Combinatorial optimization for integrating rolling stock management and railway trac scheduling in passenger stations
- Authors
- Publication Date
- Nov 20, 2020
- Source
- Hal-Diderot
- Keywords
- Language
- English
- License
- Unknown
- External links
Abstract
Railway stations that concentrate starts and ends of train journeys structure most of the passenger lines operations. Indeed, rolling stock preparation operations (cleaning, trains coupling...) which are called shunting are scheduled there. These operations are essential to ensure service quality. These operations require train movement and parking. The thesis tackles an integration of shunting operation planning and capacity management in railway stations. The Generalized Train Unit Shunting Problem (G-TUSP) is introduced to consider this integration. In the G-TUSP we assign trains which arrive in a railway station to departures and parking tracks and we schedule their maintenance operations and their movements. These decisions are made to minimize departure delays, coupling and uncoupling operations and maintenance or departure cancellations. The G-TUSP has constraints due to rolling stock and infrastructure characteristics or related the nature of the operations carried out. The G-TUSP includes four sub-problems, often considered independently in literature. The thesis proposes optimization algorithms as decision support tool for shunting planners. A mixed integer linear programming formulation which considers a detailed representation of G-TUSP aspects is set. The formulation is tested on instances representing traffic at Metz-Ville station. Relevant results are obtained within an hour of computation. Then, we propose algorithms in which we consider different combinations for the integrated or sequential solutions of the G-TUSP sub-problems. In a thorough experimental analysis, based on Metz-Ville station instances, we study the contribution of each sub-problem to the difficulty of the G-TUSP, and we identify the best algorithm. This algorithm returns very satisfying results in less than 20 minutes.
