船舶の操縦運動の多変数制御に対する学習型制御方式の適用と学習の高速化に関する研究

概要

論文の詳細を見る
Recently an optimal regulator theory is often used to the multivariable control system. There are some problem in designing the system with optimal regulator theory. First, the designer must resort to trial and error iteration to the design of the control system. And second, that theory needs accurate mathematical model of the controlled system. So we try to apply the Learning Feed-Forward Controller (LFFC) to multivariable control system. This control system uses a neural network and consists of a feed-forward controller based on an inverse-dynamics model of the controlled system and a feedback controller. The inversedynamics model needs only simple structure of the system. In our earlier paper we applied the LFFC system to single input and single output system which control heading angle using rudder. In this paper we expand its application to the multivariable control system which is 2-input 2-output system that controls heading angle and position in y-direction using a bow thruster and a stern thruster. Because it is difficult to derive the inverse-dynamics model of the system, we approximate the inverse system with linearised model. Moreover to hasten the learning speed we modified usual learning equation by adding proportional term. As a result, the multivariable manoeuvring system with LFFC has a good controllability and the neural network with proposed learning equation can be recognized to learn the inverse system in real-time.
社団法人日本船舶海洋工学会の論文
1994-03-29

船舶の操縦運動の多変数制御に対する学習型制御方式の適用と学習の高速化に関する研究

スポンサーリンク

概要

著者

関連論文

スポンサーリンク