A Mobility Prediction-based Weighted Clustering Algorithm Using Local Cluster-heads Election for QoS in MANETs Vincent Bricard-Vieu ∗ ∗ LIRSA facult´ e des Sciences Mirande 9, av A. Savary BP 47870 21078 Dijon Cedex vincent.bricard-vieu@u-bourgogne.fr nmikou@u-bourgogne.fr Nidal Nasser † Noufissa Mikou ∗ † Deprt. of Computing & Infor. Sc. University of Guelph Guelph, Ontario Canada N1G 2W1 nasser@cis.uoguelph.ca Abstract—In this paper, we propose a new distributed Mobil- ity Prediction-based Weighted Clustering Algorithm with Local cluster-heads election (MPWCA-L) based on an on-demand distributed clustering algorithm for multi-hop packet radio networks. The multi-hop packet radio networks, also named mobile ad hoc networks (MANETs) have a dynamic topology due to the mobility of their nodes. This mobility makes the challenge harder for routing protocol. Moreover, the well known routing protocols are not able to offer QoS that is why we need to manage MANETs. Such task can be done using clustering techniques but the association and dissociation of nodes to and from clusters perturb the stability of the network topology, and hence reconfiguration of the system is often unavoidable. However, it is vital to keep the topology stable as long as possible. The nodes called cluster-heads form a dominant set and determine the topology and its stability. Simulation experiments are conducted to evaluate the stability of the dominant set in terms of updates of the dominant set, handovers of a node between two clusters and the QoS in terms of packet delivery rate, end-to-end delay and overhead provided by both our algorithm (MPWCA-L) and the Weighted Clustering Algorithm (WCA), which does not consider prediction and local cluster-heads election. Results show that our algorithm performs better than WCA. keywords: Ad Hoc networks, Clustering, Quality of Ser- vice, Mobility prediction, Local election. I. I NTRODUCTION The rapid advancement in mobile computing platforms and wireless communication technology leads us to the develop- ment of protocols for easily deployable wireless networks typically termed wireless ad hoc networks. These networks are used where fixed infrastructures don’t exist or have been destroyed. They permit the interconnectivity between work- groups moving in urban or rural area. They can also help in collaborative operations, for example, distributed scientific research or rescue. A multi-cluster, multi-hop wireless network should be able to dynamically adapt itself with the changing networks config- urations. Some nodes, known as cluster-heads, are responsible for the formation of clusters each consisting of a number of nodes (analogous to cells in a cellular network) and mainte- nance of the topology of the network. The set of cluster-heads is also called Dominant set. A cluster-head is responsible of resource allocation to all nodes belonging to its cluster. Due to the dynamic nature of the mobile nodes, their association and dissociation to and from clusters perturb the stability of the network and thus reconfiguration of cluster-heads is sometimes unavoidable. The paper is organized as follows. Section II describes previous clustering algorithms and our motivations. In section III we propose a local approach for cluster-heads election in a new distributed Mobility Prediction-based Weighted Cluster- ing Algorithm using Local cluster-heads election (MPWCA-L) and compare, in section IV, using simulations, its performance to those of the Weighted Clustering Algorithm (WCA)[1]. Section V concludes our study. II. RELATED WORKS Current algorithms for the construction of clusters contained in many routing protocols, as well as clustering heuristics, such as the lowest-identifier [2], the highest-degree [3][4] and the Linked-Cluster Algorithm (LCA) [5][6], have proactive strategies. By proactive, we mean that they require a constant refresh rate of cluster dependent information. That introduces a significant background control overhead even if there is no data to send. The major difficulty comes from node mobility, which has an impact on the position of the nodes and on the neighborhood information, which is essential for clustering. To ensure the correct collection of neighborhood information, existing clustering solutions rely on periodic broadcast of the neighbor list. Mobility causes adjacency relations to change. As well as in Lowest Distance Value (LDV) and the Highest In-Cluster Traffic (ICT)[7], depending on nodes movement and traffic characteristics, the criterion values used in the election - 24 - _________________________ 1-4244-0495-9/06/$20.00 ©2006 IEEE Downloaded from http://www.elearnica.ir