A Stochastic Reachable Set Trajectory Prediction Modeling Method for Following Operation of Urban Rail Trains

Trajectory prediction of the preceding train is a vital problem in the following operation scenario of urban rail train. However, due to the existence of sensor error and communication delay, uncertainty of trains makes it great challenge to predict train trajectory accurately. To overcome the problem, a stochastic reachable set modeling method is proposed to predict the motion trajectory associated with additional probabilistic information of the preceding train, such as position and speed etc. The main idea is to achieve a complete representation of the uncertain behavior of the train, and to improve the computational efficiency by abstracting stochastic reachable sets into extended Markov chains. Two typical validation cases demonstrate the completeness of the proposed modeling method in representing train following behaviors, thus indicating its credibility in trajectory prediction.