UAV/USV Cooperative Trajectory Optimization Based on Reinforcement Learning

With the development of unmanned technology, unmanned aerial vehicle (UAV) and unmanned surface vehicle (USV) cooperative system is becoming more and more important in offshore operations. This paper consider the target search task on the sea, which requires USV to search target quickly, and UAV to track the USV in order to improve the observation accuracy and complete the task cooperatively. Based on target probability map model, a deep reinforcement learning algorithm based on proximal policy optimization (PPO) is adopted to optimize the search trajectory of USV. The PPO is also used to solve the UAV standoff tracking problem. The simulation results show that the deep reinforcement learning method can plan a reasonable search trajectory for USV, and UAV can achieve more accurate standoff tracking.