A Multi-model Train Positioning Approach Based on Robust Estimation

To address the degradation of navigation and positioning accuracy of trains in tunnels due to Global Positioning System (GPS) signal loss and non-Gaussian noise, we propose a multi-model train positioning method based on robust estimation (RM estimation). By constructing two integrated navigation models we employ the Interacting Multiple Model (IMM) algorithm to adaptively select the appropriate system model and achieve the interactive fusion of each sub-navigation model, thereby improving positioning accuracy during GPS signal loss. Simultaneously, RM estimation is integrated into the Cubature Kalman Filter (CKF) algorithm, resulting in the RMCKF algorithm for data fusion, which effectively suppresses the effect of non-Gaussian noise on positioning accuracy. The effectiveness of the proposed algorithm under conditions of GPS signal loss and the presence of non-Gaussian noise is validated through three simulation sub-experiments. The simulation experiments indicate that the proposed algorithm (IMMRMCKF) not only enhances positioning accuracy significantly compared to traditional methods, but also demonstrates strong robustness in situations involving GPS signal loss and non-Gaussian noise.