The journey towards algorithmically planned trains

摘要

Timetabling an additional service on the tight and densely operated Swiss railway network has been a feat of craftsmanship ever since the first timetable was introduced in 1847. Even though timetable planning has been performed on computers since the end of the 1980s, it has so far remained a largely manual task requiring a great deal of skill from the planners. Now for the first time, smartrail 4.0 has managed to carry out this demanding planning task almost completely automatically with the help of an algorithm (fig. 1). You might think that such complex timetable planning would only be possible nowadays with the aid of computers, but that is not the case at all. The correct train sequences or passes, for example, have to be manually identified and integrated by the timetable planners. The arrival and departure times are generated by the system, but they have to be manually amended to ensure that they provide the connections the railway companies want. Current IT systems primarily provide support for visualisations, journey time calculations and data distribution. The current timetable is rigid with the annual timetable change being almost the only time the published departure times can be altered. There are of course organisational reasons for this (and even with today's fast pace of life, there is still always an "annual timetable"), but this means that amending timetables is neither quick nor straightforward. Worksites, special event trains and maintenance work have to be "woven" into this construct and freight service customers in particular would like significantly more flexibility. This requires great planning skill from timetable planners and the required coordination creates a heavy workload.