Energy Efficient Load-Balanced Clustering Algorithm ...

Available online at www.sciencedirect.com

Procedia Technology 6 (2012) 771 – 777

2nd International Conference on Communication, Computing & Security [ICCCS-2012]

Energy Efficient Load-Balanced Clustering Algorithm for Wireless Sensor Networks Pratyay Kuilab, Prasanta K. Janaa,* a,b

Department of Computer Science & Engineering, Indian School of Mines, Dhanbad-826004, India

Abstract Clustering is an efficient technique to improve scalability and life time of a wireless sensor network. In this paper, we present an Energy Efficient Load-Balanced Clustering (EELBC) Algorithm that addresses energy efficiency as well as load balancing. EELBC is a min-heap based clustering algorithm. A min-heap is build using cluster heads (CHs) on the number of sensor nodes allotted to the CHs. We show that the algorithm runs in O (n log m) time for n sensor nodes and m CHs. The experimental results show the efficiency of the proposed algorithm in terms of load balancing, energy efficiency, execution time and also the number of sensor nodes die during the network period. © 2012 The Authors. Published by Elsevier Ltd. © 2012 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Department of Computer Selection and/or peer-review under responsibility of the Department of Computer Science & Engineering, National Institute Science & Engineering, of Technology RourkelaNational Institute of Technology Rourkela Keywords: Wireless sensor networks, clustering, load balancing, energy efficiency, network lifetime;

1. Introduction The radical advances in the fields of MEMS (Micro Electro-Mechanical Systems) technology development of high speed broadband wireless technologies and low-power radio frequency (RF) design have led to the birth of Wireless Sensor Networks (WSNs). WSNs have attracted the enormous attention towards applications in diverse areas such as disaster warning systems, crop/environment monitoring, health care, safety and strategic areas such as defence reconnaissance, surveillance, intruder detection etc (Akyildiz I.F. et al., 2002). A WSN is composed of a large number of tiny sensor nodes, which are randomly or manually deployed in a given

* Prasanta K. Jana. Tel.: +91-9431126447; fax: +91-032-62296563. E-mail address: [email protected].

2212-0173 © 2012 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Department of Computer Science & Engineering, National Institute of Technology Rourkela doi:10.1016/j.protcy.2012.10.093

772

Pratyay Kuila and Prasanta K. Jana / Procedia Technology 6 (2012) 771 – 777

coverage area. The sensor nodes consist of sensing, data processing, and communicating components along with a power unit. Sensor nodes may have a location finding system like Global Positioning System (GPS) and mobilizer to move within the coverage area. In WSN, all the sensor nodes collect local information, process them and send it to a remote base station (called sink). The sink is connected to the Internet for the public notice of the phenomena. One of the most important constraints on sensor nodes is the requirement of low power consumption. Sensor nodes carry limited and generally irreplaceable power sources. So, reducing energy consumption for maximizing network lifetime is thus considered as the most critical challenge in WSN. Many research articles have been addressed on this issue (Kyung Tae Kim et al., 2003; Emanuele Lattanzi et al., 2007). However, design of energy efficient clustering algorithms is the most promising area in this regard. In WSN, the sensor nodes are divided into several groups, called cluster. Every cluster would have a leader, often referred to as the cluster-head (CH). In a few WSN scenarios, some high-energy nodes called are deployed in the network. These gateways group sensors to form distinct clusters in the system and act as a CH. The CHs manage the network in the cluster, perform data fusion and send the processed data to the sink through other CHs or sensor nodes. Each sensor node only belongs to one and only one cluster and communicates with its CH. The functionality of a cluster based WSN with single-hop communication inside the clusters is shown in Fig. 1. The advantages of a cluster based WSN are as follows. It reduces energy consumption significantly; conserves communication bandwidth and improves the overall scalability of the network (Ameer Ahamed Abbasi and Mohamad Younis, 2007). However, improper assignment of the sensor nodes for the formation of clusters can make some CHs overloaded with high number of sensor nodes. Such overload may increase latency in communication and degrade the overall performance of the WSN. Therefore, load balancing is also a crucial issue that must be taken care while clustering sensor nodes. In this paper, we propose a clustering algorithm that addresses both the issues, i.e., energy efficiency and load balancing. The algorithm is based on min-heap on the number of sensor nodes allotted to the CHs. We show that the algorithm outperforms the similar works reported by Chor Ping Low et al. (2008) and Gaurav Gupta et al. (2003) in terms of execution time, load balancing and energy consumption.

