フィードバック線形化に基づく非線形システムの制御

概要

論文の詳細を見る
This paper addresses the problem of feedback linearization of nonlinear systems. Feedback linearization is defined as the problem of finding a change of coordinates and a suitable nonlinear feedback law for a given nonlinear system, such that it behaves like a linear one. Differential geometry application to nonlinear control systems enables the system to be globally or locally feedback linearizable. Most of the work in linearization is based on the premise that the nonlinear model of the physical system is precisely known with its parameters exactly defined. However, in practical applications, the system dynamic behavior may exist as empirical data (more likely black box systems). This paper emulates Lie derivatives which realize change of coordinates using Artificial Neural Networks (ANN). ANNs are trained by error back propagation. Finally, the relative degree of feedback linearization is determined and a nonlinear feedback law is established.
久留米工業大学の論文
1998-12-20