Robust Adaptive SINS/DVL Initial Alignment Method Based on Variational Bayesian Information Filter

Strapdown inertial navigation system (SINS) is the core navigation sensor of the autonomous underwater vehicle (AUV), and its underwater dynamic initial alignment is a hot and difficult problem in current research. Doppler velocity log (DVL), as a commonly used underwater velocity sensor, can provide external observation information for SINS dynamic initial alignment. In complex underwater environment, DVL measurement is susceptible to non-Gaussian noise pollution, such as outlier. As a result, the prior information of the observation could be inaccurate or unknown. Then, the initial alignment performance based on Kalman filter (KF) could be degraded or even diverged. To solve this problem, a variational Bayesian-based robust adaptive information filter (VBRAIKF) alignment method is proposed in this article. First, the adaptive updating strategy of measurement noise covariance matrix is designed based on the variational Bayesian method, and the variational Bayesian-based adaptive information filter (VBAIKF) method is proposed. Second, Mahalanobis distance algorithm is used to enhance the robustness of the alignment process. Thus, the VBRAIKF is proposed. Finally, the initial alignment performance of the navigation sensor SINS is significantly improved. Based on the on-board test data, the initial alignment tests of SINS were carried out using the traditional information KF (IKF), VBAIKF, and VBRAIKF, respectively, under the conditions of mixed Gaussian noise and outlier pollution. The results show that, compared with IKF and VBAIKF, the proposed VBRAIKF method can estimate the measurement noise covariance matrix adaptively under the condition of inaccurate observation prior information and different types of non-Gaussian noise pollution. Moreover, it has stronger alignment robustness and higher alignment accuracy.