Multi-Sensor Resilient Fusion Estimation for Speed Measurement and Positioning System of Trains Under Cyber Attacks and Physical Faults

This paper investigates the problem of multi-sensor resilient fusion estimation for speed measurement and positioning system of trains under cyber attacks and physical faults. To mitigate the adverse influences from both sensor attacks and faults on estimation performance, a novel distributed resilient fusion estimation method is proposed, where a saturation mechanism with adaptive bounds is subtly embedded into each local estimator to constrain the distorted innovations within a reasonable range under abnormal measurements, and the modified local estimates are then transmitted to the fusion center for generating the fused estimate. The stability condition of the local estimation error dynamics is derived, and the boundedness of the upper bound on the fused estimation errors is proved. Compared with the existing fusion estimation methods that target sparse or stochastic sensor attacks, the proposed method is able to simultaneously resist the non-sparse and ongoing sensor attacks and faults, demonstrating better robustness. The validity of the proposed method is validated via semi-physical simulation experiments.