paper suggests improved cluster head election methods consisting of various phases that help to achieve more even distribution among the cluster heads. Balancing the load among the cluster heads, our proposed algorithm can increase the sensor network lifetime. From JAVA simulations, we show that our method outperforms the existing LEACH protocol, in terms of the number of alive nodes and the average energy consumption. 1. Introduction Keywords: LEACH, Residual Energy, Clustering, WSNs. Recent years, with the development and advancement of sensor technology, wireless sensor networks (WSNs) have been widely deployed for both civil and military applications [2]. Normally, sensors in such WSNs have resource constraints like limited energy, low storage capacity, and weak computing ability. Furthermore, due to the hazardous working environment, resources, especially the energy of sensors, may not be replaced or recharged. Therefore, the lifetime of the WSNs highly depends on the energy consumption of sensors. As a typical pervasive computing application, WSNs deployed a large number of sensor nodes in the monitoring region, collecting information sent to the target users by wireless means after simple treatment. But the sensor node usually carried one-time battery powered may lead to inefficient energy problem in wireless sensor networks. Therefore, the primary design goal of the WSNs protocols is to efficiently use the energy of sensor nodes and prolong the survival time of the entire network. LEACH [1] is the first clustering routing protocol in the WSNs. It presented the clusters were formed to fuse data before transmitting to the base station. Its clustering idea runs through the many clustering routing protocols after it was proposed. LEACH makes all the nodes in cluster rotation become head node to achieve the purpose that reduce node energy consume. LEACH is also the first one to raise the data aggregation. This paper proposes a sort of proposed strategy based on the LEACH protocol which has more important theoretical meaning and practical values in WSNs system. The major parts are as follows: Section two describes the traditional LEACH algorithms and analyzes in detail its deficiency. Section three expounds the proposed algorithm, and further improve the threshold T(n) of LEACH algorithms. In Section four, we present the details of our proposed algorithm by combining the nodes average energy consumption with distribution mathematical model. Section five provides the algorithm simulation experiment. In the last Section, the whole paper is summarized, and the development direction of the next step of the research is discussed and forecasted. International Journal of Intelligent Information Processing(IJIIP) Volume3. Number2. June doi: /IJIIP.vol3.issue2.8 56

2 2. LEACH protocol This section, we mainly describe of LEACH routing protocol, and summarized the various problems existed in the LEACH routing protocol Descriptions of LEACH protocol Low-Energy Adaptive Clustering Hierarchy (LEACH) protocol is the earliest clustering routing protocol which was proposed by Heinzelman in MIT [4]. It is one of the most popular hierarchical routing protocols for sensor networks. This algorithm selects cluster head nodes randomly, the rest nodes form clusters groups based on the received signal strength from cluster head nodes. In the beginning of the set-up stage, LEACH protocol forms clusters through distributed algorithm and each node decides whether or not to become cluster head node in self-adaptive mode. This is based on a random number between 0 and 1 that the node chooses, it will become a cluster head if the number is less than the threshold T(n) as follows: T(n) P 1 P*[r*mod (1/ P)] 0,, ng otherwise (a) Here P is the desired percentage of cluster heads among all the sensor nodes, r is the current number of rounds of election, G is the set of nodes which have not been elected in the past 1/P rounds of election. In the steady stage, each node member sends data to their cluster head node within their own schedules; the cluster head node sends the aggregated or compressed data to base station after receiving data from node members LEACH Protocol Problems In the above description of the LEACH algorithm, we know that LEACH algorithm still has the following problems. First, LEACH protocol randomly selects cluster-heads at each round. Therefore, the probability of cluster heads unevenly distributed would be very high. Second, LEACH protocol unsures the number of cluster nodes to be distributed equally in each cluster. In addition, it does not consider the distance between cluster head and BS. The uneven distribution of cluster heads would lead cluster heads which are far away from BS to spend more energy forwarding collected data as showed in Fig.1: Cluster A has about 4 cluster members while cluster (D, E) contains more than 30 cluster members. (D, E) would drain its power quickly because it needs to handle more burdens to process data collection and transmission than cluster A. The members far away from cluster head (D, E) also need to consume more energy to transmit data. 57

