Online Data-Driven Adaptive Prediction of Train Event Times

This paper presents a microscopic model for accurate prediction of train event times based on a timed event graph with dynamic arc weights. The process times in the model are dynamically obtained using processed historical track occupation data, thus reflecting all phenomena of railway traffic captured by the train describer systems and preprocessing tools. The graph structure of the model allows applying fast algorithms to compute prediction of event times even for large networks. The accuracy of predictions is increased by incorporating the effects of predicted route conflicts on train running times due to braking and reacceleration. Moreover, the train runs with process times that continuously deviate from their estimates in a certain pattern are detected, and downstream process times are adaptively adjusted to minimize the expected prediction error. The tool has been tested and validated in a real-time environment using train describer log files.