Intelligent transportation scheduling: Heuristic and sequential optimization of simulated transportation systems

Transportation networks have been modeled in a variety of ways for many years. For large-scale problems, a certain level of aggregation is necessary to try to obtain a good representation of the behavior of individual objects in a closed form set of equations that adequately model the behavior of the system. This approach can give a good general idea of the dynamics of the system and can provide insight into how the system works under different loads. This general system model was used to try to develop a model of the traffic in the St. Louis highway system resulting in a model that incorporated a Kalman filter to attenuate the noise inherent in the process.; An alternative to this closed-form formulation of the equations that govern the system is the use of simulation to represent all of the entities present in the system, as well as any particular behavior by some of the objects. This approach allows a closer inspection of the system and its behavior by allowing the examination of individual objects. The major contribution of this thesis is the use of simulation first to provide a good model of all the behavior of a transportation network, specifically the Strategic Brigade Airdrop operation of the United States Air Force, and more importantly the use of constraint programming to solve sequentially the assignment problems that make up much of the decision making processes of the real world problem. The resulting optimization approach is a sequentially improving solution to the problem where the decisions are based on the current state of information about the system. This approach mitigates the effects of stochasticity on the problem by re-solving the problem many times during the simulation, each time optimizing for current known information about the system. Sequential solutions provide a better approach to an overall optimum than just a single solution does due to the presence of randomness in the system.