The navigation performance of an autonomous underwater vehicle (AUV) as the main tool for exploring the ocean greatly affects its work efficiency. Under the circumstance that high-precision GNSS positioning signals cannot be obtained, the role of the Strapdown Inertial Navigation System/Doppler Velocity Log (SINS/DVL) integrated navigation system is becoming more prominent. Due to marine creatures or the seafloor topography, DVL is prone to outliers or even failures during measurement. To solve these problems, a LSTM/SVR-VBAKF algorithm aided integrated navigation system is proposed. First, under normal circumstances of DVL, the output information of SINS and DVL are used as training samples, and they train the Long Short-Term Memory (LSTM) model. To enhance the robustness and adaptability of the filter, a novel variational Bayesian adaptive filtering algorithm based on support vector regression is proposed. When the DVL formation is missing, the deep learning method adopted in this paper will be continuously output to ensure the effect of integrated navigation. The shipboard test data is verified from two aspects: filter performance and neural network-assisted integrated navigation system capability. The experimental results show that the new method proposed in this paper can effectively handle a situation where DVL output is not available.

The integrated navigation system based on the Global Navigation Satellite System (GNSS) in conjunction with the strapdown inertial navigation system (SINS) and the Doppler Velocity Logger (DVL) is essential for accurate and long-distance navigation in maritime environments. However, the error of the integrated navigation system gradually diverges due to the inevitable velocity measurement error of DVL when GNSS outages occur. To ensure the high navigational accuracy and stability of SINS, it is necessary to dynamically adjust the damping state of SINS provided externally. In this paper, we have developed a novel method for damping state switching based on machine learning with SINS. We construct a model of the change in reference velocity error and use sliding window technology to obtain the reference velocity error for model training. Before training, the digital compass loop is designed to process and highlight the change in reference velocity change errors. In order to reduce the impact of the damping switching, a variable damping system is used to transform the traditional one-time switching of the damping coefficient into a gradual switching, effectively reducing the impact of a sudden change in the damping coefficient on the system. Simulation experiments and tests on ships show that the proposed method effectively reduces the overshoot error integrated underwater during state switching. This research is of great importance for the optimal design of integrated underwater navigation systems.