An improved variational Bayesian adaptive robust filtering algorithm for LBL/SINS integrated navigation in complex marine environments

In the context of LBL/SINS integrated navigation systems, the Kalman Filter (KF) often encounters difficulties in handling dynamic noise, signal loss, and measurement outliers. To address these challenges, this paper introduces an improved Variational Bayesian Adaptive Robust Filter (VBARKF), which estimates both process and measurement noise statistics. To mitigate the impact of beacon signal loss, the proposed method employs an adaptive beacon validity vector that dynamically adjusts the dimensionality of the measurement equation. Furthermore, to counteract the influence of gross errors, the VBARKF integrates a robust filtering mechanism based on the Mahalanobis distance. Specifically, the Mahalanobis distance between the slant distance derived from LBL and SINS is compared with a predefined threshold, enabling the dynamic adjustment of measurement weights. Experimental results from an offshore field trial demonstrate that the VBARKF improves positioning accuracy by 83.03% compared to the KF. It also outperforms Unscented Kalman Filter (UKF), VBAKF-R, VBARKF-R, robust Student’s t based Kalman filter (RSTKF) and VBAKF by 73.59%, 67.19%, 60.35%, 38.31% and 6.55%, respectively. In addition, underwater semi-physical experiments show an improvement in accuracy using VBARKF, surpassing KF, UKF, VBAKF-R, VBARKF-R, RSTKF, and VBAKF by 49.13%, 45.81%, 44.26%, 42.08%, 36.38%, and 30.11%, respectively. These findings underscore the robustness and reliability of the VBARKF, making it a promising solution for precise underwater navigation in applications such as underwater remote sensing, ocean mapping, and underwater rescue.