3 Fig1. Cluster topology not consider the distribution of cluster heads 3. Improvement of LEACH protocol Most of the papers just modified LEACH protocol on one hand, but it is still faulty and the performance is not the best. Due to all the drawbacks mentioned above, the modification of cluster generation procedure can only improve the efficiency of LEACH protocol. To achieve this goal, in this paper, we give a new algorithm, the algorithm of threshold T(n) was adjusted by taking all of the residual energy, communication mode and node distance into account in selecting satisfactory cluster head Assumption conditions Before we propose our algorithm, we must declare some assumption for modelling our network. 1) The near optimal number of cluster heads [5] has proved that system is the most energy efficient when there are between 3 and 5 clusters from the total 100 nodes in the network. That means the optimal percentage of cluster heads range from 3% to 5%. In this paper, we used from 3% to 5% as the near optimal percentage 2) Sensor nodes and base station (BS) are almost stationary and don t have any movement. 3) The sensor nodes are homogeneous and have limited energy [6] Algorithm description The proposed algorithm in this paper reduces the total energy consumption by minimizing the distance between cluster head and cluster members; also consider the residual energy to establish a better cluster head. The details mainly include the following works: 1) Determination of the cluster head number. N CH N total P (1) Here NCH is the number of cluster head, Ntotal is the total number of all nodes, and P is constant multiplier (3%~5%). 2) Improve the cluster head selection threshold T(n). The proposed algorithm still use LEACH protocol network model but the electing strategy of cluster-heads takes the energy and distance factor into account and the threshold defined as TED(n). Step1. At the first round that select cluster head used T(n) can be calculated from the equation (a). 58

4 Step2. Calculate energy as Eenergy: Nodes with higher residual energy have great probability to be selected as cluster head through importing a proportion factor on energy. Ecurrent E energy N CH Etotal (2) Here Ecurrent is the rest energy of the current node; Etotal is the initial energy when network starts. Step3. Calculate distance as DS: Nodes near cluster head have great probability to be selected as cluster head. D s D 1 node (3) Here Dnode is the between node and cluster head distance. Step4. The proposed T(n) as TED(n) can be calculate from equation (b). T ED ( n ) T 0 ( n ) E energy D s,, n G otherwise (b) By above proposed algorithm TED(n) can see, if node energy is enough and closer to cluster head, it increases the T(n) value so the probability of selecting cluster-heads rises. If energy becomes low and far from cluster head, it will lower the T(n) value so reduce the probability of select cluster heads. It can not only lead to a better equilibrium network load but also prolong the network life. 4. Mathematics and analysis In this section, we present the detail of our proposed algorithm. The main idea to assume a network environment assign suitable parameters then use the threshold of TED(n) to calculate results. And then the results are compared and analyzed to prove proposed algorithm can not only decrease the energy consumption but also prolong the network lifetime. Here we assume there is a cluster with 8 node members (A, B, C, D, E, F, G, H) shown as in Fig.2: Fig2. Cluster topology with 8 node members The initial energy of every node is set as 20, and each node member is allocated within a radius of 10. According to LEACH algorithm, three nodes D, F, H were unelected cluster head node in the latest (1 / p) round, and the distance between each other (A, B, C, D, E, F, G, H) is shown in Table1: 59

5 Table1. The distance matrix between (A, B, C, D, E, F, G, H) istance According to LEACH algorithm, the life cycle of the whole network potential to occur in different situations as following, and various cases of the cluster head s energy consumption is showed in Table2. Table2. Cluster head of energy consumption situation umber Selected State luster head Member H-energy consumption F is Cluster head (D,F) is Cluster head A,B,C,D,E,G,H 9 A,B,C 12 E,G,H 17 (D,H) is Cluster head (F,H) is Cluster head A,B,C,E,F G A,B,C,DE,G No have 5 9 (D,F,H ) is Cluster head A,B,C 2 E,G No have 0 From the above analysis, at the second cases, when D and F were elected as cluster head, the node energy consumption approached equilibrium. In order to show the performance of the proposed algorithm, using TED(n) were calculated and the result of TED(n) is shown in Fig.3. 60

