A highly robust MST-SVDCKF algorithm applicable to integrated navigation

Abstract The combination of Inertial Navigation System (INS) and odometer in mobile device positioning is of great significance, effectively compensating for the limitations of single sensors and enhancing navigation accuracy and system robustness. This paper proposes a navigation algorithm that combines an Inertial Navigation System (INS) and an odometer using a Multi-Model Singular Value Decomposition Cubature Kalman Filter (MST-SVDCKF). By introducing the singular value decomposition (SVD) technology, the algorithm enhances the numerical stability of the filter when processing the covariance matrix. It combines the multi-model strategy to improve the adaptability of the integrated navigation to the data noise and outliers in the dynamic environment. The experimental results indicate that the MST-SVDCKF algorithm introduced in this paper achieves a 79.8% reduction in average position estimation error in the straight segment of the track compared to CKF and a 53.1% reduction compared to SRCKF. In the curved section, the average position estimation error of MST-SVDCKF is reduced by 2.7% relative to CKF and by 56.4% compared to SRCKF.