Unscented Kalman Filtering on Manifolds for AUV Navigation -- Experimental Results

TL;DR

This paper introduces a real-time aided inertial navigation system for AUVs using an unscented Kalman filter on manifolds, effectively integrating multiple sensors and handling outliers, validated through field experiments.

Contribution

It develops a UKF-M-based navigation system for AUVs that explicitly models sensor offsets and includes outlier rejection, with real-world validation.

Findings

UKF-M converges to correct heading despite large initial errors

Sensor model includes compensation for lever arm offsets

System validated on Virginia Tech 690 AUV in field tests

Abstract

In this work, we present an aided inertial navigation system for an autonomous underwater vehicle (AUV) using an unscented Kalman filter on manifolds (UKF-M). The inertial navigation estimate is aided by a Doppler velocity log (DVL), depth sensor, acoustic range and, while on the surface, GPS. The sensor model for each navigation sensor on the AUV is explicitly described, including compensation for lever arm offsets between the IMU and each sensor. Additionally, an outlier rejection step is proposed to reject measurement outliers that would degrade navigation performance. The UKF-M for AUV navigation is implemented for real-time navigation on the Virginia Tech 690 AUV and validated in the field. Finally, by post-processing the navigation sensor data, we show experimentally that the UKF-M is able to converge to the correct heading in the presence of arbitrarily large initial heading error.