6 Fig3. Proposed T (n) as TED(n) results According to the above calculated results, it is easy to see that, two nodes F and D most possibly become cluster head and H has the least chance to become cluster head. It is coincidence that node F and D is located in the distribution of the dense area, and node H is located in the distribution of sparse area. Obviously, ensure the cluster head node quantity of the dense regions more than sparse regions in the networks. We find that the proposed method provides an effective way to avoid the uneven energy consumption, thus, it can prolong the network life. 5. Algorithm simulation experiment This section firstly describes the simulation experimental platform of this thesis, and then detailed description of the experiment process of the proposed algorithm. Finally, simulation the results demonstrate the better performance of proposed algorithm Experimental platform Java platform is not specific to any one processor or operating system, but rather an execution engine (called a virtual machine) and a compiler with a set of libraries that are implemented for various hardware and operating systems so Java can be developed on any device. Java language is used in a wide variety of network programming with various characteristic, such as simple, object-oriented and portability, which brought a lot of convenience for network simulation. We use Eclipse tool in this topic, and simulation system development by Java language, then use this system to validate the theory that proposed in this paper. Table3 indicate the environment parameters we use in the simulation system. Table3. Environment parameters Simulation Environment Scenario1 Scenario1 Scenario1 ode number etwork size nitial energy 100~ *400(m2) 400*400~800*800(m2) 400*400(m2) 1000(J) 1000(J) 1000~3000(J) 5.2. Design of experiment In the process of generating nodes, the first thing is to construct node class function, since all the structure of the nodes is all the same; we only need to construct one node class function. This function 61

7 includes same key attributes such as node ID(uniquely), current energy of each node, node s relative position, and node distance from the cluster-member to the cluster head, etc. In order to simulate random distribution of nodes, using the side length of deployed nodes with sides area as parameters to determine the size of the sensor node deployment area, then to generate two random number x, y, the values range of x, y should be from 0 to deployed side length, make the x and y as node s coordinates to confirm the position of the nodes. The x and y is randomly generated, therefore node is randomly distributed. Use node number as parameters to determine the total number of random distribution node in networking. The program is coded by instantiation node and assigning values to related properties of each node to produce nodes. Cluster head rotated choosing the nodes, the cluster member nodes calculate the distance between oneself and produced cluster head in this round, as well as choose the nearest cluster head to join, thus, it forms cluster topology structure. In the cluster head selection process, the final energy consumption was determined by calculating the energy cost of the distances among each node. After the completion of the cluster head election in every round, we make a judgment of all the current node energy, if the current energy is less than zero, it means the node energy exhaust, otherwise, we would continue to do the experiment and collect the data of the number of rounds Experiment process Parameters sensors node number, nodes deployed with sides area, initial energy and node percentage is input respectively in the Starting interface (Fig.4). According to the input data, the number of nodes was specified, a coordinate were randomly assigned for each nodes. Next, use improved algorithm to select the cluster head as shown in Fig.5, obviously, interface with intuitive way of displaying the results of cluster head selection. Finally, record the produced different cluster head coordinates and round numbers in the proposed algorithm and LEACH algorithm in cluster head choice process, at the same time display the contrast result as a list, shown in Fig.6. The concrete operation process is as follows. Fig.4 is a screenshot that the data of deployed node parameter in simulation system was displayed, as long as the corresponding data is available, the nodes can be assigned. Fig4. Deployed node parameter Fig.5 is a screenshots that the cluster head is selected in with node number of 200, the point are cluster head node, the big circles have been elected as a cluster head node or energy shortage node, small circle are common node. 62

8 Fig5. Cluster head is selected in with node number of 200 Fig.6 is the life cycle results figure with specified parameters, we can review the different life cycle produced by LEACH algorithm and the improved algorithm. Fig6. Life cycle results 5.4. Experimental result 1) Randomly generated 100, 200, sensor nodes. Nodes are distributed in an area within 400 * 400, and initial energy of each node is 1000, then we get performance comparison results shown in Fig.7 through the statistics. The figure suggests that proposed algorithm compared the LEACH algorithm prolonging the network life cycle as much as 38%. Fig7. Nodes deployed with different number 2) Nodes are distributed in an area respectively within 400*400,600 *600,800*800. Sensor nodes number fixed for 100, and initial energy of each node is 1000, then through the statistical we get performance comparison results shown in Fig.8. The figure suggests that proposed algorithm compared the LEACH algorithm prolonging the network life cycle as much as 32%. 63

