INPUT MATCHINGによる学習制御系の設計

概要

論文の詳細を見る
When a linear controller is applied to a nonlinear system, it sometimes causes some deterioration of the control performance. Neural control can be useful as a controller for nonlinear system. Its learning architecture can be classified into General Learning, Specialized Learning, Forward and Inverse Modeling and Feedback-error Learning Architectures. However, these architectures have the following defects, that is, the output of plant does not always track the objective value, learning law can not be determined when dynamics of plant are unknown and the structure becomes complicated, etc. In this study, to improve the above defects, a Learning Control system, which has simple structure using only one artificial neural network and has feedback function, is proposed via Input Matching.
一般社団法人日本機械学会の論文
1998-05-25