OPTIMIZATION OF WEIGHTED CLUSTERING ALGORITHM IN MOBILE ADHOC NETWORKS

Abstract

Wireless cellular network is restricted with geographical boundary and rely on wired backbone by which all base stations are connected. In utmost emergency situation such networks cannot be set and due to wired configuration cost of scaling increases. Wired (central) backbone cannot be implemented in critical area and situations such as law enforcement operations, battle field communications, disaster recovery situations, and so on. Therefore, mobile multi-hop radio networks, also called adhoc or peer-to-peer networks, play a critical role. Such situations demand a network where all the nodes including the base stations are potentially mobile, and communication must be supported untethered between any two nodes. Wireless systems with multi-cluster, multi hop packet radio network architecture should be dynamically adaptable with changing network configurations. Clusterheads which acts as central nodes are responsible for the cluster formation each consisting of supporting nodes for the cluster head. Cluster head also take responsibility of cluster maintenance. Dominant set is the set of cluster heads in the network. Responsibility of cluster head is (a) resource allocation to all nodes belonging to the specific cluster (b) serving the mobile nodes of the cluster for routing the data packet and (c) maintaining the cluster. Due to dynamic nature of the mobile nodes, cluster heads changes time to time. Reconfiguration of cluster takes place and associated mobile node gets disassociated with the particular cluster. This is an important issue since frequent clusterhead changes adversely affect the performance of other protocols such as scheduling, routing and resource allocation that rely on it. Election of clusterheads is an NP-hard problem. Probability of electing the optimal cluster heads must of high in any clustering algorithm. Clustering algorithm must be fast and cost of clustering must be as minimal as possible. Network must be stable as long as possible as well as scalable with the clustering process. Best clustering scheme must preserve its structure with dynamics of the mobile node i.e. when node changes its topology or with movement clustering must retain its structure. There is high computation overhead associated with the clustering and cost of clustering affects the network stability. Clustering is nothing but dividing the geographical region into small zones. Any node in the network can be clusterhead, depends on clustering node dynamic such as power, mobility, degree difference and sum of distances. Node associated with particular clusterhead become member of the cluster. Cluster of mobile nodes can be changed as the time progresses in the network.