CORMSIS Seminar Event

Transshipments are a means to redistribute inventory in a network of customer-facing locations in order to better meet customer demand.  Models of transshipments generally assume either a reactive (i.e. to meet existing shortages) or proactive (i.e. to meet anticipated future shortages) approach.  This research proposes a model that combines reactive and proactive transshipments.  Approximate dynamic programming is applied to develop a heuristic policy for the problem.  For comparison, a lower bound on the expected cost of meeting demand is constructed.  Experiments with test problems show that, for certain problem types, the gap between actual cost and the lower bound is 1.5-3.5% for the proposed heuristic compared to 7-14% for commonly used policies.  This improvement is due to fewer transshipments of larger size.