Comparing Robust Cost Functions for Bathymetric Point Cloud Registration

Scan matching is an established method of identifying loop closure in bathymetric SLAM. Newly developed subsea optical scanners offer much greater resolution compared to imaging sonar, but have a limited field of view. This naturally leads to higher outlier rates when performing scan registration, decreasing the success rate of traditional iterative closest point (ICP) algorithms. This work evaluates the effectiveness of different robust cost functions for outlier rejection on field data collected using an AUV-mounted optical scanner.