Cooperative Target Pursuit by Multiple Fixed-wing UAVs Based on Deep Reinforcement Learning and Artificial Potential Field

This paper presents a cooperative target pursuit algorithm based on deep reinforcement learning and artificial potential field for multiple fixed-wing unmanned aerial vehicles (UAVs) in the complex environment. In the proposed algorithm, decentralized deep deterministic policy gradient is employed to learn cooperative target pursuit policies that are adaptive to complex environments including static obstacles and dynamic evader. The artificial potential field method is combined into the learning process to avoid obstacles and collision. In order to cooperatively pursuit, the reward is designed to incentivize the capture of target and encourage a beneficial formation of pursuers. Finally, the simulation shows the feasibility of the proposed algorithm in learning multiple cooperative pursuit strategies.