Abstract
Railway companies have to create schedules for their rolling stock to operate a large number of trains by defining which route each train runs on and when maintenance is done. This scheduling task requires a lot of time and highlyskilled operators, called rolling stock managers, because of numerous complications. We developed an algorithm to help them that automatically creates a schedule for rolling stock during disruptions. We applied the Dijkstra method, which is a mathematical algorithm for searching efficient paths from a network model like the routes taken by sales staff. Our algorithm takes into consideration 1) maintenance cycles where railway companies are required by law to periodically inspect the condition of all trains. It schedules maintenance to continue these cycles. It also takes into account 2) operational patterns through scheduling patterns to create schedules that can compete with manual schedules. These patterns can represent experiences by rolling stock managers like the rotations of train routes. We carried out numerical experiments using real data (small and large numbers) at two train depots of a Japanese railway company. The small ones included ten train sets and a bi-monthly schedule. The large ones included 67 train sets and a monthly schedule. We generated experimental data on virtual disruptions. Our algorithm could mostly generate feasible schedules both from the small and large numbers of data, between 0.02 and 173.25 s, and in 632.47 s. Slight deviations in maintenance cycles were involved in some of the large ones. These results indicate that our algorithm could be feasible for real situations, even though satisfaction with constraints and calculation time should be improved to achieve highly interactive rescheduling operations.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have