Internet

Sensor Node

Cluster head

Base Station.

Fig.1. A Cluster based wireless sensor network The paper is organized as follows. The related work is presented in Section 2. The energy consumption model is described in Section 3. The proposed algorithm is presented in Section 4. Experimental results are given in Section 5 followed by the conclusion in Section 6.


2. Related Works A number of clustering algorithms for WSN have been addressed in Ameer Ahamed Abbasi et al. (2007), Olutayo Boyinbode et al. (2010), Congfeng Jiang et al. (2009) etc. In 2002, Heinzelman W. B. et al. have developed LEACH. LEACH is a popular clustering technique that forms clusters by using a distributed algorithm. However, the main disadvantage of this approach is that a node with very low energy may be selected as a CH which may die quickly. Therefore, a large number of algorithms have been developed to improve LEACH such as PEGASIS (Lindsey S. et al., 2003), HEED (Younis O. et al., 2004), TEEN (Manjeshwar A. et al., 2002), APTEEN (Manjeshwar A. and Agrawal D. P., 2002), EEPSC (Amir Sepasi Zahmati et al., 2007) etc. Compared to LEACH, PEGASIS improves network lifetime, but it requires dynamic topology adjustment and the data delay is significantly high and it is unsuitable for large-sized networks. The re of the neighbor nodes or node degree. Bencan Gong et al. (2008) have introduced MRPUC, where the authors design an unequal clustering and multihop routing to extend network lifetime. But the inter-cluster multi-hop communication may cause an additional overhead for a large-sized network. To form cluster, Chor Ping Low et al. (2008) have considered a bipartite graph of the sensor nodes and the gateways to find out the maximum matching for assigning a sensor node to a CH. The algorithm has the time complexity of O(mn2). For a large scale WSN, execution time is very high and also building a BFS tree for individual sensor node takes a substantial amount of memory space. In 2011, Pratyay K. and Prasanta K. J. proposed an algorithm of execution time O (n log n), which is an improvement over Chor Ping Low et al. (2008). However, no energy consumption issue has been addressed in this algorithm. Gaurav Gupta et al. (2003) have proposed a load balanced clustering algorithm, where distance is not considered. So, overall energy consumption of the system is comparatively high. Other clustering algorithms have been developed which can be seen in Zhixin Liu et al. (2011) or Xiang Min et al. (2010) or in Wei Li. (2009). But all such protocols do not consider both the load balancing and the energy consumption issues combinedly. The algorithm proposed in this article takes care of both these issues with the following advantages over the algorithms by Chor Ping Low et al. (2008) and Gaurav Gupta et al. (2003): 1) It is more load balanced and energy efficient. 2) It is more efficient in terms of dead sensor nodes. 3) It has less time complexity, i.e., O (n log m) in contrast to O (mn2); reported by Chor Ping Low et al. (2008). 3. Energy Model We use the same energy consumption model as discussed by Heinzelman W. B. et al. (2002). In our work, both the free space and multi-path fading channel models are used, depending on the distance between the transmitter and receiver. If the distance is less than a threshold d0, the free space (fs) model is used; otherwise, the multipath (mp) model is used. Thus, the energy required by the radio to transmit an l-bit message through a distance d is given as follows.

E T (l , d )

lEelec lEelec

l l

fs d

2

mp d

4

, ,

d

d0

d

d0

(1)

where, Eelec is the electronics energy required by the electronics circuit, fs and mp is the amplifier energy in free space and multipath respectively. The radio also expends energy to receive an l-bit message given by.

