A Robust SVD-UKF Algorithm and Its Application in Integrated Navigation Systems
Guiling Zhao,Zihao Jiang,Jinbao Wang,Shuai Gao
DOI: https://doi.org/10.1016/j.asr.2024.10.037
IF: 2.611
2024-10-24
Advances in Space Research
Abstract:To improve the positioning accuracy in Inertial Navigation System/Global Navigation Satellite System (INS/GNSS) integrated navigation and address the issue of measurement information anomalies, an integrated navigation algorithm based on an improved Huber-M estimation with Singular Value Decomposition Unscented Kalman Filter (SVD-UKF) is proposed. During the iterative update process of the prior information matrix in the UKF filtering, the Singular Value Decomposition (SVD) is introduced. This enhancement optimizes the prior information matrix, thereby enhancing numerical stability. The tail function of the Huber-M estimation is replaced with an exponential square loss function to improve the computational accuracy of the Huber-M estimation. The improved Huber-M estimation is then incorporated into the SVD-UKF framework. This approach reconstructs the UKF prior information and measurement information. The experimental results indicate that, compared to the traditional UKF and the Sage-Husa adaptive SVD-UKF algorithm, the improved algorithm enhances velocity accuracy by 68.7% and 31.2%, respectively, and improves position accuracy by 73.5% and 38.6%, respectively. The positioning accuracy and robustness of the INS/GNSS integrated navigation system were significantly improved, particularly under conditions with abnormal measurement information. The system demonstrated enhanced stability and resistance to interference in these scenarios.
geosciences, multidisciplinary,meteorology & atmospheric sciences,astronomy & astrophysics,engineering, aerospace