A Single Beacon Deep Water Navigation Method Based on the T-Distribution

In deep water environments, the autonomous underwater vehicle (AUV) operates at a longer distance from the seafloor, which means it can only measure water tracking velocity rather than bottom tracking velocity. However, this water tracking velocity is significantly affected by unpredictable ocean currents, posing a major challenge for precise navigation. To address these issues, this study proposes an integrated multi-sensor navigation system based on the inertial navigation system (INS), incorporating a single beacon, doppler velocity log (DVL), and depth meter (DM). The single beacon provides distance information to estimate current flow velocity, enabling compensation for the DVL-measured water tracking velocity. To handle time-varying and non-Gaussian measurement noise, we employ a hierarchical state-space model with t-distribution for measurement noise, and inverse Wishart and gamma distributions for unknown parameters. To comprehensively estimate the state vector, measurement noise, and these unknown parameters, a variational bayesian (VB) method is utilized. This method updates posterior distributions iteratively, optimizing all parame-ters simultaneously. Experimental results based on the field data shows that the proposed algorithm significantly improves the navigation performance compared to conventional method.