**Disclaimer**

This syllabus is to be used as a guideline only. The information provided is a summary of topics to be covered in the class.
Information contained in this document such as assignments, grading scales, due dates, office hours, required books and materials may be from
a previous semester and are subject to change. Please refer to your instructor for the most recent version of the syllabus.

MAE 509 08/23/2021

MAE 509: LMIs in Optimal and Robust Control

Syllabus
Instructor: M. Peet
email: [email protected]
Zoom link: https://asu.zoom.us/j/2031490735
website: control.asu.edu
phone: 480-965-5599
Meeting Time: M W 4:30-5:45 in ECGG335
Office Hours: F 4:30-5:30PM (online only)
Outside Office Hours: Appointment by email

Introduction Please watch the videos: Intro Part 1 and Intro Part 2 for an introduction to the material
and an overview of the course organization.

Motivation The recent introduction of Linear Matrix Inequality (LMI) methods in control has dra-
matically expanded the types and complexity of the systems we can control. For example, consider the
problems: Gain Scheduling for Missile Attitude Control (A switched system); Control of Robots over
Noisy Communication Channels (a sampled-data systems); Remote Control of Spacecraft Attitude (a
delayed system); Behavioural Therapy (A system with binary inputs); or self-driving vehicles (a case of
decentralized control). None of these systems can be studied using classical root-locus or PID control
methods. Rather, advances in these fields have been made possible through the increased power and
flexibility created by the LMI (optimization-based) approach to control.

Course Structure This class will be inverted. Lectures will be posted on canvas, with an associated
in-class meeting deadline, and must be viewed before class on that date. In-person attendance at lecture
is MANDATORY and students will be strongly penalized for failure to attend. During in person meeting
times, activities will be mixed. However, there will be a minimum of 7 classes which utilize the disputation
lecture format. At each disputation lecture, all students should be prepared to answer the questions
posed by the instructor and to support their response using material from the lecture slides and scientific
literature. Students should also be prepared to rebut or dispute answers given by other students.

Discord Please join the course discord sever at Discord Invite Link for a detailed (but tentative)
schedule of the course.

Course Objectives To be able to use LMI solvers to synthesize optimal or suboptimal controllers and
estimators for multiple classes of state-space systems.

Learning Outcomes Upon completion of this course, students will be able to:
1. Recognize the three parts of an optimization problem: Variables, Objective, Constraints
2. Categorize an optimization problem as convex or non-convex
3. Categorize a convex optimization problem as semidefinite programming, linear programming, second-
order cone programming, etc.
4. Recognize the difference between tracking and regulator frameworks
5. Construct the 9-matrix representation of an optimal control problem

1 of 4
 6. Construct an H∞ -optimal dynamic output feedback controller using LMI solvers
7. Implement and Test a dynamic output feedback controller on a state-space system

8. Construct the LFT representation of a robust control problem

Grading Policies There will be approximately 5 homework assignments which cumulatively will be
the basis for 25% of the course grade. A midterm will be given for 30% of the grade. Disputation, in-class
activities, and the wikipage project will constitute 20% of the grade. The final project or final exam will
be for 25% of the grade.

Absence Policy Attendance at all in-person lectures is required. Exceptions include excused absences
related to university sanctioned events/activities that are in accord with ACD 304–02, “Missed Classes
Due to University-Sanctioned Activities” and excused absences related to religious observances/practices
that are in accord with ACD 304–04, “Accommodation for Religious Practices”. In general, late homework
will not be graded and missing examinations cannot be retaken. Exceptions include excused absences
related to university sanctioned events/activities that are in accord with ACD 304–02, “Missed Classes
Due to University-Sanctioned Activities” and excused absences related to religious observances/practices
that are in accord with ACD 304–04, “Accommodation for Religious Practices”

Content In this course, we provide an overview of LMIs and their many applications in Modern Control
Theory. In contrast to other courses in control, our subject matter will be more focused on the variety
of applications for LMIs and less on theory. Nonetheless, there is a significant theoretic component and
students should expect to perform rigorous mathematical proofs.
The primary text for the class is “LMIs in Control Systems: Analysis, Design and Applications” by
G.-R. Duan and H.-H. Yu. This text is not organized in the same way as the course, however. The second
text we will use is “Linear Matrix Inequalities in System and Control Theory” by S. Boyd. This second
text is freely available online from the author and may be found at
https://web.stanford.edu/ boyd/lmibook/lmibook.pdf.
A more theoretical treatment of some of the material can be found in “A Course in Robust Control
Theory: A Convex Approach” by G. Dullerud and F. Paganini. Be warned, however, that many students
find this text to be difficult to follow.
I will assume that students are familiar with the basics of state-space systems or are capable to
learning about them very quickly. For an accessible text on State-Space methods, I recommend “Linear
State-Space Control Systems” by D. A. Lawrence. If you have limited exposure to state-space, it might
be a good idea to review the first few chapters of this book.

Schedule Class meets on Monday and Wednesday from 4:30-5:45 in room ECGG335. A detailed
schedule will be posted on canvas and the course discord channel.

