Observation-Differenced Point Mass Filter in Gravity-Aided Inertial Navigation

Gravity-aided inertial navigation systems (GAINS) represent a crucial advancement in underwater navigation, enabling prolonged and highly accurate passive navigation. However, achieving precise positioning in long-term navigation with large inertial navigation system position errors and severe accumulated gravimeter drift poses a significant challenge. To address this issue, this article introduces an observation-differenced point-mass filter (ODPMF). First, a differenced observation model is proposed to eliminate the drift in gravity anomaly measurement in the GAINS framework. Then, the iterative Bayesian regression is proposed based on the observation-differenced system. Each integration in Bayesian regression is approximated by a finite sum over the grid points with point-mass weights, thus, yielding the ODPMF. To speed up calculations while ensuring optimization, a method for adaptively computing the interpolated resolution based on the ratio of map error to local gradient is developed. The results from two marine experiments show that the ODPMF can accurately track the true position despite the low-frequency trend of the gravimeter and large initial position error. Compared with the sequence matching algorithm, the ODPMF has stronger robustness and can achieve real-time navigation, which offers a promising solution for precise positioning in long-term underwater navigation scenarios.