DescriptionThis dissertation introduces original analytical methodologies for decision-making in transportation systems. Moving away from the conventional yet burdensome simulation approaches, we advance closed-form solutions that describe transportation-related processes. It contains two parts. Part 1 concentrates on the problem of predicting congestion on roadways, and part 2 focuses on the problem of scheduling inspections for railway track maintenance.
Part 1 provides a faster and more efficient method to determine traffic density behavior for long-term congestion management using minimal statistical information. Applications include road work, road improvements, and route choice. The research adapts and generalizes two models (off-peak and peak hours) for the probability mass function of traffic density on a major highway. It then validates them against real data. The studied corridor experiences randomly occurring service deterioration caused by accidents and inclement weather, such as snow and thunderstorms. We base the models on queuing theory, and we compare the fundamental diagram with the data.
This research supports the validity of the models for each traffic condition under certain assumptions on the distributional properties of the associated random
parameters. Different scenarios demonstrate traffic congestion and traffic breakdown behavior under various deterioration levels. Last, we include a direct expansion of the model for non-space-homogeneous segments. These models, which account for non-recurrent congestion, can improve decision-making with no extensive datasets or time-consuming simulations.
Part 2 considers inspection and maintenance activities in railways. They are essential to preserving railways’ safety and cost-effectiveness. Still, one of the leading causes of derailments, the stochastic nature of railway defect occurrence, is rarely present in the related literature. Defect occurrence has been investigated as a standalone problem by other authors. Even then, models concentrating on defect prediction demand large datasets of obscure parameters that can be costly or infeasible to gather.
We propose a new method that relies on customary data for predicting track and geometry defects. We then develop a holistic approach to scheduling inspection and maintenance activities that integrates the prediction of railway defects into the problem. This integration is robust and allows for different constraints, such as crew limitations via a Multi-Armed-Bandit framework. Results indicate a high accuracy rate in prediction and effective scheduling policies that are adaptable to varying levels of risk tolerance. Finally, we theorize that search games can solve the final decision of where to inspect within the pre-defined segment.