A Terrain-Aided Navigation Method Incorporating Terrain Matching and Terrain Suitability Analysis

Underwater terrain-aided navigation system can effectively correct the drift error of inertial navigation system that gradually increases with the accumulation of time, among which, underwater terrain suitability analysis is an important part of underwater terrain-aided navigation system. Traditional approaches view terrain matching and terrain suitability analysis as separate processes, which leaves terrain suitability analysis devoid of the knowledge used through navigation and terrain matching in particular. Whereas terrain matching can often provide useful information for TSA. This study attempts to use a deep learning model to deal with the terrain matching problem and terrain suitability analysis problem in the dimension of abstract features using a unified feature encoder. We propose a framework for sharing knowledge during the terrain analysis and matching phases, and construct a fitted probabilistic nonlinear classification network for matching. We validate the effectiveness of the method by performing suitability analysis in terrain regions with different geomorphic differences. After that, the good performance of the method is further verified by the navigation results in regions with different suitability. Finally, the feasibility and effectiveness of the terrain-assisted navigation system proposed in this paper are verified and analyzed as a whole in the form of semi-physical simulation.