Interactive Timetabling System With Function of Estimating Train Congestion on Timetable Planned

Abstract

It is important for railway operators to make suitable timetables. On the assumption that passengers' train paths are invariant even if the timetable changes, the timetables are determined according to the various statistical data under the current timetable. In planning the timetable, it is difficult to grasp the change of passengers' paths due to timetable modifications because it is too complex. In this paper, we propose a framework of timetabling with due consideration of the path change and propose a practical timetabling system which can quickly estimate the congestion of each train by using passengers' origin-destination (OD) data collected by automatic ticket checkers. With the system, timetable planners can interactively make and modify timetables by trial and error while confirming the congestion, and finally, they can reach the most preferable one. In order to realize such an interactive system, it is important to develop a fast estimation algorithm of train congestion. We developed a new shortest path search algorithm to determine which trains each passenger gets on. The algorithm devised based on the Dijkstra method has two features. First, the shortest path search from each node to all the destination stations in the composed graph network is executed only once. Second, overlapping the path searches are omitted using stored information of the shortest paths which have already been searched. By applying this algorithm, it took only about 10 s to estimate the train congestion under a timetable of the whole day on an urban commuter line.

<b>Prediction Method of Train Delay Using Deep Learning Technique</b>

“…Since real-time congestion data between stations of each target train were unavailable, we used estimated instead of actual congestion data. Based on passenger origin-destination data (OD data) and actual timetables, we estimated train congestion between stations of each train with the system we built [9] and used this as a substitute for actual data. Concretely, OD data is the number of passengers for each pair of origin-destination stations for each time period, and it is collected automatically through ticket gates.…”

Passenger Behaviour Simulation Inside Train Car Combined With Train Traffic Simulation

“…In case if there is a grid on which more than X shortest paths of alighting passengers pass, the staying passengers on that grid become temporal alighting, because they disturb alighting passengers. In addition, as the shortest paths are changed along with changes of passengers' positions at a certain time, they are recalculated at every simulation clock, by using fast estimation algorithm for the shortest paths [5].…”