AUV robust bathymetric simultaneous localization and mapping

Bathymetric simultaneous localization and mapping (BSLAM) technique could provide long-term underwater navigation results for autonomous underwater vehicles (AUVs) and produce a self-consistent bathymetric map simultaneously. However, the inter-frame motion inside BSLAM is still difficult to estimate, and BSLAM might fail catastrophically with invalid loop closures caused by the measurement errors of vehicle states and bathymetric data. To deal with these problems, an AUV robust BSLAM algorithm is proposed based on graph SLAM. In this algorithm, weak data association is constructed via sparse pseudo-input Gaussian process (SPGP) regression to predict inter-frame motion, and a multi-window consistency method (MCM) is introduced to identify invalid loop closures. Various simulation experiments are conducted under different environments. Comparisons are made between more standard approaches, and our proposed algorithm is shown to be viable, accurate, and could robustly handle invalid loop closures.