A Study of Multi-UAV Cooperative Pursuit Based on PointNet-MATD3

In multi-UAV cooperative pursuit tasks, how to efficiently coordinate multiple agents to pursue high-speed escaping targets is an important research topic. This paper proposes a multi-agent deep reinforcement learning method based on PointNet-MATD3 (PointNet-Multi-Agent Twin Delayed Deep Deterministic Policy Gradient) to address the limitations of traditional methods in dealing with the interchangeability of multiple agents. This method combines the MATD3 algorithm with the PointNet network structure. By introducing a symmetric function to aggregate the information of all agents, the strategy learning of the agents is insensitive to the input order, thereby improving the training efficiency and convergence stability of the algorithm. Numerical simulation results show that, compared with the traditional MATD3 algorithm, PointNet-MATD3 achieves higher reward values in both the early and late stages of training, and significantly improves success rate and task completion efficiency. This study demonstrates the effectiveness of PointNetMATD3 in multi unmanned aerial vehicles(UAVs) pursuit-evasion games and provides a new solution for multi-agent cooperation in complex task environments.