A new algorithm for a multi-item periodic review inventory system

In this paper, the authors deal with a multi-item, stochastic, periodic review inventory system with general cost structure which permits partial or complete backlogging of unfilled demand. Since both the (σ,S) policy and the mixed reorder policy are not optimal, they derive several properties of an optimal ordering policy and propose a new algorithm for computing it. This algorithm is based on the policy iteration method (PIM), but reduces substantially computation times in the policy evaluation and improvement routines of the PIM.