Enhancing Maritime Safety: Ship Attitude Prediction Using a Hybrid Bi-LSTM with Self-Attention Mechanism

Accurate prediction of ship attitude is crucial for ensuring maritime safety. However, the complexity of the marine environment and the long-term dependency of ship motion pose significant challenges to this task. In this study, we propose a hybrid model that combines bidirectional long short-term memory (Bi-LSTM) neural networks with a self-attention mechanism. The proposed model leverages bidirectional LSTM to capture temporal dependencies in ship motion data while incorporating a self-attention mechanism to dynamically focus on key information within the data sequence. To evaluate our model’s predictive accuracy, we conduct experiments using real-world ship motion datasets for predicting ship roll and pitch. Performance metrics such as root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are employed for comparison against two traditional methods: ARMA and LSTM. Experimental results demonstrate that our hybrid model outperforms these traditional methods by effectively capturing complex attitude variations in long time series data.