AUV terrain-aided navigation using a Doppler velocity logger

This paper addresses the design and implementation of terrain-aided navigation (TAN) methods for small autonomous underwater vehicles (AUVs) that rely on standard navigation sensors and dispense with the need for dedicated sensors for terrain data acquisition. The research described focuses on the problem of TAN implementation in underwater scenarios characterized by smooth sea-bottom topography and very shallow water, where the terrain information available for navigation is scarce. The navigation algorithms and the data fusion methods whose tests are documented in the paper build upon and expand prior theoretical work published by the authors; the TAN solutions adopted exploit the terrain information and the navigation data acquired with an inexpensive Doppler velocity logger (DVL) and a standard motion reference unit, respectively. The position estimation methods analyzed include a bi-dimensional particle filter (PF) and a four-dimensional Rao-Blackwellized PF that was designed to estimate the unknown Doppler velocity measurement biases responsible for the unbound localization errors typically observed in dead-recknoning navigation. The positioning accuracy achieved with these filters is compared with the output of a novel method, also proposed in the paper, that mechanizes a complementary-like filter designed to fuse the output of a TAN estimator with the velocity measurements provided by a DVL. Experimental results obtained during field tests with an autonomous marine vehicle are reported and analyzed.