Real-time decision-support algorithms to optimize train order at junctions

In this paper, we present a real-time decision support system whose objective is to propose the best order of trains to run at junctions in dense areas, based on passenger punctuality. This question typically arises when two branches of a railway line (say, A and B) coming from different origins in the network merge, and trains coming from these two branches must use the same track after the junction. For each couple of trains on both branches, a choice has to be made on which train should run first after the junction: train coming from A or from B? This choice can have a significant impact on the operating performance, and, to be efficient in terms of passenger quality of service, should take into account a large amount of information on the position of trains currently running before and after the junction, their service, the number of passengers on each origin-destination pair at a given time... A first real-world experiment was carried out during fall 2015 at one of the most highly used junctions in the Parisian area. This experiment was performed using a software prototype relying on a dedicated approach combining real-time data treatment, operations research and simulation algorithms.