Digital Track Map Aided Multi-sensor Fusion for Train Occupancy Identification in Complicated Track Sections

GNSS-based (Global Navigation Satellite System) train positioning techniques have been considered to apply in the next-generation train control system, aiming to improve transportation efficiency and reduce construction & maintenance costs. However, while adopting GNSS positioning techniques to train positioning, as GNSS is vulnerable to the environment, GNSS positioning accuracy usually cannot meet the requirements in complicated track sections in station areas. Track occupancy determination using traditional map-matching algorithm will fail. This paper proposes a track occupancy identification method in railway stations based on GNSS/INS/DTM sensor fusion results is proposed. Firstly, GNSS/INS loosely coupled model is implemented. Secondly, with GNSS/INS sensor fusion result aided with track geography and topology information, probability model based on distance and heading evidence can be implemented. Combined with track topology and train running characteristics, rule sets are constructed. Finally, dynamic Bayesian network is adopted to analyse the casual dependency of variables and recursive Bayesian estimation is applied to fuse GNSS/INS/DTM and prior information. Field experiment data gathered from a highspeed railway line has been analysed to verify the track occupancy identification method. The result shows that track occupancy identification accuracy has been apparently improved, error along the track under complicated track sections scenarios has been greatly reduced. Test result fully implies the effectiveness of the method proposed in this paper.