An efficient dynamic energy replenishment and data gathering strategy based on deep reinforcement learning in wireless rechargeable sensor networks

The advent of Wireless Rechargeable Sensor Networks (WRSNs) has brought about a new reality where sensor nodes can be recharged wirelessly via mobile charging vehicle. However, most of current methods fail to adequately combine the processes of replenishing sensor’s energy and gathering data in efficient manners. In order to reduce sensor node’s mortality, this paper proposes a dynamic energy reproduction and data gathering strategy based on deep reinforcement learning for WRSNs. Firstly, a multi-objective optimization model is established to evaluate both maximizing charging efficiency and minimizing data gathering latency. Secondly, the selection of charging path of the wireless charging vehicle and the determination of dwell time of rendezvous points are optimized collaboratively. Finally, the frame work of Long Short-Term Memory Twin Delayed Deep Deterministic Policy Gradient (LSTM-TD3) is utilized to implement a dynamic adjustment of the residence points’ dwell time. Extensive experiments show that our proposed method improves energy efficiency by 7.5 % and reduces average data latency by 11.3 % approximately, which can confirm its superior effectiveness and efficiency in terms of charging performance and data gathering.