Non-iterative data-driven controller parameters tuning of feedback linearizing controller

Abstract

Non-iterative data-driven controller parameters tuning achieves sub-optimal controller parameters from one-shot experimental data. Among such approaches, Fictitious Iterative Feedback Tuning (FRIT) is one of the representative one as well as Virtual Reference Feedback Tuning (VRFT) and Non-iterative Correlation based Tuning (NCbT). While most of prior researches on FRIT consider linearly parametrized controller, the paper proposes a FRIT of a feedback linearizing controller for a certain class of nonlinear systems. The proposed FRIT assumes that the controlled plant is a single-input, single-output, continuous-time, nonlinear affine systems with unknown parameters, and provides a controller parameterization that achieves feedback linearlization. Thus, the reference model can be given as a linear transfer function. The desired control parameters are obtained by optimizing a cost criterion composed of a fictitious reference signal. Using Radial Basis Function (RBF) networks for estimating unknown characteristics, the proposed approach can be applied even if nonlinear characteristics are not known beforehand. The paper also examines the well-posedness of the proposed nonlinear feedback controller. Finally, a numerical example is shown to demonstrate the effectiveness of the proposed method.