Expert Systems with Applications vol:62 pages:302-316
Automatically generating train to platform assignments has been an active research area for some time, but systems simplementing this research are still not readily available to practitioners. However, now, our train platforming model has been implemented as the tool Leopard inside Infrabel, the Belgian railway infrastructure manager.
In practice, initial macroscopic timetables are often not yet feasible inside stations on the microscopic level. This means that a platforming tool must be able to handle cases where not all trains can be platformed or routed. Our model provides a platforming and routing plan for as many trains as possible and puts the remaining trains on a fictive platform. Contrary to the manually made platforming plans, the optimised platforming plans have no platform conflicts nor routing conflicts. Our model assigns as much trains as possible, given the timetable and the available infrastructure. Our tool can solve the platforming problems for all 530 stations in Belgium together in about 10 minutes. This means (i) it saves many man months of planning time compared to the still common manual practice to platforming and (ii) it achieves higher quality results leading to significantly less in-station train delays in practice.