Autonomous Mapping of Underwater 3-D Structures: From View Planning To Execution

Abstract

Three-dimensional mapping of submerged industrial facilities or underwater sites with complex geometries is still a challenge for nowadays autonomous underwater vehicles (AUVs), which are restricted to more simple survey missions. This letter presents a complete framework to plan and execute autonomous missions to map a region of interest, coping with the particularities of the underwater domain. The proposed solution consists of the following steps: First, if a previous model is not available, a preliminary bathymetry of the area to be covered is gathered and an octree representation is created. Next, using this a priori model, an offline view-planner computes a minimum set of sensor configurations to cover the volume of interest by taking into account the mapping sensor and the vehicle safety constraints. It also computes the free path to be followed by the AUV to traverse all these sensor configurations. Finally, the planned mission is executed while an online simultaneous localization and mapping algorithm prevents the AUV from drifting, ensuring that the planned path is correctly followed. Preliminary results regarding the view planner have been obtained in a shipwreck site with the Girona 500 AUV, while results involving the whole framework have been achieved using the Gazebo simulator and a realistic AUV model equipped with a multibeam sonar mounted on a pan & tilt unit.