9 Fig8. Nodes deployed with different sides area 3) Each node of the initial energy were 1000,2000,3000, respectively. Sensor nodes number fixed for 100, Nodes are distributed in an area within 400 * 400, and then through the statistical we get performance comparison results shown in Fig.9. The figure suggests that proposed algorithm compared the LEACH algorithm prolonging the network life cycle as much as 43%. Fig9. Nodes deployed with different initial energy By the above statistics result we can find that proposed algorithm compared the LEACH algorithm with different parameters of the simulation experiment, prolonging the network life cycle as much as 37.7%. Obviously, the proposed algorithm used in the WSNs, is more effective in reducing and balancing the network energy consumption and improvement in WSNs lifetime, it has better performance than LEACH algorithm. 6. Conclusion To prolong the survival time of the network, clustering algorithm is a key approach to reduce the energy of dissipation in WSNs. In this paper, we propose a new clustering algorithm "Energy Balancing LEACH for Wireless Sensor Networks" when residual energy, communication mode and distance were utilized in selecting cluster heads. The ability of threshold T(n) that can increase the probability of choosing appropriate nodes, to become cluster head was significantly proposed. Therefore, the network load can be more evenly distributed and the network lifetime can be prolonged. In this paper theoretical analysis and calculation results demonstrated that, not only in the stability of networks but also in the survival time, the proposed algorithm had a better performance compared with LEACH. Even though the cluster head are replaced and selected again and again, there is still a situation that the energy of a cluster head node would be exhausted, resulting in the whole network lifetime affected. Future WSNs routing protocol in addition to efficient use of resources, should also be able to support its high degree of mobility, can be expanded, at the same time be able to have a certain QoS (Quality of Service) [12] the ability of routing and security fully ensure that all the sensor nodes to work together well in a variety of applications to play a full role. Moreover, this paper fails to consider the requirements of QoS, so the conclusion just contains a certain range of the comparison results. Therefore, a leach routing protocol is still expected to consider various demands as possible. 64

10 7.References [1] Sun Limin, Li Jianzhong, and Chen Yu. "Wireless sensor network" [M].BeiJing: Tsinghua University Press [2] Akkaya K, Younis M. "A survey of routing protocols in wireless sensor networks" [J]. Ad Hoc Networks, [3] Mucheol KIM,Sunhong KIM,Hyungjin BYUN,Sangyong HAN. "Optimized algorithm for balancing clusters in wireless sensor networks" [J]. Journal of Zhejiang University SCIENCE A (10): [4] Heinzelman WR, Chandrakasan A, Balakrishnan H. "energy-efficient communication protocol for wireless microsensor networks" [C]. In Proceedings of the Hawaii International Conference System Sciences, [5] Wendi B. Heinzelman, Anathan P. Chandraskan and Hari Blakrisshnan. "An Application- Specific Protocol Architecture for Wireless Microsensor Networks".IEEE Trans. on Wireless Communications. Volume. [6] Fan Xiangning; Song Yulin. "Improvement on LEACH Protocol of Wireless Sensor Network" International Conference on Sensor Technologies and Applicationsn. [7] Shah R C, Rabaey J. "Energy Aware Routing for Low Energy Ad hoc Sensor Networks", Orlando: IEEE Wireless Communications and Networking Conference (WCNC) [C].[S.I.]:[s.n.], [8] Kulik J, Heinzekman W.R, Balakrishnan H. "Negotiation based Protocols for Disseminating Information in Wireless Sensor Networks" [J].Wireless Networks, 2002, 8 (2-3) : [9] Chi Fu Huang, Yu Chee Tseng. "The Coverage Problem in a Wireless Sensor Net-work". [A], Proceedings of the 2nd ACM International Conference on Wireless Sensor Networks and Applications[C], Oakland: IEEE Press, 2003, [10] Intanagonwiwat C, Govindan R and Estrin D. "Directed diffusion: ascalable and robust communication paradigm for sensor networks". Proceedings of the 6th Annual ACM IIEEE International Conference on MobiCom'OO[C]. Boston: MA, [11] Linlin Wang, Jie Liu and Wei Wang. "An Improvement and Simulation of LEACH Protocol for Wireless Sensor Network" First International Conference on Pervasive Computing, Signal Processing and Applications. [12] Lindsey S, Raghavendra CS. PEGASIS. "Power-Efficient gathering in sensor information systems". In: Proc. of the IEEE Aerospace Conf. Montana: IEEE Aerospace and Electronic Systems Society, 2002, [13] Ye M, Li CF, Chen GH, Wu J. EECS. "An energy efficient clustering scheme in wireless sensor networks". In: Proc. of the IEEE Int l Performance Computing and Communications Conf. New York: IEEE Press, 2005, [14] Ren F.Y, Huang H.N, Lin C. "Wireless sensor networks" [J]. Journal of Software, 2003, 4(7): [15] Akyildiz LF, Su W, Sankarasubramaniam Y, et al. "A survey on sensor network"[j].ieee Communications Magazine, 2002, 40(8):