Predicting a Vessel’s Trajectory: A Simple Machine Learning Method

Nowadays, ship safety is a significant concern, with over 50% of ship accidents occurring due to collisions. Therefore, accurately predicting the future position of a ship is essential to avoid such incidents. This is particularly crucial for Autonomous Vessels (AV), as there is no human operator to control the vessel when an obstacle appears in its path. In this study, a procedure was developed to predict a vessel’s future trajectory. Various types of machine learning methods can be employed for this purpose, including the Two-Point Method, Consecutive Average Method, Linear Regression Method, nonlinear regression methods, and deep neural network models. While deep neural network models tend to provide the best results for future position prediction, they necessitate a large dataset. This work employed the simplest machine learning method, namely linear regression, to forecast the future trajectory of a vessel. A computer program using MATLAB was created to predict the future trajectory based on previous GPS positions. The program forecasts future coordinates (Longitude, Latitude) for each second from the current position. The application of weightage and parametric equations was also demonstrated. This method exhibited good prediction accuracy for linear paths but was less effective for curved paths. Nonetheless, this procedure can serve as an initial step in designing Autonomous Vessels, particularly if the AV is intended to follow a straight path.