Deep Reinforcement Learning Based UAV Mission Planning with Charging Module

This paper focuses on UAV mission planning in an environment with charging using deep reinforcement learning, firstly proposing an improved actor-critic algorithm and constructing a reward function related to the UAV energy as well as the mission to guide the UAV behavior to achieve the completion of the mission, as well as the penalty of the action and the smoothing of the trajectory, for improving the UAV's ability to complete the mission, and secondly adding a charging module to ensure the balance of UAV energy. The simulation results show that the method can ensure the stability of UAV energy and when the energy module is added, the UAV energy level smoothes out and the mission completion is higher than before.