773

774


ER (l ) lEelec

(2)

The Eelec depends on factors such as the digital coding, modulation, filtering, and spreading of the signal, whereas the amplifier energy, fsd2 or mpd4, depends on the distance to the receiver and the acceptable bit-error rate. In our simulations, the typical parameters are set same as Heinzelman W. B. et al. (2002) i.e., Eelec =50 nJ/bit, fs = 10 pJ/bit/m2, mp = 0.0013 pJ/bit/m4, d0=60.0m. In addition to that the energy for data aggregation is set as EDA = 5 nJ/bit. 4. Proposed Clustering Algorithm Here, we present our proposed algorithm. However we first describe all the assumption about the WSN model we use along with the associated terminology. We consider two kinds of nodes in the system; sensor nodes and less-energy-constrained gateway or CH. All communication is over wireless links. A wireless link is established between two nodes only if they are within range of each other. Gateways are capable of long-haul communication compared to the sensor nodes. All nodes are assumed to be aware of their position through GPS. Network setup is performed in two phases; bootstrapping and clustering. During the bootstrapping process, all the sensor nodes and gateways are assigned unique IDs. Sensor nodes broadcast their location and IDs to the gateways within communication range of each other. So, the distance from a sensor node to all the gateways within its range is calculated. In clustering phase sink executes the clustering algorithm. When the clustering is over, all the sensors are informed about the ID of the gateway they belong to. Depending on the communication range between the sensor nodes and the gateways, there can be two kinds restricted node open node Definition 1. (Restricted Node): Restricted nodes are those sensor nodes, which can communicate with one and only one gateway. Definition 2. (Restricted Set): Restricted set is the set of all restricted nodes in the WSN. We refer this set as Rset Si is belongs to Rset, if it satisfies the following criteria: Si

Rset

[{G j

Com(Si ) | G j

} {Gk

Com(Si ) | Gk

(

G j )}]

Where, Com (Si) is the set of all those gateways, which are within communication range of Si and is the set of all gateways. Definition 3. (Open Node): Open nodes are those sensor nodes, which can communicate with more than one gateway. Definition 4. (Open Set) sensor node Si is belongs to Oset, if it satisfies the following criteria:

Si

Oset

[Si

Oset

Rset ]

The basic idea of our clustering algorithm is as follows. We first assign the restricted nodes to their corresponding gateway. We then build a min-heap using the gateways depending on their respective number of assigned sensor (restricted) nodes. Let = {G1, G2, G3 Gm} be the list of gateways after building the minheap. So, G1 has been assigned by the minimum number of sensor nodes. Now we will assign that Si to G1 which is the nearest to G1 and G1

Com (Si). Now, we rearrange the min-heap. So, the gateway with next

775


minimum number of sensor nodes assigned to it will be at the root of the min heap. We pick up this gateway and assign a sensor node from Oset which is nearest to this gateway and within its communication range. The same procedure is continued until all the sensor nodes are allotted to their correct gateway. The algorithm considers both the issues, i.e., load balancing and the energy efficiency as follows. At each iteration, we assign a sensor node to that CH, which has the minimum number of sensor nodes already assigned to it. Therefore, the load is distributed over minimum loaded CHs. Thereby, balancing the load over all the CHs. On the other side, consumed energy heavily depends on the distance between two nodes. We assign that sensor node to a CH which is nearest to it; thereby reducing the overall energy consumption. The algorithm is formally presented as follows. Algorithm: Energy-Efficient-Load-Balanced-Clustering (EELBC) Input: A set of sensors T= {S1, S2, n}. A set of cluster heads {G1, G2, ..., Gm}; where, m < n. dij : For each Si and Gj, the distance between Si to Gj; where, Gj Com (Si). Rset and Oset. Output: An assignment A : T such that the overall maximum number of sensor node of CHs and total consumed energy is minimized. Step 1: While (Rset { Assign successive sensor nodes Si to their corresponding gateway Gj such that Si }

