关键词: 异构无人集群系统, 线性时序逻辑, 协同任务规划, 意图预测, 动态目标, 在线规划

Abstract: Cooperative planning for heterogeneous multi-agents system is an important enabling technology for complex scenarios such as future space-air-ground integrated operations, intelligent security, and unmanned combat. In such scenarios, tasks are often specified with temporal constraints, cooperative requirements, and coupling with dynamic targets. Moreover, dynamic targets are driven by their own high-level intention and may exhibit reactive maneuvers. Therefore, relying solely on short-term trajectory prediction is hard to ensure the feasibility of cooperative planning solutions and the efficiency of task execution for heterogeneous system, which poses significant challenges to online decision-making, cooperative allocation, and real-time planning. To address these problems, the paper proposes an intention-aware cooperative planning method for heterogeneous multi-agents system with dynamic temporal tasks. By leveraging environmental semantics and target behavioral information, the method performs high-level intention estimation and long-horizon trajectory prediction for dynamic targets, and explicitly incorporates the prediction results into the cooperative planning process of heterogeneous system. In this way, an online closed-loop decision-making mechanism integrating “intention prediction, cooperative decision-making, and online adjustment” is established. Simulation results show that, under dynamic task updates and changes in target behavioral patterns, the proposed method can maintain the logical consistency of task progression and improve decision quality. Compared with conventional planning methods, the proposed method reduces the mean task execution time by 22.9% and variance by 32.3%, overcoming the limitations of low execution efficiency and weak dynamic adaptability in complex dynamic scenarios. This work provides new technical support for real-time and efficient cooperative planning for heterogeneous multi-agents system with dynamic temporal tasks.

Key words: heterogeneous multi-agents system, linear temporal logic, cooperative task planning, intention prediction, dynamic target, online planning