Optimizing multiproduct assemble-to-order (ATO) inventory systems is a long-standing difficult problem. We consider ATO systems with identical component lead times and a general "bill of materials." We use a related two-stage stochastic program (SP) to set a lower bound on the average inventory cost and develop inventory control policies for the dynamic ATO system using this SP. We apply the first-stage SP optimal solution to specify a base-stock replenishment policy, and the second-stage SP recourse linear program to make allocation decisions. We prove that our policies are asymptotically optimal on the diffusion scale, so the percentage gap between the average cost from its lower bound diminishes to zero as the lead time grows.
ASJC Scopus subject areas
- Computer Science Applications
- Management Science and Operations Research