Experimental Study of Intelligent Autopilot for Surface Vessels Based on Neural Network Optimised PID Controller

Abstract

As all ships are required to operate with sufficient reliability and appropriate economy, it is necessary to achieve good controlling at reasonable costs. Autopilot systems have a momentous influence on the performance of ships, enabling them to cruise in various sea conditions without human interventions. This paper introduces a Radial Basis Function Neural Network (RBFNN) based Proportional Integral Differential (PID) autopilot system for a surface vessel. In the proposed control algorithm, the RBFNN trained by adaptive mechanism was utilized to approximate the realistic ship’s behaviours, thereby updating the parameters of the discretising PID based controller in real time, so as to compensate for the environmental disturbances and uncertainties during the ship’s sailing. In order to validate the efficiency of the proposed algorithm, the experiments were conducted in a lake by using the free running model scaled ship ‘Hoorn’. The experimental results indicate that the proposed RBFNN PID based autopilot can decrease the course keeping deviations with reasonable rudder actions.