Energy centric reputation index and fuzzy-based clustering for wireless sensor networks

Abstract

Energy conservation is essential for a wireless sensor network to enhance its network lifetime. Clustering is one of the most important energy conservation techniques that prolong the WSN by providing balanced energy depletion across the network. Various clustering strategies have been proposed in the past to enhance the efficiency of the network till now. In the last two decades, fuzzy inference system based clustering solution has been proven to provide promising solutions. The existing fuzzy-based clustering algorithms consider the naïve network parameters as input that saves only a limited amount of energy in the network. However, in wireless sensor networks, the requirement is to make them highly energy efficient through an effective clustering strategy. To achieve effective clustering there is a need for energy-centric network parameters for the clustering instead of naïve network parameters. In this article, a novel energy-centric reputation index is proposed in which the naïve network parameters are unified with energy parameters in the background. Further, the reputation index scores are passed into a fuzzy-based approach to obtain the fitness value of the node chosen as the cluster head. Thus, in this paper, an energy-centric reputation index and fuzzy-based clustering(ECRIF) model is proposed for wireless sensor networks. The proposed model provides a highly energy-efficient wireless sensor network and is designed to work in two phases. In phase 1, three reputation indexes concerning internodal distance, distance to the sink, and the degree of the node is calculated while considering their energy consumption pattern in the background. In phase 2, the three reputation indexes are passed onto the fuzzy system as fuzzy input parameters to obtain the fitness value of the nodes for cluster head selection. Simulation results confirm that the proposed protocol enhances the overall lifetime of the network on an average by 70% in terms of the first node dies, 61% in terms of half node dies, and 78% in terms of the last node dies when compared with two other existing clustering models.