Underwater Gravity-Matching Navigation Algorithm Based on Adaptive-Scale Feature Descriptor

Abstract

Gravity matching provides an absolute position reference for correcting the accumulated errors of an inertial navigation system (INS) during long-endurance underwater navigation. Its performance, however, can deteriorate when the available sampling data are limited, the gravity field is weakly distinctive, or the measurements are contaminated by noise. To improve matching accuracy and robustness under these conditions, this paper proposes a gravity-matching navigation method based on an adaptive-scale feature descriptor (ASFD). A coarse-to-fine framework is first established by introducing the feature extraction mechanism of SURF into gravity sequence matching. An adaptive multi-scale descriptor is then constructed to screen candidate positions efficiently. During fine matching, the matching position estimates obtained using phase correlation, random sample consensus, and least squares are integrated through a reliability-aware adaptive fusion strategy. Simulation experiments evaluate the effects of measurement length, measurement accuracy, and regional gravity-field characteristics, followed by validation using three independent shipborne gravity survey trajectories acquired in different matching areas with two marine gravimeters. Across the three trajectories, the ASFD achieves APEs of 0.80–1.13 n miles, representing reductions of approximately 23.1–50.9% relative to the best-performing traditional methods, with MSRs of 89–99% at the 2 n mile threshold. These results indicate that the ASFD maintains relatively high and consistent gravity-matching accuracy under different simulated and measured-data conditions.