Roll Prediction of Unmanned Boat Based on LSTM-SVR

During the navigation of unmanned boats on the sea, complex movements are generated due to environmental influences. Among them, roll motion can affect navigation safety and even cause harm. A roll prediction method aiming to take control measures ahead of time to ensure the safe navigation of unmanned boats is proposed in this paper. Firstly, the LSTM (Long Short-Term Memory) network is used for performing the single-step prediction. Secondly, the errors of the LSTM model's prediction can be predicted by the SVR (Support Vector Regression), aiming to compensate for the forecasting errors. Simulation verification is conducted using actual ship data, and the prediction effect illustrates that the LSTM-SVR obtains higher forecast accuracy compared to the LSTM, and its prediction ability is better than the LSTM model.