DRG-MAPPO is a hierarchical dynamic role-graph multi-agent reinforcement learning approach for cooperative air combat. It models structured, time-varying relationships among battlefield entities to support more sophisticated tactical coordination.