Você está na página 1de 6

EECS 562

AE 551

NONLINEAR SYSTEMS AND CONTROL


Instructor: Prof. Jessy W. Grizzle, 4421 EECS Bldg., (734) 763-3598,
grizzle@umich.edu
Text: Hassan Khalil, Nonlinear Systems, Third Edition, Prentice Hall, 2002
Assumed background: EECS 560 and Matlab.
Lectures: Tu-Th 12:10 PM to 1:30 PM (no break)
Final Exam: 4:10 PM to 6 PM, Thursday, April 21, 2016. By CoE policy, the time
of the final exam cannot be changed.
Outline: The course will be roughly 20% mathematical basics and fundamentals
of differential equations, 50% stability theory, and 30% feedback design. The
field of nonlinear control system analysis and design is extremely vast, and is still
very much incomplete. A lot is known, and much, much more is unknown. We will
go through many proofs. I will try my best to illustrate ideas on examples. Many
of the examples will seem simplistic as they will be chosen to illustrate concepts
or computations AND to be doable on a whiteboard or a HW set. You will be able
to find many papers in the literature where the material we cover is used in real
engineering systems. My web site has many examples in automotive systems
and robotics, for example. Everything we cover will have practical implications,
even though you may not see this right away.
1. Fundamental Theory (Existence of solutions of ODEs, uniqueness, continuity,
real analysis, etc.)
2. State Space Stability Theory (First and Second Methods of Lyapunov)
3. Nonlinear State Variable Feedback Design (Control Lyapunov Functions,
Backstepping, Feedback Linearization)
4. Classic Input Output Techniques (Absolute Stability, Small Gain Theorem,
Describing Functions.)

Grading:
Weekly homework 10%
Inv. Pend. Project 10%
Exams
80%

(equally weighted)

There will be one (1) midterm exam and the final. The midterm exam will be of
two hours duration and will be given at night. You will have at least two weeks
warning of when the midterm exam will take place. The final is at the time set by
the registrar.
Two lowest HW set scores will be dropped. No late HW accepted; see below.
The project is like an extended homework problem. It uses linearization methods
to control the classic inverted pendulum on a cart; you will have very precise
instructions on what to do, and a written report to turn in. There is no oral report.
The goal is for you to apply what you learned in linear systems theory (i.e., EECS
560) to a nonlinear system. In later HW sets, we will design nonlinear controllers
that perform much better!
Other references for nonlinear control design and analysis: these are not
on reserve.

H. Khalil, Nonlinear Control, Pearson, 2015 [A newer textbook by our


textbooks author. It focuses much more on feedback control and less on
stability analysis. Some of you may like this book.]

M. Vidyasagar, Nonlinear Systems Analysis, Second Edition, Prentice Hall,


1993.

M. Kristic, I. Kanellakopoulos and P. Kokotovic, Nonlinear and Adaptive


Control Design, Wiley Interscience, 1995.

J.-J. E. Slotine and W. Li, Applied Nonlinear Control, Prentice Hall

S. Sastry, Nonlinear Systems Analysis, Stability, and Control, Springer, 1999.

R. Freeman and P. Kokotovic, Robust Control of Nonlinear Systems,


Birkhauser, Boston, 1996.

A. Isidori, Nonlinear Control Systems, Springer-Verlag, third edition, 1995.

A. Kelkar and S. Joshi, Control of Nonlinear Multibody Flexible Structures,


Springer Lecture Notes in Control and Information Sciences 221, Springer,
1996.

P. Kokotovic, H. Khalil and J. OReilly, Singular Perturbation Methods in


Control: Analysis and Design, Academic Press, 1986.

H. Nijmeijer and A.J. van der Schaft, Nonlinear Dynamical Control Systems,
Springer-Verlag, New York, 1990.

R. Sepulchre, M. Jankovic and P. Kokotovic, Constructive Nonlinear Control,


Springer, 1997.

G. Tao and P. Kokotovic, Adaptive Control of Systems with Actuator and


Sensor Nonlinearities, Wiley Interscience, 1996.

My expectations: This is included due to apparent confusion in past classes.


Decorum:
a) Arrive on time. Entering late bothers me and other students in the class.
b) Once class has started, put away your cell phones.
c) Sitting in the back of the room and exchanging comments with your
neighbor is very visible to me, the instructor at the front of the room.
Learning the material is not so different than learning a new game, sport, or
dance. It takes repetition and practice. Here is one very good method:
a) Read the book before coming to class. Just scan the material to see what
are the general concepts and which part is the most difficult for you.
b) Take good notes in class. Re-read your notes before going to the next
class. Some students find it advantageous to re-copy their notes after
class and to add remarks in the margins. Another good idea is to
summarize the key parts of the lecture in half a page or less.
c) Re-read the appropriate part of the textbook after attending the lecture.
This time, spend extra effort to learn the main ideas.
Following the above gives you a minimum of four passes through the material.
This will give you a high probability of storing the material in your long term
memory instead of in your short term memory.
Remark: Various studies have shown that sitting in the front of the room
increases the odds of you focusing on the material at hand. But hey, its your
money; as long as you are not disrupting the class, you can do as you please.

