Tag
A hierarchical multi-agent reinforcement learning framework combining graph attention and dynamic role assignment improves tactical coordination and win rates in air combat.