An Application of the Semi Markovian Decision Approach for Train Conflict Resolution on a Vital Dutch Railway Section

This paper shows how the Semi-Markovian Decision (SMD) approach can be used to assist train dispatchers resolve train conflicts in real-time. The approach produces SMD tables which can be used in a similar way as the Trein Afhandelings Document (Dutch for Train Management Document) [TAD] rules used by Prorail. In order to use this approach, the area of interest, line section Utrecht – Gouda was divided into a number of sub-areas. In an extensive simulation, the SMD strategy has been evaluated, and compared to TAD and a number of heuristics. Even though a number of assumptions of the SMD model do not match the real world (Poisson arrivals, maximal train speeds), the SMD strategy clearly out performs the TAD strategy and improves the punctuality of trains with 10 percent points. From the heuristics the authors have looked at some that performed quite well. The drawback of heuristics is that their performance heavily depends on the system configuration. For instance, a change in the number of the freight trains running through the segment affects the performance of such a heuristic. In Section 8 of the paper it has been shown that the changes in performance can be considerable. The SMD approach, however, is dynamic, i.e., it produces different strategies for different system configurations. The computation of the SMD strategy is done off-line. When on-line a conflict is detected and the situation in the conflict area is translated into a state in the SMD model and the corresponding SMD decision is read. This SMD strategy can be formulated in an SMD table which lists intuitive rules that are simple to apply and are easy to interpret. This should contribute to the acceptance of the computed decisions by train dispatchers.