(A) reinforcement learning approach to lot size scheduling

Abstract

In this paper, we study a lot size scheduling problem similar to the actual factory size. In fact, building process and curing process are important processes that determine product types and produce finished products in tire factories. The curing processing is less variable in scheduling and is expensive due to idle time because it maintains energy even in the absence of production. Therefore, lot size scheduling is carried out to prevent idle time of the curing process during the building process, which is a large scale scheduling problem, which poses difficulties. In this paper, we define this as a lot size scheduling problem in large scale and propose reinforcement learning methodology for this. We compare and analyze intuitive rule based heuristic algorithms and reinforcement learning based methodology, and compare the differences in performance with mixed integer programming models. Finally, we want to analyze the behavior of reinforcement learning models and confirm their applicability to real-world sites.