Development of an Artificial Neural Network Model for Predicting Ship Berthing States based on Model Test Data

Autonomous ship berthing is a complex maneuver that necessitates precise predictions of ship motion to ensure safe and efficient docking. This study developed an Artificial Neural Network (ANN) model to forecast a ship's motion during the berthing process. Training and testing data were gathered from physical model tests conducted in a square tank, utilizing a small-scale ship model under various berthing conditions. The hyperparameters of the ANN model, including the number of hidden layers, nodes per hidden layer, and training epochs, were selected through iterative testing and cross-validation to optimize prediction accuracy and model stability. The ANN model was trained using time-series input data, such as the ship’s position and heading angle. The target outputs included surge and sway velocities, yaw rate, position, heading angle, and control variables like steering angle and propeller revolution. To assess the model's performance, its predictions were compared with actual test results. The findings indicate that the ANN model accurately tracks the real motion trends of the ship with minimal prediction error. The predicted responses closely aligned with the experimental data, demonstrating the model’s capacity to capture realistic berthing dynamics. This study confirms that the ANN model can reliably and swiftly predict ship behavior during berthing, thus enhancing port safety and supporting autonomous berthing systems.