Artificial neural network and mathematical modeling of automatic ship berthing

Abstract

Automatic berthing has been known as one of the most challenging problems in ship control. During port approach and berthing maneuvers, the ship master takes into account many factors before any maneuver action, i.e. ship speed, wind speed, wind direction, current water direction, available power, heading angle, and ship response. Many methods related to automatic berthing were developed by recent research, such as Artificial Neural Network, Adaptive Backstepping, Nonlinear Programming, and Proportional-Integral-Derivative. However, most of these researches adopted a simplified dynamic model that reduces the validity of the optimal solution and may lead to dynamics that do not express real-time conditions. In this paper, a feed-forward controller using a non simplified mathematical model is developed. Numerical simulations were performed to verify the effectiveness of the proposed controller and test its ability to control the ship safely to reach the goal points of the berthing plan. The agreement between the ANN model results and experimental data is impressive.