为提高柔性作业车间动态调度的灵活性与对车间扰动的实时响应能力，提出一种基于多智能体(Agent)动态博弈的柔性车间实时调度方法(Multi-Agent and Dynamic game Real-time Scheduling method，MDRS)。首先构建了基于多 Agent的实时调度系统，设计各类Agent之间的协商机制，通过Agent的分工合作以实现车间的实时生产调度与管控。在此基础上考虑机器故障与新订单加入两类异常事件，提出一种基于非合作动态博弈的实时分配策略。在异常事件发生时刻，根据工件和机器的实时信息重新建立动态博弈模型，形成以完工时间、生产总能耗和关键机器负荷为局中人的三方非合作动态博弈。设计基于逆向归纳的纳什均衡搜索算法以求解子博弈精炼纳什均衡，优化实时分配方案。最后借助JADE平台实现该系统，并在不同的测试场景下进行仿真，仿真结果表明3个优化目标可以达到纳什均衡，且相比其他算法均有不同程度的提升，验证了多Agent实时调度系统的可行性和MDRS的有效性。
Multi-Agent and dynamic game real-time scheduling decision method for flexible shop
To improve the flexibility of flexible job shop dynamic scheduling and real-time response ability to disturbance,Multi-Agent and Dynamic game Real-time Scheduling method(MDRS)based on multi-Agent dynamic game is proposed.First,a real-time scheduling system based on multi-Agent was constructed,and the coordination mechanism among Agents is designed.Through the division of labor and cooperation of Agents,the real-time production scheduling and control of the workshop are real-ized.On this basis,a real-time allocation method based on non-cooperative dynamic game is proposed considering machine failure and new task insertion.When the abnormal event occurs,the dynamic game model is re-established according to the real-time in-formation of the workpiece and the machine,and a three-party non-cooperative dynamic game is formed with completion time,total production energy consumption and key machine burden as players.Nash equilibrium search algorithm based on backward induc-tion is designed to solve sub-game refined Nash equilibrium and optimize real-time distribution scheme.Finally,the system is re-alized with JADE platform and simulated in different test scenarios.The simulation results show that the three optimization objec-tives can reach Nash equilibrium,and are improved to varying degrees compared with other algorithms,which verified the feasibil-ity of the multi-Agent real-time scheduling system and the effectiveness of the MDRS.
flexible job shopmulti-Agent systemreal-time schedulingdynamic game
国家科技期刊平台
NSTL
万方数据
