Charging Efficiency Optimization Based on Swarm Reinforcement Learning Under Dynamic Energy Consumption for WRSN

Wireless rechargeable sensor networks (WRSNs) have been widely used to solve the energy constraint problem of wireless sensor networks (WSNs). Improving the charging efficiency of the mobile charger (MC) is crucial for WRSN. However, the time-varying states of WRSN caused by dynamic changes in node energy consumption during the charging process of the MC make the optimization of charging efficiency rather difficult in WRSN. To solve this problem, a kind of optimization approach based on swarm reinforcement learning (SRL) is presented in this article. The presented approach lets the multiple agents better adapt to the dynamic energy distribution in WRSN by designing a dynamic energy consumption model. Then, it utilizes a rank-based ant system (AS(rank)) to ensure that the MC gets the initial optimal requesting sensor nodes (SNs), which contributes to accelerating the convergence speed of SRL. Finally, it adopts particle swarm optimization (PSO) to improve the learning effectiveness during the exchanges of experience among multiple agents, which contributes to optimizing the charging path of MC. Extensive simulations show that the presented approach achieves better charging performance than the existing typical approaches, and it has significant advantages in terms of charging efficiency, SN dead ratio, and MC energy efficiency ratio.