ROBUST AUTOMATIC SHIP HEADING CONTROLLER FOR VARIOUS CONDITIONS

The paper presents a ship heading controller with the specific training algorithm. The controller is based on the use of a feed forward neural network and is relatively robust to simulated disturbances, changes in ship's speed and environmental conditions. After training on the linear Nomoto model, the controller switches to a non-linear model or a real vessel without the need for additional adaptation, which does not reduce the quality of its operation in the mode of stabilization or change in the heading of a ship. The objective function of maximum management efficiency with a minimum of deviations from the heading and rudder deflections throughout the entire range of tasks, was obtained. A mathematical model of the double-response controller of the ship's heading has been developed. According to the simulation results, the characteristics of a configured PID controller and a neural heading controller are compared. The simulation results showed, on average, two times higher efficiency of the neural controller compared to the PID controller when performing maneuvers with different ship speeds and wind disturbances.