On the synthesis of adaptive parameter-dependent output feedback controllers through LMI-based optimization
G. Chesi
Abstract:This paper addresses the problem of designing adaptive output feedback controllers for stabilizing plants affected by parameters. A novel approach is proposed that allows one to des ign a fixed-order fixed-degree adaptive parameter-dependent ou tput feedback controller by solving convex optimization problems with Linear Matrix Inequalities (LMIs). The proposed approach i s based on the construction of a function that provides a stabi lity margin of the closed-loop system depending on the controlle r. The conservatism of the proposed approach can be reduced by increasing the size of the LMIs. Keywords–Parameter-dependent; Adaptive Controller; Stability; LMI. I. I NTRODUCTION Ensuring stability is of fundamental importance in engineering. Given an unstable plant, this is generally achieve d by designing a stabilizing output feedback controller, i.e ., a controller that elaborates the output of the plant in order t o provide an input to the plant such that the so obtained closed loop system is stable. The design of such a controller is base d on the model of the plant, and several techniques can be used. Real plants are often affected by parameters. These can happen due to various reasons. One reason is that such parameters can represent quantities that the user can modif y, such as the gain of an amplifier, in order to achieve a differen t performance. Another reason is that such parameters can represent quantities that are unknown or subject to changes , such as the mass, resistance, temperature, etc. Whenever the plant is affected by parameters, the output feedback controller should be able to ensure stability for a ll admissible values of the parameters. For this, the controll er should be dependent on the parameters in general, i.e., shou ld be able to adapt to different plants corresponding to differ ent values of the parameters. Such a controller would be, hence, adaptive, in particular parameter-dependent. Unfortunately, the design of stabilizing output feedback controllers for plants affected by parameters is a difficult problem. Indeed, several conditions do exist in the literat ure for establishing stability of systems affected by paramete rs, in particular conditions based on convex optimization constrained by LMIs; see for instance [1]–[5]. However, such conditions lead to nonconvex optimization problems whenev r a controller is searched for, generally due to the product of the Lyapunov function and the controller that generates Biline ar Matrix Inequalities (BMIs); see for instance [6] [7]. Also, several non-LMI strategies are available for the design of stabilizing feedback controllers for plants that are not af fected by parameters, however, for plants affected by parameters, such strategies could not be used due to the lack of analytical expressions (such as factorizations dependent on the paramet ers) or could lead to controllers of unacceptable order and degre e; see for instance [8]. This paper addresses the problem of designing adaptive output feedback controllers for stabilizing plants affect ed by parameters. A novel approach is proposed that allows one to design a fixed-order fixed-degree adaptive parameterdependent output feedback controller by solving convex optimization problems with LMIs. The proposed approach is based on the construction of a function that provides a stabi lity margin of the closed-loop system depending on the controlle r. The conservatism of the proposed approach can be reduced by increasing the size of the LMIs. A numerical example illustrates the proposed approach. This paper extends our previous work [9]. The paper is organized as follows. Section II introduces the preliminaries. Section III describes the proposed appr oach. Section IV presents an illustrative examples. Lastly, Sect ion V concludes the paper with some final remarks. This work is supported in part by the Research Grants Council of Hong Kong under Grant HKU711213E. II. PRELIMINARIES Notation:N,R,C: sets of nonnegative integers, real numbers, and complex numbers; re(A): real part of matrixA; A: transpose of matrixA; A ≥ 0: symmetric positive semidefinite matrix A; spec(A): set of eigenvalues of A. Let us consider the plant
Mathematics,Engineering,Computer Science