Dispatching with Reinforcement Learning: Minimizing Cost for Manufacturing Production Scheduling
In the manufacturing industry, a factory receives production job orders from customers, starts processing job orders, and finally delivers products to customers. There are costs in this process sequence. Each job comes with a due time. Past-due cost is the cost when a job cannot be delivered on time. Inventory cost is the storage cost when the job is finished before due time and stays in factory. It is obvious that all manufacturing managers want on-time delivery and minimum past-due cost and inventory cost. However, achieving these goals requires an efficient dispatching rule. A dispatching rule is used to select the next job to be processed from a set of jobs awaiting service. This article describes work being conducted at Hitachi America’s R&D Division focusing on dynamic dispatching for due date-based objective where we proposed a reinforcement learning based approach for this problem.
