Long-Range Gravity-Aided Autonomous Underwater Vehicle Navigation

Abstract

Autonomous underwater vehicles (AUV) are a mobile platform for underwater sensing, an environment relatively unexplored. Georeferencing measurements is difficult due to the challenge of AUV localization. The rapid attenuation of radio frequencies underwater restricts AUVs from using the global position system (GPS), the above-water solution to localization. Underwater localization relies on dead-reckoning, the integration of vehicle inertia measurements to arrive at a position estimate. However, the dead-reckoned position error is unbounded. This error can be bounded using a source of position feedback. Terrain aided navigation (TAN) - using georeferenced geophysical terrain maps can provide that feedback. TAN shows significant promise as a method for long-range, passive underwater AUV navigation, especially gravity-aided navigation (GAN). This thesis presents a TAN algorithm that uses a gravity gradiometer and gravity gradient maps to successfully limit dead-reckoning error by a factor of 25 over a 500 km long AUV mission, with a localization accuracy of 1 km. The TAN algorithm exploits the correlation between terrain and the gravity anomaly to use a global database of bathymetry maps (GEBCO) with 400 m resolution. The mission was simulated in the AUV navigation testbed (ANT), a collection of tooling developed during this thesis to accelerate research in TAN. Among the contributions made by the ANT, is a inertial navigation system (INS) that emulates the uncertainty characteristics of a commercial navigation grade INS (Kearfott Seanav) \textemdash~to simulate dead-reckoning error growth. Parts of the ANT have been released to the research community as open-source, and are being used by researchers in the Intelligent Systems Laboratory (ISL) at Dalhousie University.