Online 3D Underwater Exploration and Coverage

Abstract

This work proposes a novel 3-dimensional (3D) exploration and coverage methodology for autonomous underwater vehicles (AUVs). Exploration can be defined as the task of discovering an unknown region in the space. Most robotic exploration algorithms seek full coverage of the unknown region with data from one sensor (in the underwater domain, it is usually occupancy data from profiling sonars, or optical data from cameras). Our method is able to guide the robot so that full discovery and coverage is obtained in a single exploration for both occupancy data and optical data. Since the algorithm has to run in the AUV's onboard computer, with limited processing power, our method is based around efficient operations that can be performed to an octree data structure. This work extends our previous 2-dimensional (2D) method, which has proved to work in real tests conducted with the Sparus II AUV. In this work, we have used three real datasets to generate environments for its exploration in simulation: an underwater seamount, a shipwreck and an underwater cave. Results using the Girona 500 AUV show the feasibility of our exploration approach in all tested scenarios.