| Abstract |
In this paper, a general approach of observer and control design for a class of nonlinear mechatronic systems is presented. It is assumed, that the system is separable into a known linear part and an unknown static nonlinearity. This combination is called system with an isolated nonlinearity. The observer starts from the known linear part and is extended by a neural network to approximate the nonlinearity. This leads to a mathematically proven stable identi_cation and observation method that provides information about all system states and the nonlinearity. The presented control concept is based on the nonlinear controllable canonical form (NCCF). It is implemented as a nonlinear state feedback law, which, due to the observer, does not require full state measurement. The result is a globally linearized system where free pole placement is possible. Compared to input{output{linearization the relative degree of the system does not a_ect the design procedure. Furthermore, only the linear part of the system has to be known in advance, the nonlinearity is supposed to be unknown. |
