Reactive Robustness and Integrated Approaches for Railway Optimization Problems

Abstract

Planning railway operations is not a simple task as it entails solving multiple interdependent optimization problems. These problems have been subject to study in the literature for the last few decades, and are still profoundly researched. The robustness of a plan or schedule denotes the ability to absorb or withstand unexpected events such as delays. Making robust plans is central in order to maintain a safe and timely railway operation. This thesis focuses on reactive robustness, i.e., the ability to react once a plan is rendered infeasible in operation due to disruptions. In such time-critical situations, new plans must be found quickly. Integration of the different planning problems is also considered in this thesis as these problems are strongly interdependent in many cases. In contrast, finding feasible plans for each problem in isolation can lead to an overall infeasibility, e.g., during a disruption the updated timetable may be impossible to realize due to the lack of rolling stock units at certain positions. It is important to avoid creating problems for later or subsequent planning stages. Several railway problems are studied in this thesis. The main contributions are summarized in individual chapters, some of which are papers that have been submitted to international scientific journals in operations research. The problems have been formulated as optimization problems and solution methods have been proposed to solve them using optimization theory and various solution techniques. In collaboration with industry and academic partners real-life and realistic data has been used to benchmark and test the solution methods. A central actor and theme of the thesis is the rolling stock running on the railway networks. A public timetable is given, and in order to service the departures and passengers a rolling stock schedule (or circulation) is sought that provides the best compromise between operational cost, robustness, contract requirements and passenger satisfaction. In between train services the rolling stock units must be parked in the available depots. As trains cannot overtake each other easily, special attention must be given to avoid conflicting movements. Furthermore, rolling stock units are heavy and consume a considerable amount of energy in operation; with proper optimization tools a significant amount of the energy can be saved. A prompt optimization of individual train journeys helps the driver to drive efficiently and enhances robustness in a realistic (dynamic) environment. Four international scientific prizes have been awarded for distinct parts of the research during the course of this PhD project. The first prize was awarded for work during the \2014 RAS Problem Solving Competition", where a freight yard optimization problem was considered. The second junior (PhD) prize was awared for the work performed in the \ROADEF/EURO Challenge 2014: Trains don't vanish!", where the planning of rolling stock movements at a large station was considered. An honorable mention (and second place) was awarded in recognition for excellent work in the \Discrete Optimization Challenge", where the aim was to minimize energy consumption in timetables. Finally, a second place was awarded in the \2015 RAS Student Paper Award", where a comparison of solution methods for planning shunting yard movements was considered.