Heterogeneous Multi-Mission Task Planning for UAV Swarm Based on Hierarchical Deep Reinforcement Learning

Fixed-wing unmanned aerial vehicles (UAVs), as one of the most prevalent UAV types, are particularly suited for large-scale battlefield operations, especially in UAV swarms, where effective attack planning is critical. However, existing swarm planning algorithms often fail to account for the complexity of heterogeneous tasks and their coupling. To address this problem, we propose a hierarchical deep-reinforcement learning framework for attack planning in fixed-wing UAV swarms. The proposed algorithm balances competition among UAVs while optimizing both individual performance and overall swarm effectiveness. Our approach decomposes the problem into high-level task allocation and low-level path planning. The task allocation phase incorporates trajectory complexity to enhance global mission coordination, while the path planning phase optimizes real-time execution. Ensuring a balance between individual UAV performance and overall swarm efficiency. Experimental results demonstrate that our method improves mission completion by 10.77%, reduces path length by 15.81%, decreases path risk value by 22.8%, and increases battlefield combat power by 10.77%, outperforming existing approaches. These findings provide a promising solution for UAV swarm coordination in complex multitask scenarios