Vessel Trajectory Prediction Using Radial Basis Function Neural Networks

This work presents a novel data-driven modeling approach for the direct prediction of a vessel’s trajectory through the use of AIS data. The proposed method is based on radial basis function neural networks trained with the fuzzy means algorithm, a combination which produces models of high accuracy and simple structures. The produced model is applied on real AIS data in order to approximate the behavioral patterns of cargo ships when moving in the vicinity of a busy port. Results show that the proposed method outperforms a well-established machine learning technique, namely multi-layer perceptrons, not only in terms of accuracy for one-step and multi-step-ahead prediction, but also by providing lower computational times; these facts make it suitable for use in receding horizon integrated control frameworks.