OPTIMIZED CLUSTERING APPROACH FOR WIRELESS SENSOR NETWORK

Abstract

Wireless sensor network play an important role in today’s networks which are mainly focusing on automating the computing needs of the world. These networks can work better in an environment which is unattended and inaccessible. As such networks have constrained potentials such as battery power, radio transmission range, memory and processing hardware of the sensor nodes which makes them less robust in case of node failures, inefficient, poor scalability. One of the main unique features of WSN is that they have wireless connectivity among the small size sensor nodes which makes them work even in harsh conditions with limited human support. WSNs is a type of an ad-hoc networks that involves the large number of small, lightweight sensor nodes. Here, the energy of nodes is a critical issue in WSN because of limited battery life. There are some of the challenging issues in WSN like energy consumption, traffic load balancing, data aggregation, data redundancy, limited wireless communication range. In this thesis, we have worked towards designing of new energy efficient clustering algorithms to achieve higher energy performance along with longer life span of the network in WSNs. we have proposed two new improved clustering algorithms for the purpose of efficient clustering analysis of data sets to offer high energy efficiency and reduced communication overheads. The first algorithm will solve the difficulty of finding the optimal number of clusters with their cluster centroids and arbitrarily shaped clusters set by uniting model and density based algorithms. It also presents an idea for estimating input parameters to minimize intra and inter-cluster distance between sensor nodes. The proposed hybrid algorithm reduces wireless data transmissions and increases the speed of transferring the accumulated data to the mobile sink node located at the cluster centroid that patrols by following the shortest path. The second algorithm presents a new idea for cluster formation phase and cluster head selection phase based on density of the sensor node. It not only finds the optimal number of clusters in every round, but also gives an idea of electing cluster head based on density using residual energy, average energy and distance from the location of cluster centroid. After cluster formation phase, all the member nodes convey their data to selected CH node during each round. The CH node aggregates the received data to remove duplicate number of packets so that the it can be communicated to static or mobile sink node in less time with higher transmission speed when the distance between them is minimal. This suggested algorithm guarantees uniform distribution of CH in the sensing field to preserve the energy of sensor nodes.