Abstract:
Tiny devices known as sensor nodes capable of sensing computing and communicating
simultaneously to form a wireless sensor network. WSN has emerged as a revolutionary
technology which has brought with it numerous applications. Sensor nodes are placed in
the hostile environment to monitor different type activities or changes. Different
topologies to create a WSN from these sensor nodes have been proposed. Clustering is a
form of topology in which sensor nodes organize themselves in the form of cluster which
are common function to all the sensor nodes. It is important to describe an efficient
topology discovery algorithm to find a set of distinguished nodes. The main objective of a
clustering algorithm is to make the whole network connected in the form of efficient
topology. A clustering algorithm organizes the nodes in two phases: set up is the first
phase and maintenance of the cluster is the second phase. To optimally choose self
healing cluster heads is an NP-hard problem. The research discusses and compares
various strategies of self healing cluster formation in wireless sensor networks. Three
popular techniques of clustering, namely highest degree, topology discovery and
weighted clustering have been implemented. Two new algorithms, named maximal
weight and bounded degree have been proposed for the design and implementation of self
healing cluster based topology creation and management system. The detailed design and
working of maximal weight and bounded degree algorithms are presented with the help of
examples. The new proposed algorithms generate self healing clusters based, on
minimizing reconfiguration, thus saving energy and optimizing the communication. The
algorithms perform better than in comparison with the popular techniques in terms of
number of rearrangements i.e. reaffiliations and dominant set updates, number of clusters,
stability of clusters, and ratio of clusterheads to number of nodes. A detailed comparitive
analysis of the proposed algorithms against the topology discovery and weighted
clustering algorithm has been provided. The research work also highlights some open
research areas for securing wireless sensor networks.
