Trajectory Planning of UAV in Wireless Powered IoT System Based on Deep Reinforcement Learning

In this paper, a UAV-assisted wireless powered communication system for IoT network is studied. Specifically, the UAV performs as base station (BS) to collect the sensory information of the IoT devices as well as to broadcast energy signals to charge them. Considering the devices' limited data storage capacity and battery life, we propose a multi-objective optimization problem that aims to minimize the average data buffer length, maximize the residual battery level of the system and avoid data overflow and running out of battery of devices. Since the services requirements of IoT devices are dynamic and uncertain and the system can not be full observed by the UAV, it is challenging for UAV to achieve trajectory planning. In this regard, a deep Q network (DQN) is applied for UAV's flight control. Simulation results indicate that the DQN-based algorithm provides an efficient UAV's flight control policy for the proposed optimization problem.