ME 450: ADVANCED TOPICS IN CONTROLS

DATA-DRIVEN MODELING & ROBUST CONTROL

http://www.lehigh.edu/~eus204/teaching/ME450_SIRC/ME450_SIRC.html

Instructor: Eugenio Schuster

Room 454, Packard Laboratory

Phone: (610) 758-5253

Email: schuster@lehigh.edu

Class Times	Tuesday/Thursday: 10:45AM to 12:00 PM
Location	PA 454
Course Description:	Theoretical and practical knowledge on methods to develop mathematical models from experimental data. Multi-input and multi-output feedback control; robustness analysis of control systems; H-infinity feedback control; performance limitations in control systems; system model reduction.
Textbook:	“System Identification: A Theory for the User,” by Lennar Ljung, 2^nd ed., Prentice Hall, 1998 (ISBN: 0136566952). “Multivariable Feedback Control”, by S. Skogestad and I. Postlethwaite, Second Edition, John Wiley & Sons, 2005. (ISBN-13 978-0-470-01168-3).
Prerequisites:	ME 343 (Classical Control) or equivalent course, or consent of instructor.
Office Hours:	By appointment.
Syllabus:	1. Mathematical Background: Random Variables and Stochastic Processes; Discrete-time Signals and Systems; Model Parameterization and Prediction; Commonly used Signals: Spectral Properties; Persistent Excitation. 2. Nonparametric Identification: Impulse and Step Response; Correlation Methods; Spectral Analysis. 3. Parametric Identification: Determining Model Dimension; Minimizing Prediction Error; Linear Regression and Least Square Estimation; Identifiability, Convergence and Consistency; Asymptotic Distribution of Parameter Estimates; Instrumental-Variable Method; Realization Method; Kalman Filtering. 4. MIMO Systems: Closed-loop performance evaluation, Loop-shaping, Transfer function matrices; Smith-McMillan form; Poles, zeros, principal gains; Norms. 5. Limits of Performance: Feedback properties; Weighting functions; RHP poles/zeros; Bode gain/phase; Bode’s sensivity integral. 6. Uncertainty Robustness: Model Uncertainty; Linear Fractional Transformations; Structured singular value; Robust stability & performance. 7. Design of Robust Controllers: H¥ and m-synthesis; H¥ loop shaping. 8. Controller Reduction: Balanced realization; Optimal Hankel norm reduction; Balanced realization in closed form.
Grading	Assignments
Assignments:	Assignments will be open-book, open-notes, and take-home. In addition to written exercises, Matlab assignments will be given to demonstrate the theory in the text. You CANNOT work with others on assignments. Late assignment will not be accepted. Assignments should be neat, the pages should be stapled with one staple in the upper left corner, and the problems should be in order. Matlab codes must be provided.