Multi-altitude Multi-sensor Fusion Framework for AUV Exploration and Survey

Abstract

In this paper, we propose a path planning framework for underwater exploration and rugosity estimation using Autonomous Underwater Vehicles (AUVs). Rugosity, a measure of variation in the height of a surface, is commonly used to characterize seafloor habitat. The goal of this work is to optimize the survey of an unknown area in order to efficiently estimate its rugosity. To this end, we propose a novel rugosity approximation on 3D voxel grids and a novel framework for using that approximation to adaptively plan AUV paths. The proposed method employs a heterogeneous set of sensors - multi-beam sonar and stereo cameras - whose varied resolution and range make them complimentary for this task. For broad-scale exploration, sonar is used to produce a coarse sense of the area's structure. Fine-scale exploration is completed using the stereo cameras to refine the high-resolution estimate of rugosity. Results display the simulation of two scenarios on real structural data gathered with an AUV and diver held sensor. The first scenario explores the situation where no broad-scale information is available and the robot must explore the terrain optically. The second simulates the two-pass case and demonstrates our method's ability to achieve high accuracy rugosity estimation faster than other survey planning approaches.