A Nonparametric Approach to Design Fixed-order Controllers for Systems with Constrained Input

This paper presents an approach for designing fixed-structure controllers for input-constrained linear systems using frequency domain data. In conventional control approaches, a plant model is needed to design a suitable controller that meets some user-specified performance specifications. Mathematical models can be built based on fundamental laws or from a set of measurements. In both cases, it is difficult to find a simple and reliable model that completely describes the system behavior. Hence, errors associated with the plant modeling stage may contribute to the degradation of the desired closed-loop performance. Due to the fact that the modeling stage can be viewed only as an intermediate step introduced for the controller design, the concept of data-based control design has been introduced, where controllers are directly designed from measurements. Most existing data-based control approaches are developed for linear systems, which limit their application to systems with nonlinear phenomena. An important non-smooth nonlinearity observed in practical applications is the input saturation, which usually limits the system performance. Here, we attempt to develop a nonparametric approach to design controllers from frequency-domain data by taking into account input constraints. Two practical applications of the proposed method are presented to demonstrate its efficacy.