A self-consistent bathymetric mapping algorithm

The achievable accuracy of bathymetric mapping in the deep ocean using robotic systems is most often limited by the available guidance or navigation information used to combine the measured sonar ranges during the map making process. This paper presents an algorithm designed to mitigate the affects of poor ground referenced navigation by applying the principles of map registration and pose filtering commonly used in simultaneous localization and mapping (SLAM) algorithms. The goal of the algorithm is to produce a self-consistent point cloud representation of the bottom terrain with errors that are on a scale similar to the sonar range resolution rather than any direct positioning measurement. The presented algorithm operates causally and utilizes sensor data that are common to instrumented underwater robotic vehicles used for mapping and scientific explorations. Real world results are shown for data taken on several expeditions with the JASON remotely operated vehicle (ROV). Comparisons are made between more standard mapping approaches and the proposed method is shown to significantly improve the map quality and reveal scene information that would have otherwise been obscured due to poor direct navigation information. (c) 2007 Wiley Periodicals, Inc.