Constrained Zonotope Terrain-Aided Navigation Method for Long-Range Autonomous Underwater Vehicles

The single beam echo sounder (SBES) terrain-aided navigation (TAN) method, using publicly available seabed terrain maps, facilitates long-range autonomous underwater vehicle (AUV) navigation. However, the particle filter TAN method often requires knowledge of the noise distribution, which is difficult due to low resolution and confidence maps. In addition, limited SBES measurement data hinders effective navigation evaluation. This article proposes a constrained zonotope TAN (CZTAN) method for long-term underwater missions and a strategy for evaluating navigation results. Constrained zonotopes are used to estimate the unknown distribution of noise, which does not require prior knowledge of the noise distribution. Bathymetric data from gridded and contour maps enhance the accuracy of the constrained zonotope. Twin support vector regression is employed to establish the relationship between real-time information and navigation results for navigation evaluation. The CZTAN method and evaluation strategy were validated using SBES datasets and a shipborne SBES experiment. Compared to state-of-the-art methods, the CZTAN method reduced the maximum and end-mission positioning errors by 5.55% and 19.22%, respectively, in the shipborne experiment, while also decreasing computational consumption by 19.61%. In the SBES dataset, the average positioning error was reduced by 21.80%. The results show that the CZTAN method achieves accurate and robust navigation results.