学習型フィードフォワードコントローラによる船舶の操縦運動制御システムの研究

概要

論文の詳細を見る
Recently, neural networks have been studied using the back propagation learning algorithm for the manoeuvring control systems of ships. But usually it is difficult to consider the dynamic characteristics of the controlled system and to determine the number of layers, units and the value of synaptic weights. In this paper, we try to apply the Learnig Feed-Forward Controller (LFFC) to the manoeuvring control system of the ship. This control system consists of a feed-forward controller based on an inverse-dynamic model of the controlled system and a feedback controller like a Proportional, Integral and Derivative controller (PID). The inversedynamic model expresses the values of the necessary manipulated variables to the manoeuvring motion and with the feedback error learning method the output of the feedback controller tends to zero as learnig proceeds. That is, this system is controlled by the feedback controller at first and subsequently the LFFC begins to learn and takes over the control. As a result of learning, the system attains fast responsibility to the demand. A series of numerical experiments or computer simulations were executed for a container ship as an example and the fundamental data for the development of the new control system are obtained.
社団法人日本船舶海洋工学会の論文
1992-03-29