Robuust optimalisatie voor lotsizing.

Traditional decision support for inventory and production management of products with a time-varying demand pattern has typically focused on solving deterministic lot sizing problems, assuming demand to be fully known during the planning process, whereas stochastic inventory management models have also been around for multiple decades but have usually studied settings where the average demand remains approximately constant with time. Variants of stochastic inventory policies have been devised also for situations where the demand is nonstationary, for instance because the average demand rate may vary appreciably with time, but exact analysis turns out to be very difficult here. Our goal is to examine how to proceed with inventory and production planning with time-varying (nonstationary) demand patterns when demand is uncertain. Forecasting tools are available for estimating the demand distribution, but these techniques are typically developed for stationary demand or for demand following very particular patterns (e.g. with trend, seasonality, etc.). Also, in many cases, there is a large gap between the estimated and the actual demand distribution. We will use techniques from robust optimization, which do not require detailed distributional information. Our goal is the “robustification” of lot sizing problems, which seems to offer a promising approach to production and inventory control with time-varying uncertain demand.