Models and algorithms for dynamic real-time freight train re-routing

This dissertation focuses on solving a train re-routing problem in a near real-time context for a freight train carrier operating over a large network. A holistic evaluation framework is developed using a time-space network model. Computational results using data from a class I railroad in the United States are used to determine the subset of problem instances that can be solved using the evaluation framework by systematically evaluating all solutions. Solving the remaining problem instances is addressed by developing two solution methodologies that leverage the evaluation framework: an optimization-based approach and a search-based heuristic approach. A further problem variant is also introduced where rail terminal processing rate is a non-constant function of the traffic at the rail terminal. The above approaches are extended to address the problem variant. Computational results are presented for a comprehensive set of problem instances created using data from a class I railroad in the United States. Results indicate the tractability of the optimization-based approach for large-scale instances, practical solution time with reasonable compute resources, as well as the robustness of solution quality to increases in the number of candidate trains. The solution time of the search-based heuristic approach is furthermore shown to be robust to increases in network traffic volume. The results are discussed in detail, along with implications for future research.Ph.D