Robust underwater SLAM fusing bathymetric and range information

Abstract

Bathymetric simultaneous localization and mapping (BSLAM) can accurately locate an autonomous underwater vehicle (AUV) in deep-sea missions, but it cannot provide real-time location results for vehicles prior to revisit due to the lack of overlap between bathymetric measurements. A real-time estimate of the vehicle’s location can not only help the vehicle to follow the given path accurately but reduce the number of invalid bathymetric associations caused by the nearly flat seafloor terrain. In this paper, we proposed a bathymetric and range information fused underwater SLAM (BRSLAM) method that estimates the vehicle’s position in real time by combining both bathymetric data and range measurements. In particular, a probabilistic graph partitioning algorithm (PGPA) was proposed to identify range outliers, and an information-theoretic node reduction method was presented to reduce the computational complexity of BRSLAM. Playback experiments showed that the BRSLAM can provide accurate navigation results with no more than 4 beacons.