Task Allocation and Trajectory Planning in UAV Formation Counter-Swarm Missions

Unmanned aerial vehicle (UAV) swarms present significant threats to critical infrastructure due to their low cost and high maneuverability. Traditional counter swarm methods face effectiveness limitations when deployed individually and incur prohibitive costs when combined. Defensive UAV swarms leveraging similar attributes offer a promising swarm-on-swarm countermeasure strategy. These operations involve three critical stages: swarm aggregation, formation movement, and confrontation. While formation control technologies have matured, research neglects subsequent challenges in dynamic target allocation and tracking during confrontation. To address these challenges, this paper proposes an optimal controller based on unified reinforcement learning (RL) utilizing a single Actor-Critic (AC) network architecture that seamlessly handles both formation control and target tracking. Stability is rigorously guaranteed through Lyapunov theory, facilitated by an adaptive network update rate. Additionally, a novel Multi-step Sampling Task Allocation (MsSTA) algorithm incorporating formation communication constraints is proposed. This approach enables one-to-one dynamic target allocation through real-time position updates, batches allocations for computational efficiency, and minimizes communication disruption by restricting participation to peripheral UAVs.