Efficient Multi-Agent Reinforcement Learning for Pursuit-Evasion Game of Unmanned Surface Vessels with a Faster Evader

This paper presents a multi-agent reinforcement learning method that effectively learns the strategy for cooperative pursuit in the pursuit-evasion game of unmanned surface vehicles (USVs) with a faster evader. The proposed method is based on the multi-agent deep deterministic policy gradient (MADDPG) algorithm, which allows us to learn a position-free policy for the cooperative pursuers. Furthermore, we shape the reward function according to the Apollonius circle, which alleviates the sparse reward problem during learning. Experiments show that the proposed method can successfully learn a cooperative pursuit strategy for USVs to capture the faster evader and outperforms the vanilla MADDPG.