A basestock level for a buffer is a quantity s such that if x denotes
the amount of available inventory, the amount to order is max(0, s - x).
Basestock policies are extremely popular in supply-chain applications; under appropriate stochastic models of demand uncertainty, they are optimal. In this talk we discuss algorithms for computing optimal basestock levels under the robust optimization paradigm. Our algorithms are based on Benders' decomposition and scale well (under several models of demand uncertainty) with problem size, e.g. up to several hundreds of periods. We will also discuss the extension to safety stocks, to ambiguous demand uncertainty and to model superposition.
