Timetabling optimization of a single railway track line with sensitivity analysis

The paper deals with the timetabling problem of a single-track railway line. To solve the timetabling problem, we propose a three-stage approach combining several optimization criteria. Initially and mainly, the maximum relative travel time (ratio of travel time to minimum possible travel time) is minimized subject to a set of constraints, including departure time, train speed, minimum and maximum dwell time, and headway at track segments and stations. Since this problem has many solutions, the process is repeated for other trains, keeping the relative travel times of the critical train fixed, until all trains have been assigned their optimal relative travel times. In the second stage, the prompt allocation of trains is a secondary objective, and finally, in the third stage, the one minimizing the sum of the station dwell times of all trains, keeping the relative travel times constant, is selected to reduce fuel consumption, as a tertiary objective. To consider the user preferences in the optimization problems, the user preference departure time is used instead of the actual planned departure times. In order to guarantee that the exact or a very good approximate global optimum is attained, an algorithm based on the bisection rule is used. This method allows the computation time to be reduced in at least one order of magnitude for 42 trains. The problem of sensitivity analysis is also discussed, and closed form formulas for the sensitivities in terms of the dual variables are given. Several examples of applications are presented to illustrate the goodness of the proposed method. The results show that an adequate selection of intermediate stations and of the departure times are crucial in the good performance of the line and that inadequate spacings between consecutive trains can block the line. In addition, it is shown that, in order to improve performance, regional trains must be scheduled just ahead of or following the long distance trains, rather than having independent schedules. The sensitivities are shown to be very useful in identifying critical trains, segments, stations, departure times, and headways and in suggesting line infrastructure changes.