ECE 720

Multivariable and Robust Control


Spring 1996: Thursday, 7:20-10:00 p.m., ST2, Room 360, Dr. Beale
Prerequisites: Any ECE 600/700 level or INFT 800/900 level controls class, or POI
Text: Multivariable Feedback Design, J.M. Maciejowski, Addison-Wesley

Objectives:
To introduce students to the concepts, terminology, and issues that are of current research interest in multivariable and robust control theories.
To provide students with the ability to design control systems and to conduct research using the LQG/LTR and H_infinity design methodologies.

Topics Covered:

Eigenstructure assignment for multivariable systems
The Smith-McMillan form and multivariable poles and zeros
Internal stability
Modeling system uncertainty
Performance specifications and principal gains
Parametrization of all stabilizing controllers
LQG problem, loop shaping, and loop transfer recovery
The H_infinity design methodology

Course Outline:

Chapter 1 -- Introduction, definitions of multivariable and robust control, form of the standard control problem, performance characteristics and limitations - 1.5 class periods.
Chapter 2 -- Multivariable systems, the importance of eigenvectors, getting performance improvement through eigenstructure assignment, defining the poles and zeros of a multivariable system, and the concept of internal stability - 2.5 class periods.
Chapter 3 -- Singular values and characteristic values, open-loop and closed-loop gains, operator norms, structured and unstructured uncertainty, stability and performance robustness - 3 class periods.
Chapter 5 -- Robust control using LQG/LTR methodology: the Linear Quadratic Gaussian (LQG) problem, robustness of the linear regulator and of the Kalman filter, loss of robustness with the LQG structure, and maintaining robustness through loop transfer recovery (LTR) - 4 class periods.
Chapter 6 -- Robust control using H _infinity methodology: justifying and formulating the H_infinity problem, parametrization of all stabilizing controllers, the model matching problem, and a numerical robust control design algorithm - 4 class periods.

Course Requirements:

There will be two design projects and one technical paper review in this course. The first design project will deal with eigenvector assignment in multivariable systems. The second project will deal with robust control design using the LQG/LTR methodology. The third assignment will be a critical review of a current paper in the area of robust control theory.

Design Project #1

Design Project #2

Project #3

Click the icon to return to Dr. Beale's home page

Latest revision was made on 05/09/01 04:47 PM