Underwater sound speed is one of the most significant factors that affects high-accuracy underwater acoustic positioning and navigation. Due to its complex temporal variation, the forecasting of the underwater sound speed field (SSF) becomes a challenging task. Taking advantage of machine learning methods, we propose a new method for SSF forecasting based on the least square support vector machine (LSSVM) and a multi-parameter model, aiming to enhance the forecasting accuracy of underwater SSF with hourly resolution. We first use a matching extension method to standardize profile data and train the LSSVM with the parameters of observation time, temperature, salinity, and depth. We then employ radial basis function kernels to construct the forecasting model of SSF. We validate the feasibility and effectiveness of the LSSVM model by comparing it with the polynomial fitting (PF) and back propagation neural network (BPNN) methods, using hourly data obtained from the measured data and open data. The results show that the means of the root mean square for the LSSVM based on the observation time parameter and the LSSVM based on the multi-parameter model achieve 0.51 m/s and 0.45 m/s, respectively, presenting a significant improvement compared with the PF (0.82 m/s) and BPNN (0.76 m/s) methods.
The geometric distribution of seabed beacons significantly impacts the positioning accuracy of underwater acoustic navigation systems. To address this challenge, we propose a depth-constrained adaptive stochastic model optimization method based on singular value decomposition (SVD). The method quantifies the contribution weights of each beacon to the dominant navigation direction by performing SVD on the acoustic observation matrix. The acoustic ranging covariance matrix can be dynamically adjusted based on these weights to suppress error propagation. At the same time, the prior depth with centimeter-level accuracy provided by the pressure sensor is used to establish strong constraints in the vertical direction. The experimental results demonstrate that the depth-constrained adaptive stochastic model optimization method reduces three-dimensional RMS errors by 66.65% (300 m depth) and 77.25% (2000 m depth) compared to conventional equal-weight models. Notably, the depth constraint alone achieves 95% vertical error suppression, while combined SVD optimization further enhances horizontal accuracy by 34.2–53.5%. These findings validate that coupling depth constraints with stochastic optimization effectively improves navigation accuracy in complex underwater environments.
Model-aided navigation is a key approach for enhancing the positioning accuracy of autonomous underwater vehicles (AUVs). However, its precision is often degraded by model-based velocity errors arising from attitude-induced deviations and uncertainties in the mapping between propeller rotational speed and vehicle velocity. To overcome these limitations, this study proposes an attitude-compensated and acoustics-calibrated model-aided navigation framework for AUVs. The framework derives the vertical velocity from pressure sensor depth data to correct attitude-related model errors. It also dynamically refines the mapping between propeller speed and velocity using long-baseline (LBL) acoustic positioning data when LBL measurements are available. A sea trial was conducted in the South China Sea at a depth of 2000 m to verify the proposed method. The results showed that the system maintained a positional accuracy of 509 m over 5 h beyond LBL coverage. This outcome demonstrates its ability to achieve sustained high-precision navigation without external assistance.
Rapid and accurate estimation of sound speed profile (SSP) is of great importance for underwater precise positioning and navigation, as the propagation mode of sound signals is influenced by the distribution of sound speed. Among the existing methods for estimating SSPs, artificial neural networks (ANN) have better real-time advantages compared to matched field processing (MFP). However, under few-shot learning situations, they are prone to be overfitting and the inversion accuracy decreases. To achieve accurate SSP inversion in few-shot learning, we propose a spatio-temporal information driven meta learning (ST–ML) method. By learning different types of SSP distribution, common features are extracted, thus accelerating the learning process on given tasks, and reducing the demand for reference samples, so as to improve the accuracy of SSP inversion in few-shot learning situation. To verify the feasibility and effectiveness of ST–ML, a deep–ocean experiment was held in April 2023. Results show that ST–ML outperforms the state–of–the–art methods in terms of accuracy for SSP inversion, while inherits the real-time advantage of ANN during the inversion stage.
1 January 2026 in IEEE Transactions on Instrumentation and Measurement