Lecture Format Pre-recorded lectures will be posted on canvas. Links to these lectures can also be
found on my website http://control.asu.edu under the classes heading - MAE 598.

Canvas Please monitor canvas for important class announcements. Lecture Notes and pre-recorded
lectures will be posted on Blackboard, along with all assignments and supplementary material. Grades
will also be posted on canvas.

Prerequisites There is no prerequisite for this class. However, it is assumed that all students have
some background in controls and linear algebra. The first two modules will review standard material on
optimization and state-space systems. Access to Matlab is required, including the robust control toolbox.

2 of 4
Final Project In lieu of a final exam, this class will have a final project. This project is expected to be
development of a new LMI. Part of the project will be documentation of this LMI on the course Wikibook.
This project may be based on your own research or may be on an entirely new topic. The project should
leverage LMI methods to solve a problem which has not already been solved. Some problem suggestions
are listed at the end of this syllabus. You may work in groups of up to 2 people. However, if you work in
a group of 2, I will expect double the results as of a group of 1. All project topics should be submitted
to me in advance. If you do not wish to do a project, you may take a final exam instead.

Classroom Behavior Students are expected to participate in in-class activities, including disputation.

Violence See the Student Services Manual, SSM 104–02, “Handling Disruptive, Threatening, or Violent
Individuals on Campus”

Disabilities A reminder to students that when requesting accommodation for a disability, they must
be registered with the Disability Resource Center (DRC) and submit appropriate documentation from
the DRC

Academic Integrity Collaboration on exams will result in an E for the course. Copying of homework
or duplication of material found online will result in a “0” on the homework and a referral to the ASU
office of academic integrity. Reminder: Two referrals to the office of academic integrity is grounds for
expulsion from the university. If in doubt about a specific case, ask me.

Plagiarism Plagiarism applies to homework, numerical code, and the final project. In all cases, it
refers to the use of outside material without attribution. In the case of code, it can mean re-arrangement
of lines, renaming of variables or other cosmetic changes. In the case of the final project, it includes
“rephrasing” of published or unpublished material. This occasionally includes proofs. The penalty for
plagiarism on exams is an E for the course. The penalty for plagiarism on homework is as defined in the
paragraph on Academic Integrity.

References Aside from the text, there are several excellent optional sources which may be consulted.
Although not directly required for the course, students are encouraged to browse the following references.

The following is an introduction to classical control and state-space theory.

The following are references for LMI methods in control.

ˆ Boyd, El Ghaoui, Feron and Balakrishnan. “Linear Matrix Inequalities in Systems and Control
Theory”, SIAM, 1994.

The following is a thorough reference on matrix analysis.

The following is a clearly written text on mathematical analysis.

3 of 4
 Assignments:
ˆ Week 1-3 Read: Duan, chapter 1, 3, Assignment: HW 1

ˆ Week 4-5 Read: Duan, chapter 4.1,4.2, 6.1, 6.2, 7.1, 7.2, 9.1.1, Assignment: HW 2

ˆ Week 6-7 Read: Duan, chapter 8.1,8.2,8.4, Assignment: HW 3

ˆ Week 8-9 Midterm Examination

ˆ Week 10-11 Read: Duan, chapters 7.3, 8.3, 11, Assignment: HW 3

ˆ Week 12-13 Read: Course Notes on SOS/PDEs/Switched Systems

ˆ Week 14-15 Final Examination

Outline: The following is a rough outline of the set of lectures which will be presented.
Lecture 1 - Introduction
Lecture 3 - Introduction to Optimization
Lecture 4 - Introduction to Positive Matrices and LMIs
Lecture 5 - Our First LMI
Lecture 2 - LMIs for Stability, Controllability and Observability.
Lecture 6 - The Optimal Control Framework
Lecture 7 - An LMI for Full-State Feedback Controller Synthesis
Lecture 8 - An LMI for H2 -Optimal Full-State Feedback Control (LQR)
Lecture 9 - The H∞ norm
Lecture 10 - An LMI for H∞ -Optimal Full-State Feedback Control
Lecture 11 - The Luenberger Observer and The Kalman Filter
Lecture 12 - An LMI for Output-Feedback Stabilization
Lecture 13 - An LMI for H∞ -Optimal Output Feedback Control
Lecture 14 - Systems with Uncertainty and the Structured Singular Value
Lecture 15 - An LMI for Robust Stability Analysis
Lecture 16 - The D-K iteration
Lecture 17 - Switched Systems
Lecture 18 - An LMI for Control of Systems with Switching
Lecture 19 - Delayed Systems
Lecture 20 - An LMI for Stability of Systems with Delay
Lecture 21 - Nonlinear Systems
Lecture 22 - An LMI for Stability of Nonlinear Systems

Examples of Proposed Project Topics:

ˆ Control of Robots over Noisy Communication Channels (Sampled-Data Systems)

ˆ Spacecraft Attitude Control with delayed communication (Delay Systems)

ˆ Social Cognitive Therapy using Discrete Inputs (Mixed-Integer Control)

ˆ Self-Driving Vehicles (Decentralized Control)

4 of 4