Resilient Train Localization Based on GNSS/INS/Trackmap Integration Using a MAP-AICKF Method

GNSS-base train localization is confronted with rigorous safety level requirements, and merged with railway scenario features of polytropic operation environments and constrained trackmap. Integrating GNSS with extra assistant sensors is highly required for the effectiveness of GNSS-enabled train localization under the signal constrained or even the GNSS interference environments. Under an integrated localization architecture, how to address the nonlinearity of the system is a significant issue for precision and performance stability. The Bayesian filter provides a unified recursive framework for the data fusion which approximates the posterior density based on the Linear Minimum Mean Square Error (LMMSE). However, the filter may suffer from a failure under large prior errors. The probability distribution of the measurement noise is required as a priori knowledge to be filtered out which is usually modeled as an additive Gaussian random variable with an invariant variance. Nevertheless, with the potential GNSS signal interference from a deliberate attack, the priori probability cannot describe the likelihood probability correctly. The approximation process of the nonlinearity and the noise distribution of the fusion system is important to guarantee the reliability and resilience of GNSS-based train localization. An Adaptive Iterated Cubature Kalman Filter (AICKF) based on Maximum A Posteriori (MAP) is applied for the integrated train localization to cope with the vulnerability of GNSS under the signal interference scenarios. The solution is capable of realizing autonomous train localization with the enhanced nonlinearity approximation capability. It allows us to deal with specific GNSS interference situations where the probability density of the pseudorange noise for different visible satellites is dynamically adjusted. The proposed solution is tested based on a practical operational scenario on Qinghai-Tibet Railway. GNSS interference injection (e.g., amplitude modulation jamming) is carried out to build simulative scenarios and illustrate the performance of the proposed solution. The numerical results demonstrate the interference tolerance capability of the proposed solution with an adaptive nonlinear filtering framework, which is of great significance in trustworthy localization for GNSS-based railway train control systems.