Terrain-Aided Navigation of Long-Range AUV Based on Cubature Particle Filter

Terrain-aided navigation (TAN) is a promising technology for achieving accurate positioning of long-range autonomous underwater vehicles (AUVs). In TAN systems, particle filters (PFs) are widely used to estimate the AUV's location. However, PFs suffer from inefficient particle resampling and the assumption of constant particle process noise, both of which reduce the AUV positioning accuracy. In this article, we propose an improved TAN method based on cubature PFs (CPFs). Specifically, we design an adaptive particle process noise that varies the particle search range according to positioning errors. Additionally, a CPF-based resampling mechanism is introduced, in which particles are forced into the surrounding region of high-weight particles, rather than a single point. Simulations performed on actual bathymetric maps demonstrate that our proposed CPF-based TAN can improve the average positioning accuracy by 4.2%, 9.4%, 12.7%, 15.8%, and 17.2%, respectively, compared with the TANs based on neural network PF (NNPF), auxiliary PF (APF), data-driven PF (DDPF), fuzzy PF (FPF), and PF, which verifies the effectiveness of our proposed method.