Rolling-Horizon Heuristics for Capacitated Stochastic Inventory Problems with Forecast Updates

In this paper we propose a practical optimization approach based on the rolling-horizon paradigm to address general single-product periodic-review inventory control problems. Our framework supports many constraints and requirements that are found in real inventory problems and does not rely on any assumption on the statistical distribution of random variables. Ambiguous demand and costs, forecast updates, constant lead time, lost sales, flexible inventory capacity and product availability can all be taken into account. Three, increasingly sophisticated, solution methods are proposed and implemented within our optimization framework: a myopic policy, a linear programming model with risk penalization and a scenario-based stochastic programming model. The effectiveness of our approach is proved using a dataset of realistic instances.