10 March 2026 in IEEE Transactions on Instrumentation and Measurement
Accurate alignment between the Doppler Velocity Log (DVL) and the Strapdown Inertial Navigation System (SINS) is essential for reliable Autonomous Underwater Vehicles (AUVs) navigation, especially in dynamic and GNSS-denied environments. Conventional DVL calibration methods often suffer from local optima and limited robustness, leading to inaccurate estimation of scale factors and misalignment angles. While optimized calibration parameters can improve the performance of traditional filters such as EKF or UKF, their effectiveness remains sensitive to environmental disturbances, outliers, and non-Gaussian noise. To address these challenges, this paper proposes a unified calibration and navigation framework. A Simulated Annealing Adaptive Particle Swarm Optimization (SA-APSO) algorithm is developed to globally and robustly estimate DVL error parameters. These parameters are then used in conjunction with a Modified Residual Adaptive Unscented Kalman Filter (MRUKF) with M-estimation, which enhances resilience to measurement anomalies and dynamic noise. Unlike conventional filters, MRUKF adaptively adjusts residual weighting, allowing it to better exploit calibration benefits under non-ideal conditions. Simulations conducted on four representative AUV trajectories demonstrate the superiority of the proposed framework. Compared with EKF and UKF, endpoint errors are reduced by up to 50.2% and 54.1%, respectively, while overall position RMSE decreases by 30.6% and 29.6%. These results confirm that combining accurate calibration with robust filtering is essential for high-precision navigation in realistic, uncertain underwater environments.