66 A novel clustering algorithm for efficient energy saving in Wireless Sensor Networks 0. Moussaoui, A. Ksentini, M Naimi LICP EA 2175, Universitei de Cergy-Pontoise- 2 Av. Adolphe Chauvin 95302 Cergy-Pontoise - France tomar. moussaoui, adlen. ksentini, mohamed.naimi@,dept-info. u-cergy.fr Abstract-Energy efficiency operations are essential in extending Wireless Sensor Networks lifetime. Among the energy- saving-based solutions, clustering sensor nodes is an interesting alternative that features a reduction in energy consumption through: (i) aggregating data; (ii) controlling transmission power levels (iii) balancing load; (iv) putting redundant sensor nodes to sleep. This paper introduces a novel clustering algorithm that uses a distributed approach to set up non-overlapping clusters, while performing cluster heads rotation as well as carrying out other energy intensives tasks. Besides the transmission power and the energy level of each node, the proposed clustering algorithm considers nodes number that a cluster head can handle aiming at ideally partitioning the network. In fact, the nodes number constituting the cluster is confined around a pre-defined threshold, facilitating thus the optimal operation of the medium access control (MAC) protocol. Further, intending to reduce the computation and the communications cost as well as the messages exchanged, the proposed algorithm invokes the cluster formation only at the system activation, and delays as long as possible the cluster head turn over. Keywords: sensor networks; clustering; energy efficiency; network lifetime. I. INTRODUCTION Advances in micro-electro-mechanical systems (MEMS) technology, wireless communications as well as digital electronics have prompted the development of extremely small and low-cost sensors, which possess the capabilities of: (i) sensing; (ii) signal processing; (iii) communicating through wireless channel [1]. Wireless Sensor Network (WSN) is an emerging technology that allows regrouping sensor nodes in order to form a collaborative network. Many application fields can be considered for using WSN, such as target detection, tracking in a battlefield situations and habitat monitoring in remote areas. WSNs are highly affected by the energy dissipation of the nodes. In fact, sensor nodes are usually used to collect and report application-specific data to the monitoring node, known as a "sink node". In this context energy dissipation are due to either "useful" or "wasteful" sources. On the one hand, useful energy consumption results from transmitting/receiving data, processing query requests and forwarding queries/data to neighbouring nodes. On the other hand, wasteful energy consumption arises from idle listening to the media, retransmitting due to packet collisions, overhearing, and generating/handling control packets. Since sensor nodes carry limited and non-replaceable power sources, the efficiency of M. Gueroui PRISM CNRS, Universitei de Versailles-45, Av des Etats-Unis 78035 Versailles - France mogue@,prism. uvsq.fr energy use determines the lifetime of the sensors and, consequently, the duration of the sensing task and the network lifetime. Here, the network lifetime is defined as the time elapsed until the first node (or the last node) in the network depletes its energy (dies). Energy saving in sensor networks is an open challenging issue, where a substantial amount of works has been done. Among these works, clustering-based approaches are showing the most exiting result through their ability to reduce energy consumption by multiple ways. In fact, hierarchical-clustering reduces: (i) the amount of query packets via inter-cluster query dissemination; (ii) the amount of data packets by aggregating collected data. Further, hierarchical-clustering needs a minimal number of active nodes to cover the target area. This is achieved by putting redundant sensor nodes in sleep mode. At this point, hierarchical-clustering is particularly useful for applications that require scalability to hundreds or thousands of nodes. However, hierarchical-clustering scheme deals with some constraints. The most relevant concerns the Cluster-Head (CH) reconfiguration process. Actually, to alleviate the large amount of energy consumption required by a CH, frequent reconfigurations of clusters are needed. The existing hierarchical-clustering approaches, select at first a set of CHs among the nodes in the network, and then cluster the rest of the nodes with these heads. In this context, CHs are responsible for coordination among the nodes within their clusters (intra-cluster coordination) as well as communicating with other CHs (inter-cluster communication). The other sensor nodes have just to transmit their information to their respective CH, which aggregates the received information and forwards it to the sink. Actually, partitioning the network in clusters as well as choosing a CH optimally are an NP-hard problem[2]. The existing solutions to this problem are based on heuristic approaches and none attempts to retain the stability of the network topology[2]. We believe that a good clustering scheme should preserve its structure as much as possible when nodes are moving and/or the topology is slowly changing. Otherwise, re-computation of cluster heads and frequent information exchange among the participating nodes will result in high computation overhead. In this paper, we introduce a novel clustering algorithm towards minimizing the energy consumption in wireless sensor- based networks. The proposed algorithm partitions the network into different clusters based on: Proceedings of the Seventh IEEE International Symposium on Computer Networks (ISCN' 06) 1-4244-0491-6/06/$20.00 © 2006 IEEE