Short-term ship roll motion prediction using the encoder–decoder Bi-LSTM with teacher forcing

• DOI: 10.1016/j.oceaneng.2024.116917
• Publication Date: 2024-01-31T20:56:00Z
• Authors: Shiyang Li; Tongtong Wang; Guoyuan Li; Robert Skulstad; Houxiang Zhang
• Keywords: Forcing (mathematics)

The safety of maritime operations has become a paramount concern with the advancement intelligent ships. Ship stability and are directly impacted by roll motion, making prediction short-term ship motion pivotal for assisting navigators in timely adjustments averting hazardous conditions. However, predicting poses challenges due to nonlinear dynamics. This study aims predict leveraging encoder–decoder structure Bidirectional Long Short-Term Memory Networks (Bi-LSTM) teacher forcing. model is accomplished employing an map input sequences output varying lengths, forcing enhance network’s ability extract information. To refine analyze model, aspects such as quantity training data guarantee generalization, establishing apposite length relationships between sequences, assessing performance various sea states investigated. Additionally, comparative experiments intervals 10s, 30 s, 60 120 s conducted substantiate necessity effectiveness proposed network. dataset originates from commercial professional simulator developed Norwegian company Offshore Simulator Center AS (OSC).