A residual energy gradient cognitive scheme for cluster reconstruction in wireless sensor networks

In the wireless sensor network (WSN), the energy efficiency is the most important issue and, for that, clustering mechanisms have been extensively studied in recent years. One of the most well-known clustering mechanisms for WSNs is the LEACH protocol in which cluster heads (CHs) are periodically selected, resulting in large energy consumption. To overcome this problem, ACAWT performs re-clustering when the residual energy level of a CH reaches a given threshold. In this paper, we propose an energy-efficient clustering mechanism that adopts multiple energy level thresholds for re-clustering and the waiting timer for CH election whose value is determined based on the residual energy level and the number of neighbors of a sensor node. Using a Markov Decision Process (MDP), we analyze the optimal policy consisting of the decision sequences of sensor nodes with a given number of battery energy levels and with the function of either a general sensor node or a CH, so as to maximize the generalized system performance in terms of the network lifetime. Based on simulations, we verify that our proposed cluster reconstruction mechanism outperforms ACAWT in terms of the network lifetime.