Improved vehicle based multibeam bathymetry using sub-maps and SLAM

This paper presents an algorithm to improve sub-sea acoustic multibeam bottom mapping based on the simultaneous mapping and localization (SLAM) methodology. Multibeam bathymetry from underwater water vehicles can yield valuable large scale terrain maps of the sea door, but the overall accuracy of these maps is typically limited by the accuracy of the vehicle position estimates. The solution presented here uses small bathymetric patches created over short time scales in a sub-mapping context. These patches are registered with respect to one another and assembled in a single coordinate frame to produce a more accurate terrain estimate and provide improved renavigation of the vehicle trajectory. The mapping is implemented using a delayed state extended Kalman filter (EKF) and results are shown for a real world multibeam data set collected at the mid-Atlantic ridge using the JASON ROV.