大型生產系統之多智能體動態派工與預保養架構

Abstract

本篇論文開發具泛用性的分散式多智能體動態派工與排程方法，預計應用在大型智慧製造系統中，針對系統狀態提出對應的動態派工與預保養決策。本研究將使用動態規劃結合馬可夫決策過程求解考慮了系統即時動態特性與未來動態特性的排程方法。同時，為了解決動態控制方法面臨的維度詛咒，本研究採取多智能體架構針對大型問題進行求解。透過拆解大系統、求子系統最佳解與整合子系統最佳解的流程，得到整個系統的較佳初始解。其中，本研究主軸在於多智能體間的合作與協調，透過虛擬分派機的作用，組合子系統的動態規劃收斂值並得出整個系統的成本估計值，以利未來進行全域最佳解的搜索。

This study presents a dynamic dispatching and scheduling method under the structure of distributed Multi-agent System. The method can be applied in large intelligence manufacturing system, making dynamic dispatching and preventive maintenance decisions according to the state of agent. To generate the solution under reasonable time complexity, we work under Multi-agent System. By decomposing the problem into sub-problems and solving the sub-problems using dynamic programming and Markov Decision Process, we get the optimal solution of each agent. Finally, we develop an algorithm to coordinate the solution of sub-problems and get a good initial solution of the large system.