Leveraging Past DVL Velocity Measurements for Acceleration-Aided AUV Navigation
arXiv:2308.11762
Abstract
Autonomous underwater vehicles (AUVs) commonly rely on the fusion of an inertial navigation system (INS) with a Doppler velocity log (DVL) through a nonlinear filter to maintain accurate navigation in underwater environments. In this paper, we propose a cost-free, software-only extension to the standard INS/DVL error-state extended Kalman filter (EKF). The proposed method derives an acceleration estimate from a polynomial fit of past DVL velocity measurements and incorporates it as an additional measurement update. We provide a closed-form derivation of the acceleration measurement model and its associated noise covariance. The proposed approach requires no additional hardware, operates at the existing DVL sampling rate, and is implemented through a one-time software modification. We evaluate our approach using real sea-trial data collected with the Snapir AUV across three trajectories with distinct maneuvering profiles. Compared to the baseline EKF, the proposed method achieves up to 50% improvement in position accuracy, 59% in velocity accuracy, and 64% in orientation accuracy, with the percent covariance improvement metric confirming a genuine reduction in estimator uncertainty rather than a mere shift in the point estimate. These results demonstrate that DVL-derived acceleration measurements offer a practical and low-cost means of enhancing underwater navigation performance across diverse AUV operating conditions.