Prediction Model of Marine Drifting Trajectory Based on Multilayer Perceptron
摘要
This study introduces a trajectory prediction method for the drift of marine buoys, utilizing the Multilayer Perceptron (MLP) model, designed to enhance maritime search and rescue operations as well as ocean resource development. We preprocessed marine buoy drift data to build an MLP model for trajectory prediction. The model excels in forecasting linear and large curvature paths and shows accuracy in more complex trajectories. It outperformed 1D CNN, CNN-MLP hybrids, SVR, and LSTM in metrics like MSE and R2. The MLP's simplicity and efficiency make it ideal for real-time predictions, though improvements are needed for complex trajectory changes. This research provides a practical tool and foundation for future studies in maritime search, rescue, and ocean resource management.