BDS for Train Localization Performance Forecast in Railway Environments Using Machine Learning

Debiao Lu, Haoyu Wang, Baigen Cai, Jian Wang, Jiang Liu, Guohao Zhang, and Li-Ta Hsu

Peer Reviewed

Abstract: Railway environments are typically formed by open area, tunnels, railway stations, urban canyons, viaducts etc. Open area is the most ideal environment for GNSS signal reception. The other environments may cause GNSS signal propagation effects as multipath, and signal blockage, which delivers degraded GNSS performance which needs to be mitigated or isolated for train control purposes. To deliver safe train localization for train control purposes, the GNSS localization performance needs to be forecasted before the train puts into operation following the given timetable. Thus, this paper investigates the BDS and other satellite constellation performance in the railway environments especially the signal degradation environments as long-sequence-time-series-forecasting (LSTF) and short-sequence-time-series-forecasting (SSTF) for accuracy and availability performance study, the LSTM/informer for time-series forecasting are applied for data collected in 2 different railway lines. The results delivered both LSTM and Informer model for accuracy forecast, which set up forecast model structure and data training procedures, the vertical-track-error (VTE) is analyzed using the digital track map (DTM) as reference and along-track-error (ATE) is also studied using the reference system, the plausibility of the forecast method is validated using the selected environmental scenarios. Using this method, the GNSS performance can be forecasted for a specific train timetable before it is put into operation, which gives a solution for time-table-relevant-GNSShealthy in railway safety-relevant applications.
Published in: Proceedings of the ION 2024 Pacific PNT Meeting
April 15 - 18, 2024
Hilton Waikiki Beach
Honolulu, Hawaii
Pages: 206 - 217
Cite this article: Lu, Debiao, Wang, Haoyu, Cai, Baigen, Wang, Jian, Liu, Jiang, Zhang, Guohao, Hsu, Li-Ta, "BDS for Train Localization Performance Forecast in Railway Environments Using Machine Learning," Proceedings of the ION 2024 Pacific PNT Meeting, Honolulu, Hawaii, April 2024, pp. 206-217. https://doi.org/10.33012/2024.19652
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In