Vision Based Localization for Subsea Intervention

Abstract

A vision based underwater localization system using fiducial markers is proposed. The system is implemented as a new control mode in a shared control system for ROVs and tested experimentally in a pool. A high accuracy underwater motion capture system in the pool is used as ground truth for performance assessment of the proposed system. The main objective has been to assess the feasibility for using a vision system and fiducial markers for localization of an underwater vehicle performing automated manipulation at subsea facilities. This research show that it is feasible to perform autonomous manipulation using the proposed system. The assessment of the performance in a set of experiments show that all degrees of freedom have a an approximate Gaussian error distribution with zero mean. The largest position errors are experienced in the Y-direction. This is seem to be caused by inherent coupling between yaw and lateral position in the camera projection of the fiducial marker and errors in the yaw direction. The main contribution of this paper is an experimental performance assessment of a localization method commonly used for terrestrial robots in a marine environment.