The dynamic fleet management problem with uncertain demand and customer chosen service level

In this paper, we study a dynamic fleet management problem with uncertain demands and customer chosen service levels. We first show that the problem can be transformed into a dynamic network with partially dependent random arc capacities, and then develop a structural decomposition approach which decomposes the network recourse problem into a series of tree recourse problems (TRPs). As each TRP can be solved by an efficient algorithm, the decomposition approach can solve the problem very efficiently. We conduct numerical experiments to compare its performance with two alternative methods. Numerical experiments show that the performance of our method is quite encouraging.