A Whitening-based Tracking Algorithm for Autonomous Underwater Vehicles

This paper presents a tracking algorithm for Autonomous Underwater Vehicles (AUVs) using a whitening approach. Localization of underwater vehicles faces challenges that span from the spatiotemporal aspects of the properties of the water to unavailability of GPS signal. To overcome some of these problems, bathymetric information has been used to estimate the state trajectory of AUVs but it has limitations in regions of little to no vertical relief. To address this issue, we developed a new method for localizing the state trajectory of an AUV by considering not only the bathymetric information but also sensor data e.g., dissolved oxygen concentration, temperature, and turbidity. Nonetheless, the correlation of these parameters masks the true distances between data points. Therefore, we decorrelate the sensed data using a zero-phase components analysis, also known as Mahalanobis whitening; thus, revealing the true distances between data-points and finally being able to localize with higher precision. Results are shown for deployments carried out on two consecutive days in Big Fisherman's Cove on Santa Catalina Island, CA, USA. Tests were performed for each day where a new list of sensor data observations (without GPS information) is localized based on the depth, turbidity and dissolved oxygen values of the historical data. Using GPS information as a measurement of the errors in localization, a median of 4.79 meters was obtained when tracking was performed on correlated data and a median of 1.16 meters was obtained when tracking was performed on the whitened data. Results indicated that ZCA Mahalanobis whitening is a promising methodology to aid navigation of AUVs in GPS-denied environments.