Reinforcement Learning for a Novel Mobile Charging Strategy in Wireless Rechargeable Sensor Networks

Abstract

The charging strategy for the mobile charger (MC) has been a hot research topic in wireless rechargeable sensor networks. We focus on the charging path for the MC, since the MC stops at each sensor node until the sensor node is fully charged. Most of the existing reports have designed optimization methods to obtain the charging path, with the target like minimizing the charging cost. However, the autonomous charging path planning for the MC in a changeable network is not taken into consideration. In this paper, Reinforcement Learning (RL) is introduced into the charging path planning for the MC in WRSNs. Considering the influences of the energy variation and the locations of the sensor nodes, a novel Charging Strategy in WRSNs based on RL (CSRL) is proposed so that the autonomy of the MC is improved. Simulation experiments show that CSRL can effectively prolong the lifetime of the network and improve the driving efficiency of the MC.