Coalition Formation Game-Based Multi-Level Multi-Stage Dynamic Task Decision for Large-Scale Heterogeneous UAVs

Considering the diversity of tasks and the variability of environments in practical decision-making process, it is essential to maximize the advantages of heterogeneous UAVs and adjust decision strategies in real-time. In this paper, a coalition formation game (CFG)-based multi-level multi-stage task decision (MLMSTD) strategy is proposed to achieve a rapid response of heterogeneous UAVs to dynamic environments and their demands. UAV needs and tasks are classified into multiple levels based on priority queue technique, ensuring the safety of UAVs and the strategic nature of decision-making results. To reduce computational burden and improve the dynamic task decision speed, a task decision strategy is proposed to decompose the complex dynamic task decision into multiple stages. First, UAVs and tasks participating in dynamic task decisions are identified based on their needs and levels. Then, task leaders are determined based on weighted maximum matching approach. Next, the remaining UAVs communicate with the leaders, respond to their optimal tasks through switching operations, and form intermediate coalitions. Finally, the leaders resolve conflicts based on the cost-optimal principle to obtain the final task decision result. Simulations in various scenarios are conducted to verify the effectiveness, scalability, and superiority of the proposed strategy. In a specific scenario involving 80 heterogeneous UAVs and 5 tasks (comprising 15 subtasks), the proposed MLMSTD strategy achieves a solution speed improvement of at least 21.43% compared to baseline methods.