AUV Terrain-Aided Doppler Navigation using Complementary Filtering

Abstract

This paper addresses the challenging problem of achieving truly autonomous long-range navigation of underwater vehicles relying on affordable navigation sensors. Terrain-aided navigation (TAN) is a methodology that holds potential to solve this problem by dispensing with the need to use high-grade, inertial navigation sensors and the deployment and calibration of acoustic beacons. To implement TAN without incurring the additional cost of expensive dedicated sensors, we propose the utilization of a Doppler velocity logger (DVL) which is an equipment of widespread utilization in oceanography and also a standard instrument in underwater navigation. We avail ourselves of a less exploited characteristic of DVL sensors that consists in the ability to acquire, simultaneously with the velocity data, a set of accurate range measurements relatively to a reflective interface. The combination of these sensing capacities enables Doppler units to be used not only in dead-recknoning navigation but also in terrain-based localization of underwater vehicles. The main contribution of the paper is the design of a complementary filter (CF) for fusion of TAN estimates with DVL measurements. The CF approach is motivated by the need to reduce the short-term variability of position estimates that is typically observed in conventional TAN. The solution proposed is analysed in comparison with a well-known Rao-Blackwellized particle filter set-up which is shown to implement a data fusion filter designed in a stochastic estimation framework. The results obtained in Monte Carlo simulations performed with real bathymetric data evidence the superior performance of the complementary filter approach in terms of position estimation accuracy and long term output signal stability.