Game theory based node clustering for cognitive radio wireless sensor networks

Abstract

Cognitive radio wireless sensor networks (CRWSN) is a promising technology for developing bandwidth constrained applications. Future Internet of Things (IoT) applications may extensively use CRWSN. CRWSN consists of cognitive radio enabled sensor nodes which are energy constrained, in general. Hierarchical cluster based approach for overall network management is suitable for network stability and scalability. Thus node clustering is an important problem in CRWSN setup. Objective of this work is to develop a suitable node clustering algorithm for CRWSN, in which nodes are expected to be mobile. A node clustering protocol for CRWSN has been proposed in this paper. The proposed clustering protocol is based on evolutionary game theory (EGT). Initial clusters are formed through a simple partitioning approach. Eventually, initial clusters are merged to form the final clusters. After forming the clusters by the resourceful sink node, the cluster head nodes for respective clusters are determined. The sink node runs the EGT based algorithm to identify the most capable node as the cluster head. The strength of this approach is that while identifying the cluster head node, various parameters such as residual energy level, geographic location, mobility, and the probability of primary user (PU) arrival are considered. The clusters and therefore, the cluster head nodes are distributed uniformly in the geographical area. The proposed clustering protocol has been compared with LEACH, RARE and the spectrum aware clustering algorithm. The simulation results show that the proposed clustering protocol outperforms all these similar node clustering protocols. On average, the proposed protocol outperforms the benchmarks protocols by 25% in terms of number of high energy nodes selected as cluster head, by 37% in terms of uniform geographical distribution of cluster head nodes, by 23% in terms of total energy consumed during the simulation time, and by 27% in terms of network lifetime. The future scope of the work has been outlined.