Hybrid Long/Inverted Ultra-Short Baseline (LBL-iUSBL) Acoustic Pose Estimation for Underwater Sonar Mapping

Abstract

In this article, we present an acoustic system capable of estimating the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">position and orientation</i> of an underwater robot by combining long baseline (LBL) and inverted ultra-short baseline (iUSBL) capabilities in an approach we term <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">hybrid LBL-iUSBL acoustic pose estimation</i> . We propose the use of this system to map complex, man-made underwater structures using multibeam sonar. The magnetic fields exhibited by such structures interfere with inertial measurement unit magnetometers, and thus, prevent the use of these sensors for accurate underwater mapping. Our hybrid LBL-iUSBL system makes use of acoustic range measurements from multiple beacons to estimate robot position, combining this position estimate with beamformed acoustic angle measurements to each beacon to obtain an estimate of robot heading. Acoustic pose is tracked using a particle filter to eliminate outlier measurements caused by undesirable acoustic effects, such as multipath. Particle filter pose estimates are fed into a factor graph framework along with constraints from scan matching of multibeam sonar returns to improve pose accuracy via simultaneous localization and mapping. Initial data from experiments in an 8.7-m diameter outdoor saltwater tank demonstrate promising preliminary results.