Abstract
In recent years, with the focus of the maritime industry on the development of Maritime Autonomous Surface Ships (MASS), intelligent algorithms have been developed to be used as the decision support layers for these types of ships. The implemented navigation systems on such vessels require advanced control strategies to guarantee safe maneuvers during complex operational scenarios. In this context, leveraging data-driven methods for the development of various systems, such as collision avoidance systems, offers promising results in handling challenging maneuvers. This paper applies a human-centric computational intelligence method called Adaptive Neuro-Fuzzy Inference System (ANFIS) to make automated ship maneuvers. In this approach, the ship’s next control inputs during a maneuver are predicted by cloning and replicating the human’s navigation behavior. This method, which is a data-driven technique, utilizes the ship’s states, including heading and positions, as the inputs to the ANFIS model. The proposed approach is trained on the simulation data obtained from a high-fidelity simulator, meaning ANFIS learns the nonlinear underlying relationship between the ship states and the necessary control inputs to follow the desired trajectory.