Real-Time Underwater Localization in Three-Dimensional Environment

Abstract

Localization is a critical component to enable autonomous visual inspection of underwater structures by using autonomous underwater vehicles (AUVs). Underwater structures, such as bridge piers, are often vertically nonuniform and have 3-D features, such as inclination and structural overhangs. This article proposes a real-time localization method using a particle filter with a novel 3-D observation model. The acoustic ray casting is introduced into the observation model to estimate multiple echoes of a single fan beam from varying geometries at different depths, so as to consider the vertical beamwidth of the mechanical scanning imaging sonar. The proposed method was implemented based on parallel computation using an embedded computer equipped with a graphics processing unit to enable real-time onboard navigation for the AUV. Photogrammetric survey missions were conducted in an area with large bridge structures of the Tokyo Wan Aqua-Line Expressway using an AUV. The effectiveness of the proposed method was validated by the successful demonstration of its ability to provide precise navigation and guidance in real time for very close distance visual inspection of underwater structures.