A Two-Stage Robust Underwater Gravimetry Method Based on Position Observations

Abstract

The accurate underwater gravity measurements are essential for revealing marine geological structures and resource distributions. Most existing methods primarily rely on the strapdown inertial navigation system (SINS) as the core component, heavily depending on underwater sensors, such as depth gauges (DGs) and ultrashort baseline (USBL) systems. However, the USBL data are susceptible to interference, which degrades the accuracy of gravity determination. To address these challenges, we propose a two-stage robust underwater gravimetry method based on positional observations, which deeply integrates data cleaning with gravity solution to jointly suppress USBL outliers and improve the accuracy of gravimetry. In the data cleaning stage, a trend-residual decomposition smoothing (T-RDS) method is used to preprocess USBL horizontal position data, effectively suppressing outlier interference through trend fitting and adaptive residual correction. In the gravity solution stage, a robust forward–backward filtering (RF-BF) algorithm is further introduced to fuse the preprocessed horizontal positions, DG depth observations, and SINS data. The RF-BF method incorporates a robust mechanism based on standardized residuals into forward filtering to dynamically mitigate observation anomalies and is then followed by backward smoothing to achieve optimal state estimation. Validation using sea trial data demonstrates that the proposed method improves the repeat-line internal accuracy from 0.756 to 0.683 mGal under the original favorable sea-trial conditions. To further test robustness, the representative USBL anomalies, including impulsive jumps, sustained offsets, and data losses, are injected into the measured USBL data. Under this anomalous condition, the proposed method improves the internal accuracy from 1.002 to 0.687 mGal, corresponding to a 31.4% improvement over the Hampel + Kalman filtering (KF) baseline. These results show that the proposed approach mitigates unstable USBL observations and provides more reliable underwater gravity measurements under both normal and disturbed observation conditions.