Towards Geo-Referenced AUV Navigation Through Fusion of USBL and DVL Measurements

This paper reports progress on an improved navigation system for an Autonomous Underwater Vehicle (AUV) which combines navigation data from a Doppler velocity log (DVL) and an ultra-short baseline (USBL) acoustic tracking system to provide superior three-dimensional position estimates to the AUV. The AUV motion and the USBL observations are highly nonlinear processes which contain non-Gaussian noise. A particle filter is therefore chosen as a suitable data fusion technique, as an alternative to traditional methods based on applying the Kalman filter to linearised models. The particle filter is used to fuse USBL observations with the dead-reckoned DVL measurements to give a geo-referenced position estimate that does not drift. The performance of the data-fusion algorithm is evaluated using data obtained from a recent field deployment