A Position Correction Model for AUV Navigation with Sequential Learning-Assisted State Estimation

Abstract

Global Positioning System (GPS) is unavailable in underwater environments, causing many challenges for Autonomous Underwater Vehicle (AUV) navigation. Currently, state estimation techniques such as Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) are typical AUV navigation methods performing multi-sensor data fusion to obtain AUV pose estimations. However, the state estimation-based AUV navigation method will inevitably introduce the state estimation errors, which will affect the AUV positioning accuracy. Therefore, in this paper, we propose a Position Correction Model (PCM) for AUV Navigation based on sequential learning-assisted state estimation. We build a deep neural network that transforms AUV trajectory estimation into a sequential learning problem. The constructed deep network learns the relationship between the state estimation predicted position sequence and the true position to capture the AUV motion trend and reduce the adverse effects of multi-errors on AUV navigation.