Modeling AUV Localization Error in a Long Baseline Acoustic Positioning System

Abstract

This paper develops a linearized Bayesian inversion algorithm for high-precision localization of an autonomous underwater vehicle (AUV) on a test range using time difference of arrival acoustic data. The Bayesian approach allows the uncertainties of several factors, including precise receiver locations and bias and lateral variability of the water-column sound-speed profile, to be included in the inversion as unknown parameters with varying levels of prior information. This provides better estimates of these parameters (and hence AUV locations), and includes the effect of their uncertainties in the uncertainty estimates for AUV locations. A modeling study comparing AUV localization uncertainties from the linearized inversion that results from nonlinear Monte Carlo analysis shows that linearization errors are small, and hence linearized analysis is used to efficiently map AUV localization uncertainty as a function of position over the test range. These approaches are applied to model localization uncertainty for an AUV test range associated with the Victoria Experimental Network Under the Sea cabled observatory system, examining the factors mentioned above as well as the effects of water-column refraction and receiver geometry.