Rset and Gj

Com (Si) and delete Si from Rset and T

Step 2: Build a min-heap using the gateways on the number of allotted sensor nodes to the gateways Step 3: While (T { Step 3.1: Pick up the root node of the min-heap say Gj Step 3.2: Select and assign a sensor (Open) node Si to Gj such that Gj Com (Si) and Si is nearest sensor node to Gj Step 3.3: Delete Si from T Step 3.4: Adjust the min-heap so that the minimum loaded gateway will be at root } Step 4: Stop Time Complexity: Step 1 requires O(n) time for assigning Rset to their corresponding gateways. Step 2 requires O (m log m) time to build a min-heap using m number of gateways. In worst case, step 3 iterates n times in which Step 3.1 through Step 3.3 require constant time and Step 3.4 requires O (log m) time alone. Therefore, Step 3 can be executed in O (n log m) time. Thus the above algorithm requires overall time of O (n log m).

776


5. Experimental Results We performed extensive experiments on the proposed algorithm with the following experimental set up. The experiments were performed using MATLAB (version 7.5) on an Intel Core 2 Duo processor with T9400 chipset, 2.53 GHz CPU and 2 GB RAM running on the platform Microsoft Windows Vista. The experiments are performed with diverse number of nodes placed in a 1000 X 1000 square meter area by varying the number of sensor nodes from 100 to 500 and the number of CHs from 4 to 10. Each sensor node is assumed to have an initial energy of 2 joules. A node is considered dead if its energy level reaches to 0 joules. For 6 gateways.

For 6 gateways.

14 EELBC GLBCA LBC

10

3

Time in miliseconds.

12

Standard Deviation

3.5

8 6 4 2 0

2.5 2 1.5 1 0.5 0 100

100 150 200 250 300 350 400 450 500 No. of Sensor Nodes

2.(a) 4

400

EELBC GLBCA LBC

3

2

1

0

0

400

For 6 gateways.

For 6 gateways.

x 10

200 300 No. of Sensor Nodes.

2.(b)

Dead Sensor Nodes.

Consumed Energy(J).

4

EELBC GLBCA LBC

5

10 15 Time.

2.(c)

20

25

EELBC GLBCA LBC

300

200

100

0

5

10 Seconds

15

2.(d)

Fig. 2 Comparison between GLBCA, LBC and our proposed method EELBC in terms of (a) Load balancing, (b) Execution time, (c) Consumed Energy (J) and (d) Number of Sensor nodes dies. For the sake of comparison, we also ran the Chor Ping BCA (Chor Ping (Gaurav Gupta et al., 2003). In order to judge the quality Low et al., 2008) and Gaurav of the load balancing, we measure the standard deviation of the loads of the CHs and plot against the number of sensor nodes. It can be observed that our proposed algorithm EELBC is better than GLBCA and far better than LBC as shown in Fig. 2(a). We also obtain the execution time for run of the same experiments. As shown in Fig. 2(b) that the proposed algorithm EELBC is better than LBC and far better than GLBCA in terms of


