A multi-criteria decision support methodology for real-time train scheduling

This work addresses the real-time optimization of train scheduling decisions at a complex railway network during congested traffic situations. The problem of effectively managing train operations is particularly challenging, since it is necessary to incorporate the safety regulations into the optimization model and to consider key performance indicators. This paper deals with the development of a multi-criteria decision support methodology to help dispatchers in taking more informed decisions when dealing with real-time disturbances. Optimal train scheduling solutions are computed with high level precision in the modeling of the safety regulations and with consideration of state-of-the-art performance indicators. Mixed-integer linear programming formulations are proposed and solved via a commercial solver. For each problem instance, an iterative method is proposed to establish an efficient-inefficient classification of the best solutions provided by the formulations via a well-established non-parametric benchmarking technique: data envelopment analysis. Based on this classification, inefficient formulations are improved by the generation of additional linear constraints. Computational experiments are performed for practical-size instances from a Dutch railway network with mixed traffic and several disturbances. The method converges after a limited number of iterations, and returns a set of efficient solutions and the relative formulations.