Traffic Modeling and Rescheduling for High-speed Train Based on Block Sections

Abstract

Affected by unexpected events, the nominal operation of high-speed trains will become invalid. To maintain the efficiency of trains, train dispatchers need to reschedule the train timetable, which is a challenging task. On the one hand, the dispatchers need to take into account complex conflicts between trains on the track; on the other hand, the rescheduled timetable should be efficient to reduce operating costs. To address the above issues, this study proposes a traffic modeling method for high-speed trains based on a block section to describe in detail the operation conflicts between trains. A train rescheduling approach combining reinforcement learning and model predictive control is proposed to accomplish train rescheduling efficiently. The experiments show the effectiveness of the proposed method.