EECS 562 HW GRADING SCHEME


Policy: Homework solutions should be legible, clear, complete and correct.
They do not have to match the solutions provided by the professor. Each week
you should compare your solutions to the handed out solutions; your goal is to
understand the reasoning employed in the solution of the problem.
Scoring: Each problem is graded as follows:
a) three (3) points if the problem is perfectly correct or nearly so. Of
course, "nearly so" is a subjective evaluation. I don't consider a single numerical
mistake to be important if it doesn't change the basic problem nor lead to greatly
simplified reasoning.
b) two (2) points if there are several minor errors or at least one major
error, but it is clear that you had a good idea of how to work the problem.
c) one (1) point if the problem was attempted, but the reasoning is quite
flawed, extremely incomplete, or if the solution was unreadable.
d) zero (0) points only if the problem was not attempted.
Further remarks:
1) As noted above, each problem will be worth 3 points. Each HW set will have a
varying number of problems and hence will be worth a varying number of points.
2) I will drop your two lowest HW scores. Hence, if you do not finish one or two
HW sets on time, due to travel or illness, it will not adversely affect your HW
score, though it may, in the long run, have an effect on your exam performance if
you do not do the HW problems for yourself.
3) Because of (2) above, no late HW will be accepted, no matter how good your
excuse. Everyone has the same opportunity to miss two HW sets and it should
not matter whether they have an ill family member, a class trip, a conference, or
they are just extra busy that week. HW solutions will be posted promptly on the
day HW is due.

EECS 562 Handout: Grizzle


Required Background

1. If you have completed the graduate linear systems course at Michigan, EECS 560 = ME 564 =
AERO 550, then you have adequate background and do not need to read further. If you have not
had that course, I will try to list the key things you should learn on your own. The list is not so
long.
2. Part of what EECS 560 gives you is mathematical maturity. That course does a lot of theory. So
does EECS 562. If you have taken ROB 501, then you have even more mathematical maturity than
is acquired in EECS 560, so in that regard, you will be ne.
3. Stability of a linear time-invariant state variable model: x = Ax, with x0 Rn . You need to learn
when a linear time-invariant model is stable in the sense of Lyapunov, when it is asymptotically
stable in the sense of Lyapunov, and when it is unstable. The conditions are given in terms of the
e-values of A and its Jordan canonical form.
4. Stabilizability of a linear time-invariant control system: x = Ax + Bu, with x Rn and u Rm .
When does there exist a linear state variable feedback u = F x such that the closed-loop system
x = (A + BF )x is asymptotically stable? It is enough to know that a sucient condition is for
the system to be completely controllable, and that complete controllabilty can be checked via the
Kalman controllability rank condition: rank [B | AB | | A(n1) B] = n. Also, you can compute
the feedback using the place command in Matlab.
5. For the PROJECT, you will need to construct a (full-order) Luenberger observer for a linear system:
x = Ax + Bu
y = Cx

This knowledge will not be used on any exam. If it is used in HW, it is a very minor part of a HW.
6. We do not use much from undergrad control. Late in the course, during a two week period, we will
use transfer functions, transfer matrices, and Nyquist plots.
7. The enrollment this term is huge. I have never had more than 60 students in the course before.
With this many students, it is going to be really hard to get special help from me. I will do my
best, but no matter how much I may wish to help you, it will not possible to spend much time with
any one person. So, it is very important that you have adequate background to take the course.
I am not going to police your background. At the end of the term, if you say that you are failing
because you do not have adequate background, then I may be empathetic to your case, but I will
not be able to be sympathetic! That is a sad thing to say in writing!! I cannot just give you a
passing grade. You have to earn it.
8. The rst half of the term, I do lots of proofs in lecture. Then in the middle of the term, we do more
design. Then toward the end, there are a couple more hard things where proofs are required. I keep
the lectures at a high level, the HW at a medium high level, and the exams, while challenging, do
not require that you provide proofs. There will be Multiple Choice or T/F questions that test your
1

knowledge of the theory. There will also be work-out problems where you have to demonstrate the
ability to apply the concepts of the course to simple examples.
9. I think this is a great course. I welcome you to take it. HOWEVER, if you have not yet completed
the graduate linear systems course, you need to learn on your own the topics listed above. It is
NOT enough to be taking the course concurrently with EECS 562.

Você também pode gostar