Clustering in wireless sensor networks 

Abstract:   

        Sensor networks lack any infrastructure due to the limitations of individual nodes: current protocols for these networks, hence, are imported from mobile ad-hoc networks where frequent changes in topology critically drive the design choices. However, the need for every node to “re-evaluate” the network introduces scalability problems and inefficient usage of network resources. A hybrid solution to addresses these issues is by partitioning the network into clusters, with a designated cluster-leader node maintaining and providing networking services such as routing, data aggregation etc.

        The cluster-leaders, however, will deplete their battery faster than the rest of the sensor nodes and could become critical points of failure. We recently identified a distributed algorithm to identify size-bounded clusters. We had encouraging results about the stability of the clusters and message-complexity of this approach. The multi-hop nature of these clusters allows the flexibility of identifying multiple nodes in the cluster to serve as clusterheads: this ability to rotate the clusterhead will facilitate balancing the battery consumption across multiple sensors and overcome a single point of failure. We are currently working on such clusters for sensor networks and studying the consistency, stability and message-complexity issues.

Publications: