UAV Swarm Confrontation Decision Optimization Based on MADDPG

This paper proposes an optimal decision making method based on the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm to solve the problem of sparse gradient and local optimal in cooperative decision making in UAV swarm confrontation task. By integrating a self-attention mechanism to improve global information capture and adopting a curriculum learning strategy to increase task difficulty, the proposed method effectively mitigates issues such as sparse gradients and local optima that frequently occur in complex adversarial scenarios. Experimental results show that, compared to conventional MADDPG, the proposed approach enhances key performance indicators such as attack success rate, thereby demonstrating its superior capability in cooperative confrontation decision-making for UAV swarms.