On optimizing the charging trajectory of mobile chargers in wireless sensor networks: a deep reinforcement learning approach

Wireless rechargeable sensor networks (WRSNs) are broadly utilized in numerous areas. However, the limited battery capacity of sensor nodes (SNs) is considered as a critical issue. To extend the battery life of SNs, mobile chargers (MCs) equipped with wireless power transfer (WPT) technology have been proposed as a key solution for charging SNs. Using directional antennas to focus energy within a specific area, as opposed to an omnidirectional antenna, increases the energy efficiency of an MC. In this paper, we focus on the travel path charging scheduling problem with a directional MC in on-demand WRSNs. Our goals are to develop a mechanism to reduce the changing delay time and boost the energy efficiency of MC. In this case, the MC receives the charging requests of SNs and responds to them by selecting appropriate stopping points (SPs) and the charging orientation angles in each SP. We propose a mobile directional charging scheduling (MDCS) solution based on a deep reinforcement learning technique. The simulation results demonstrate the superior performance of our method to existing studies in terms of the energy consumption of the MC, the number of dead SNs, and charging delay time.