Identification of nonlinear systems with noisy data: a nonlinear mapping-based concept in time domain

Abstract

It is important to guarantee the accuracy, robustness, and computational stability in identification of nonlinear dynamic systems. This problem becomes a very. challenging and complex one if high power noise is imposed on the measured variables. In control systems, linear and nonlinear filters are used to attenuate the noise in order to implement feedback. This paper studies model-based identification for continuous-time nonlinear multivariable systems in the time-domain by using a nonlinear mapping-based identification concept. The application of linear and nonlinear filters to perform the identification is studied and demonstrated. In particular, the identification problem is formulated and solved using realizable filters to recover the variables to be used to identify the unknown parameters. The importance of the time-domain identification and filtering lie on the need to guarantee accuracy, convergence, computational stability, and robustness. The identification algorithm is demonstrated through its application to nonlinear dynamic systems.