Multi-robot multi-station manufacturing system has gained significant popularity in welding workshop for processing large jobs. The implementation of cooperative welding robots effectively enhances the efficiency and ...
详细信息
Multi-robot multi-station manufacturing system has gained significant popularity in welding workshop for processing large jobs. The implementation of cooperative welding robots effectively enhances the efficiency and flexibility of the system. However, robot failures also bring challenges for rescheduling to determine the rescheduling timing and re-assign appropriate robots for the affected processes so as to minimize the losses in performance and the maximum completion time. Distinct from open-looprescheduling approach, this paper proposes a closed-loopreschedulingmethod via deep reinforcement learning and improved artificial bee colony algorithm. This method addresses two crucial questions: when to trigger a rescheduling, and how to provide a rescheduling scheme. For the first issue, current methods make decisions according to indicators derived from expert experience, failing to adapt to the dynamic environment. Therefore, we propose a deep reinforcement learning method to learn the adaptive trigger policy for the trigger agent under real-time running state. For the second issue, we present an improved artificial bee colony algorithm to re-assign robots without changing processing sequence and feed the performance back to trigger agent. Specially, a variable local search method is incorporated to avoid falling into local optimum. Finally, the proposed method is applied to solve the instances with different scales and a practical case. The experimental results of both demonstrate that the proposed closedloopreschedulingmethod can greatly reduce the deviation of the total completion time and the maximum completion time and have superior performance over other algorithms.
暂无评论