Section: New Results

Railway Time Tabling Optimization with CMA-ES and Greedy Heuristics

Participants : François Fages, David Fournier, Thierry Martinez, Sylvain Soliman.

The problem of reducing energy consumption in public transportation has received increasing attention over the last years. Most metros have energy regenerative braking systems, which allow them to produce electric energy when they brake. We study the problem of optimizing the energy consumption of a metro line by modifying the timetable, in order to maximize the actual reuse of the regenerative energy. This is achieved by synchronizing the braking and acceleration phases of the metros, through slight modifications of the stopping times in stations. In an article in preparation, we present a constraint-based model of the electric network of the line, which is used to evaluate the energy consumption at each instant, and to compute a distribution matrix for approximating the potential energy transfers between metros. The optimization of the timetable is then performed by an evolutionary algorithm using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES from Nikolaus Hansen, EPI TAO). On real data, this strategy shows energy savings ranging from 2.38% to 4.54%. Furthermore, these savings are shown to be robust with respect to perturbations of the dwell times.