INTEGRATING SIMULATION AND OPTIMIZATION FOR INVENTORY CONTROL OF PERISHABLE ITEMS

Ordering perishable items for sale at a retail store is a difficult problem, when demand is subject to predictable and unpredictable variability due, e.g., to seasonality within the week and to significant uncertainty. Ordering fresh produce with a long shelf life is a way to avoid scrapping stuff, while ensuring suitable service level, but this happens at the expense of quality. We consider here ordering rules inspired by classical order-up-to, periodic review policies, which are easily implemented at retail stores. Such rules are known as S policies in the inventory control literature. In our case, several control parameters must be determined in order to account for predictable variability, with the aim of maximizing long-term average profit. Relevant data are the profit margin and the loss due to scrapped items. Due to the complexity of the underlying phenomena, including different patterns of customer behaviour, we search for an optimal setting of the control parameters by integrating a simulator with derivative-free optimization methods such as pattern search, Nalder-Mead simplex search, and genetic algorithms. This yields a remarkably flexible framework which can be adapted to peculiar situations.