Analytical and Reasoning Methods for Fault Detection and Fault Tolerant in GNSS/INS-Based Train Multi-Sensor Fusion Localization
Debiao Lu, Meng Yang, Baigen Cai, jian Zi Wang, Jiang Liu, Wei Jiang, Xiaohui Ba · Proceedings of the Satellite Division's International Technical Meeting (Online)/Proceedings of the Satellite Division's International Technical Meeting (CD-ROM) · 2025
While the Next Generation Train Control (NGTC) is driven by the need for enhanced safety, transportation efficiency, and interoperability in rail transportation. The key trend for the NGTC is modular architecture of the trainborne and safety functions. The train localization methods are the key input for the train operation and control process. The Global Navigation Satellite System provides continuous location when the train moves dynamically along the railway track, but the safety aspects of GNSS is always a question to be answered. Traditionally the GNSS safety is evaluated using the terminology as integrity. Integrity demonstrates the trustworthiness of the information provided by GNSS; the protection limit (PL) and integrity risk (IR) are the parameters for the concept of integrity. Since the train is moving in various environments (as viaduct, train station with ceiling, train station without ceiling, bridge, tunnel etc.), which causes the GNSS performance being degraded and the GNSS integrity cannot be determined. Thus, the GNSS and the INS sensor fusion architecture is proposed in many application scenarios to eliminate the GNSS shortage. This research introduces the GNSS/INS tightly coupled positioning algorithm based on the extended Kalman filter, the state model and the measurement model of EKF is constructed and the GNSS/INS measurement faults are analyzed using both analytical and reasoning methods. The data collected in the Tibet-Qinghai railway line is studied with fault injection and error estimation of the in-field data measurements. The protection level of the GNSS/INS fusion is studied first to detect the faults in the measurement data; the discrete wavelet transform for multi-sensor fault detection is studied.