UAV Trajectory Design Based on Reinforcement Learning for Wireless Power Transfer

We studied wireless power transfer (WPT) system where unmanned aerial vehicle (UAV) broadcasts power to energy receivers (ERs) on the ground to solve the fairness problem. Since the design of the optimal UAV trajectory based on the location information of all ERs is not practical for UAV and requires high complexity, we apply Q-learning among reinforcement learning techniques to design the suboptimal trajectory with lower complexity. We confirmed that the proposed reinforcement learning-based trajectory design approaches the outer bound of the achievable region of two ERs.