2012 2nd IEEE International Conference on Parallel, Distributed and Grid Computing Adaptive Lagrangean Clustering Protocol Ravi Tandon, Biswanath Dey and Sukumar Nandi Department of Computer Science And Engineering Indian Institute of Technology Guwahati Guwahati, Assam 780139 Email: r.tandon@alumni.iitg.ernet.in, {bdey. sukumar }@iitg.ernet.in Abstmct-A Wireless Sensor Network is a dis tributed network of sensor nodes deployed in order to monitor physical or environmental conditions. One of the major challenges in the deployment of wireless sen sor networks is the efcient usage of the sensor nodes' limited energy. In this paper we propose a novel cluster ing protocol for homogeneous wireless sensor networks. The clustering protocol, Adaptive Lagrangean Cluster ing Protocol (ALCP), is a hybrid scheme. The protocol consists of a centralized and a distributed phase for cluster formation. It considers residual energy of sensor nodes for the election of cluster heads. The problem of cluster head election is modeled as a facility location problem. Lagrangean Surrogate heuristic is used as a solution to the problem. For energy efcient propaga tion of messages to Base Station hierarchical clustering is done. ALCP outperforms existing protocols in terms of network lifetime and energy distribution. 1. INTRODUCTION A vireless Sensor Network is a spatially distributed autonomous system of sensor nodes used for health moni toring, environment surveillance, military intelligence etc. A wireless sensor network consists of a Base Station and a set of scattered sensor motes. Sensor nodes collect data and send it to the Base Station. The Base Station acts as a sink. The characteristic features of wireless sensor networks such as the limitation of resources (processing power, energy) make them diferent from other distributed networks. Consideration of lifetime becomes essential for any network deployment. Network lifetime is one of the most essential parame ters for measuring the quality of network deployment. A sensor network is useful as long as it remains "alive" [1]. Therefore maintaining the aliveness of sensor nodes in a network becomes utmost important. Parameters such as connectivity, coverge and node availability depend upon the lifetime of a network. Data aggregation and clustering are two mechanisms that are employed to reduce energy consumption within a network. Data aggregation is defned as the process of aggregating data from multiple sensors to eliminate redundant trans mission and provide fused information to Base Station [2]. The process of grouping sensor nodes in a densely deployed wide-scale sensor network is known as clustering. A sensor network is divided into small independent groups called clusters. A single node from each cluster is chosen to be the cluster head. A cluster head performs scheduling, 978-1-4673-2925-5/12/$31.00 ©2012 IEEE 95 data aggregation and forwarding of data messages to Base Station. One of the major issues in clustering is the election of cluster heads. The cluster heads consume a large amount of energy per round. The problem of cluster head election is modeled as a facility location problem. A p-median solution is used to elect cluster heads. The p-median problem is the problem of locating p medians so as to minimize the transmission cost from each of the demand points to its nearest facility. The sensor nodes act as demand points and the cluster heads act as facilities. Lagrangean Surrogate heuristic is used to as a solution to the p- median clustering. Existing clustering protocols such as LEACH [3] and CTPEDCA [4] do not take the residual energy of sen sor nodes into consideration while forming cluster heads. Protocols such as HEED [5] and EECS [6] elect cluster heads stochastically using residual energy as a parameter during the election process. ALCP provides a deterministic cluster head election heuristic based up on the residual energy of the sensor nodes. It considers the position and energy remaining in sensor nodes as the two major factors for election of cluster heads. Hierarchical protocols such as CTPEDCA [4] incur overheads for the formation of a hierarchy of clusters. ALCP-H (hierarchical version of ALCP) circumvents the overhead by making use of over heard messages during the formation of hierarchy amongst cluster heads. Rest of the paper is organized as follows. Section II is a review of protocols for vSNs that have been proposed in the past. Section III provides the energy model. Section IV contains the network model. Section V is the mathematical formulation of the Lagrangean Surrogate heuristic applied to a \ireless Sensor Network. Section VI analyzes the clustering algorithm. Section VII describes the simulation study. The work is concluded in section VIII. II. RELATED WORK Signifcant research eforts have been devoted into the design of network clustering protocols. LEACH [3] uses randomization to elect sensor nodes as cluster heads. The drawback of LEACH is that it assumes the distribution of sensor node energy to remain uniform. As a sensor network progresses in its lifetime, residual energy in sensor nodes becomes unevenly distributed. LEACH does not elect sensor nodes based on their residual energy. HEED