基于Petri网的异构多机器人系统任务和路径组合规划

多机器人系统在智能制造、搜索救援及运输物流等国防、工业、社会服务诸多领域的应用呈现高速增长的态势，具有广泛的应用前景。现有的多机器人系统任务和路径规划方法只能处理特定条件下的简单任务，难以处理存在多种约束的复杂任务场景下的规划问题。因此，本项目针对复杂任务场景，建立包含复杂任务、异构多机器人系统和多种约束的统一可扩展赋时-颜色Petri网模型，并设计基于事件域模型和图神经网络的任务和路径组合规划算法，实现任务目标的合理分配及执行任务过程中的路径优化。统筹考虑系统资源不足、复杂环境以及任务限制等约束导致的死锁问题，设计多机器人系统死锁避免规划策略，提高系统处理多约束问题的能力。最后，搭建仿真和实验平台验证模型和算法的有效性。项目研究将进一步发展离散事件系统监督控制理论和多机器人系统任务和路径规划方法，为推动异构多机器人系统在实际复杂任务场景中的应用提供关键技术支撑。

In fields of national defense, industry, and social services, such as intelligent manufacturing, search and rescue, transportation, and logistics, the applications of multi-robot systems show a rapid growth trend and a wide range of application prospects. However, the existing methods for task-planning and path-planning in multi-robot systems can only handle simple tasks, which is challenging to deal with the planning problems in complex task scenarios. Therefore, this proposal focuses on developing a unified framework of task and path planning for complex tasks. Specifically, complex tasks and heterogeneous multi-robot systems will be modeled with Timed Colored Petri Net (TCPN). Based on their TCPNs, a combined task and path planning scheme will be developed via graph neural network and deep reinforcement learning, thus realizing the task assignment and path optimization in the mission process. Considering multiple constraints including resource limits, complex environment, and task sequence, deadlock avoidance methods are proposed when performing task and path planning, thus improving the ability to deal with multiple constraints. Finally, simulation and experimental platforms are developed to verify the effectiveness of the proposed planning method. This research will make further development for supervisory control theory of discrete event systems and the task-planning and path-planning methods in multi-robot systems, consequently providing key technical support for the application of heterogeneous multi-robot systems in actual complex tasks scenarios.