Joint Energy Replenishment and Data Collection Based on Deep Reinforcement Learning for Wireless Rechargeable Sensor Networks
Lingli Zhang
Abstract
Rechargeable wireless sensor networks have been extensively studied to provide a stable and sustainable power supply for the sensors in many applications for years. However, the complex scenarios and variation of the sensor node’s energy consumption imposes a great challenge in obtaining an optimal charging scheduling. In order to solve this problem, we propose a joint energy replenishment and data collection based on deep reinforcement learning for Wireless Rechargeable Sensor Networks (WRSNs). Firstly, the charging efficiency of Mobile Charging Vehicle (MCV) is analyzed, and the energy utilization rate of MCV can be measured based on the constraints of energy carried by MCV and the time cost of data collection. Then, to maximize the energy utilization of MCV and minimize the delay of data collection form sensor nodes, a multi-objective model with constraints is established. Finally, an optimal path planning algorithm for MCV based on Deep Reinforcement Learning (DRL) is presented. By using DRL to learn path planning strategies from experience, the optimal path planning and balanced charging solution can be obtained under the premise of limited driving and charging energy carried by MCV. The simulation results show that compared to typical methods, our proposed algorithm can effectively optimize the data collection from sensor nodes and improve the energy utilization efficiency of MCV.