Robust inventory management with multiple supply sources

Abstract

We study a robust rolling horizon model for a periodically reviewed inventory system with multiple supply sources with general lead times. We restrict the demands in an uncertainty set without knowing the distributions. We prove that under some appropriate conditions, the robust optimal policy for multiple sourcing is a combination of the base-stock policy and a gap-of-base-stock policy with capping effect on supply sources except the fastest source. In particular, the structure of this policy is not a natural extension of the robust optimal policy for the system with two supply sources, for which the recent literature shows that the robust optimal policy is a dual-index, dual-base-stock policy that caps the slow order. We also derive closed-form expressions of the robust optimal policy for the central limit theorem (CLT) uncertainty set. We numerically testify to the effectiveness of the structure of the robust optimal policy for more than two supply sources using both simulation demand and real sales data. The computational results are promising.