A Trajectory Planning Algorithm for Data Collection in UAV-aided Wireless Sensor Networks

In this paper, we study an unmanned aerial vehicle (UAV)-aided wireless sensor network (WSN), where a UAV is dispatched as a mobile sink to collect data from distributed sensor nodes. A system model is introduced in terms of the event driven sensor nodes and value of information (VoI). We focus on the trajectory planning problem in UAV-aided WSN for data collection. We propose a multi-objective model of data VoI maximization and UAV trajectory path minimization, meanwhile, satisfying the UAV's charging requirements. The Prioritized Replay Deep Q-Network (PR-DQN) algorithm is proposed for efficient trajectory planning, which is based on the discrete motion deep reinforcement learning architecture and the prioritized experience replay. Simulation results show that our method outperforms traditional Q-Learning and DQN in the rate of convergence.