execution time. In Fig. 2(c) and 2(d), we show energy (J) consumption and number of dead sensor node against per round respectively. It is observed that our proposed algorithm EELBC outperforms the GLBCA and LBC in terms of energy consumption and number of sensor nodes dies too. 6. Conclusion In this paper we have presented a clustering scheme for wireless sensor networks, in which some high energy gateways are treated as cluster heads (CHs). The algorithm takes care of the load balancing as well as energy efficiency. The algorithm has been shown to run in O (n log m) time for n sensor nodes and m CHs assuming equal load of the sensor nodes. Experimental results show that the proposed algorithm is more efficient with respect to load balancing and energy consumption than the similar works reported by Chor Ping Low et al. (2008) and Gaurav Gupta et al. (2003). Our future research will be towards the development of load balancing and energy efficient clustering, for the sensor networks with variable loads of the sensor nodes. We also make an effort to devise a scheme for the cluster head selection. References Akyildiz I.F., W. Su, Sankarasubramaniam Y., Cayirci E., 2002. Wireless Sensor Networks: A Survey. Computer Networks, Vol. 38, Issue 4, pp. 393-422. Kyung Tae Kim et al., 2009. An Energy Efficient Routing Protocol in Wireless Sensor Networks, Proc. of Int. Conf. on Comp. Sc. and Engg., IEEE, pp. 132 139. Emanuele Lattanzi et al., 2007. Energetic sustainability of routing algorithms for energy-harvesting wireless sensor networks, Computer Communication, Vol. 30, pp. 2976 2986. Chor Ping Low et al., 2008. Efficient Load-Balanced Clustering Algorithms for wireless, Computer Communications, Vol. 31, pp. 750 759. Gaurav Gupta, Younis M., 2003. Load-balanced clustering of wireless sensor networks, ICC '03, IEEE International Conference, Vol. 3, pp. 1848-1852. Ameer Ahamed Abbasi, Mohamad Younis, 2007. A Survey on clustering algorithms for wireless sensor networks, Computer Communications, Vol. 30, pp. 2826-2841. Olutayo Boyinbode et al., 2010. A Survey on Clustering Algorithms for Wireless Sensor Networks, 13th International Conference on Network-Based Information Systems, IEEE, pp. 358-364. Congfeng Jiang, Daomin Yuan, Yinghui Zhao, 2009. Towards Clustering Algorithms in Wireless Sensor Networks-A Survey, Wireless Communications and Networking Conference (WCNC), IEEE, pp.1-6. Heinzelman W.B., A.P. Chandrakasan, H. Balakrishnan, 2002. Application specific protocol architecture for wireless microsensor networks, IEEE Transactions on wireless communications, Vol. 1, No. 4, pp. 660-670. Lindsey S., C.S. Raghavendra, 2003. PEGASIS: power efficient gathering in sensor information systems, Proc. of the IEEE Aerospace Conference, Vol. 3, pp.1125-1130. O. Younis, S. Fahmy, 2004. HEED: A Hybrid, Energy-Efficient, Distributed clustering approach for Ad Hoc sensor networks, IEEE Transactions on Mobile Computing, Vol. 3, pp. 366 379. Manjeshwar A., Agrawal D. P., 2002. TEEN: A Routing Protocol for Enhanced Efficiency in Wireless Sensor Networks. Proc. of the 15th International Workshop on Parallel and Distributed Computing Issues in Wireless Networks and Mobile Computing, pp. 2009 2015, San Francisco, CA. Manjeshwar A., Agrawal D. P. 2002. APTEEN: a hybrid protocol for efficient routing and comprehensive information retrieval in wireless , pp.195-202, Fort sensor networks. in: Proc. of International Parallel and Distributed Processing Symposium (IPD Lauderdale, Florida (April 2002) Amir Sepasi Zahmati et al., 2007. Energy-Efficient Protocol with Static Clustering for Wireless Sensor Networks, World Academy of Science, Engineering and Technology, Vol. 28, pp.69-72. Bencan Gong et al., 2008. Multihop Routing Protocol with unequal Clustering for Wireless Sensor Networks, International Colloquium on Computing, Communication, Control, and Management, (ISECS2008), Vol. 2, pp. 552-556. Pratyay K., Prasanta K. J., 2012. Improved Load Balanced Clustering Algorithm for Wireless Sensor Networks, ADCONS 2011, LNCS, Vol. 7135, pp. 399-404. Zhixin Liu et al. 2011. A distributed energy-efficient clustering algorithm with improved coverage in wireless sensor networks, Future Generation Computer Systems, pp. 1-11. Xiang Min et al., 2010. Energy efficient clustering algorithm for maximizing lifetime of wireless sensor networks, Int. J. Electronics Communication. (AEÜ), Vol. 64, pp. 289-298. Wei Li, 2009. Energy Efficient Clustering Algorithm in Wireless Sensor Networks based on Geometric Programming. 2nd International Symposium on Electronic Commerce and Security, IEEE, Vol. 2, pp. 525-527.

777