Stochastic modeling for airlift mobility

Abstract

This thesis is a continuation of optimization modeling research conducted at the Naval Postgraduate School for the U.S. Air Force Studies and Analyses Agency. That work resulted in Throughput II, a multi-period model for determining the maximum on-time throughput of cargo and passengers that can be transported with a given fleet over a given network, subject to appropriate physical and policy constraints. This and other existing deterministic strategic airlift models assume all data are known prior to making a decision; often times these assumptions are unrealistic. One such assumption is aircraft reliability. This thesis addresses the uncertainty of aircraft reliability, which, if ignored, can result in models that are overly optimistic with respect to throughput capability. To address this issue, this thesis adds a stochastic extension to Throughput II, resulting in a two-stage stochastic linear program with recourse that is solved using Benders decomposition. To analyze the stochastic program, a simulation model of the strategic airlift system is also developed. This simulation model allows the user to analyze the deterministic and stochastic models and to compare solutions. The stochastic model, in addition to the features of Throughput II, accomplishes the following: (1) selection of aircraft routes by anticipating potential bottlenecks in the system, (2) prevents unreliable aircraft from using capacity limited airfields and (3) a flow of cargo from origin to destination that is not interrupted by the random events of aircraft reliability.