Optimization of Train Departure Times at an Interchange Considering Passenger Flows

The authors of this paper present the results of the numerical optimization of departure times of trains at an interchange station when services are disrupted. The evaluation function is based on accumulated, or aggregated, passenger delays and the dynamic model of the interchange accounts for variations in passenger walking times. The authors have developed a passenger flow model using petri nets, which enables the representation of passenger flows inside a station to be made with sufficient accuracy and short calculation times. The evaluation function of aggregated passenger times is defined, and the optimization is carried out using the multi-dimensional downhill simplex method. The results show that the optimized train departure times essentially represent compromises between the benefit generated by the extended holding of trains to allow more passengers to catch an earlier service and the disbenefit generated by longer waiting times of passengers already on the train being held. This paper concludes with recommendations for combining these techniques with other models and evaluation criteria in order to move towards a practical train controller. Examples of refinements are a model of passenger behaviour when guidance information is provided, variations in dwell times when different numbers of passengers get on or off a train, congestion inside trains, and the impact of holding a train on the overall operation of trains on the line.