A new spectral classification for robust clustering in wireless sensor networks

Abstract

Wireless sensor network has recently become an area of attractive research interest. It consists of low-cost, low power, and energy-constrained sensors responsible for monitoring a physical phenomenon and reporting to sink node where the end-user can access the data. Saving energy and therefore extending the wireless sensor network lifetime, involves great challenges. For these purposes, clustering techniques are largely used. Using many empirical successes of spectral clustering methods, we propose a new algorithm that we called Spectral Classification for Robust Clustering in Wireless Sensor Networks (SCRC-WSN). This protocol is a spectral partitioning method using graph theory technics with the aim to separate the network in a fixed optimal number of clusters. The cluster's nodes communicate with an elected node called cluster head, and then the cluster heads communicate the information to the base station. Defining the optimal number of clusters and changing dynamically the cluster head election probability are the SCRC-WSN strongest characteristics. In addition our proposed protocol is a centralized one witch take into account the node's residual energy to define the cluster heads. We studied the impact of node density on the robustness of the SCRC-WSN algorithm as well as its energy and its lifetime gains. Simulation results show that the proposed algorithm increases the lifetime of a whole network and presents more energy efficiency distribution compared to the Low-Energy Adaptive Clustering Hierarchy (LEACH) approach and the Centralized LEACH (LEACH-C)one.