Stochastic modeling of airlift operations

Large-scale military deployments require transporting equipment and personnel over long distances in a short time. Planning an efficient airlift system is complicated and several models exist in the literature. Particularly, a study conducted on a deterministic optimization model developed by the Naval Postgraduate School and the RAND Corporation has shown that incorporating stochastic events leads to a degradation of performance. In this paper we investigate the applicability of network approximation methods to take into account randomness in an airlift network. Specifically, we show that approximation methods can model key performance features with sufficient accuracy to permit their use for network improvement, while requiring only a small fraction of the computational work that would have been needed had simulation been used for all of the performance evaluations. Also, we predict that combining simulation and approximation may work substantially better than either one of these alone.