Eecs563 Fall23

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

EECS 563

Hybrid Control Systems


Fall 2023
Updated on1 : 08/24/2023
Instructor: Necmiye Ozay, (she/her/hers), necmiye@umich.edu
GSI: TBD
Schedule: Lectures: MW, 3:00pm-4:30pm
Office Hours: MW, 4:30pm-5:30pm, or by appointment
Prerequisites: EECS 562 (or equivalent), or EECS 560 (or equivalent) and consent of
the instructor. It is highly recommended that the students take at least two graduate level
control courses before taking this course. Students are expected to be comfortable with
state-space methods and should have a basic understanding of Lyapunov theory and Lya-
punov functions. Some programming experience with MATLAB or Python will be useful.
Grading and evaluation: Class work consists of a few homework assignments, critiquing
research papers and “in-class” discussions, and a (team) term project (see pages 4-5 for
details).
• 40% Homework (7-8 assignments)

• 12% Paper presentation

• 8% Paper critiques (4 critiques)

• 5% Class participation (including Perusall annotations and Piazza activity)

• 35% Term project

Summary
Hybrid systems, dynamical systems where continuous dynamics and discrete events inter-
act, are ubiquitous and can be found in many different contexts. Examples are as diverse
as manufacturing processes, biological systems, energy systems, medical devices, robotics
systems, automobiles and aircrafts. Advances in computing and communication technolo-
gies have enabled engineering such systems with a high degree of complexity. Most of
1
This is a tentative syllabus.

1
these systems are safety-critical, hence their correctness must be verified before they can
be deployed. This course will provide a working knowledge of several analysis and design
techniques that can guarantee the satisfaction of certain safety and performance specifica-
tions for such systems.
The course will introduce a combination of tools from computer science (automata
theory, ω-regular languages, temporal logics) and control theory (Lyapunov functions, op-
timization based control) for modeling, formally specifying, verifying and controlling hy-
brid systems. We will cover both theoretical and computational aspects. Recent research
progress in control of networked cyber-physical systems will be discussed. We will present
methods for synthesizing hierarchical control architectures with low-level continuous con-
trollers and high-level protocols for decision making. Finally, we apply the theory and
algorithms in case studies to complex problems such as automated highway systems, ve-
hicle management systems, motion planning and neural networks. Various software tools
will be used in the course.

Topics
• Specifications: Automata theory, temporal logics, discrete transition systems
• Model checking for discrete systems
• Hybrid systems modeling: hybrid automata, switched systems, piecewise-affine sys-
tems
• Stability of (linear) switched and hybrid systems
• Verification of hybrid systems: direct methods (barrier certificates, reachable set
computations), abstraction-based methods (simulation, bisimulation relations)
• Correct-by-construction controller synthesis: reactive control synthesis, two-player
discrete games, hierarchical control protocols, switching protocols
• Model-predictive control of hybrid systems
• Advanced topics (depending on time and interest): falsification, stochastic systems
and probabilistic model checking, quantitative objectives, system identification for
hybrid systems, distributed protocols

Apart from the above mentioned topics, we will spend two-three lectures on advanced
optimization methods (e.g., sum-of-squares) used in control and verification.

2
References
There is no textbook for the course. There will be some lecture notes supplemented
by reading assignments, mostly from recent research articles. The following books and
manuscripts can be used as reference.

1. Principles of Model Checking, C. Baier and J.-P. Katoen, The MIT Press, 2008 (li-
brary link).

2. Model Checking, E. M. Clarke, O. Grumberg, D. A. Peled, The MIT Press, 1999.

3. An Introduction to Hybrid Dynamical Systems, A. J. van der Schaft, J. M. Schu-


macher, Springer, 2000.

4. Switching in Systems and Control, D. Liberzon, Birkhauser, 2003.

5. Verification and Control of Hybrid Systems: A Symbolic Approach, P. Tabuada,


Springer, 2009 (library link).

6. Formal Methods for Discrete-Time Dynamical Systems, C. Belta, B. Yordanov, E.


Aydin Gol, 2017 (library link).

7. Verifying Cyber-Physical Systems: A Path to Safe Autonomy, S. Mitra, 2021.

8. Automata, Logics, and Infinite Games: A Guide to Current Research, E. Grädel, W.


Thomas, T. Wilke, eds., Vol. 2500. Springer, 2003 (library link).

9. Predictive Control for linear and hybrid systems, F. Borrelli, A. Bemporad, M. Morari,
2013 (manuscript).

10. Hybrid Systems: Foundations, advanced topics and applications, J. Lygeros, S. Sas-
try, and C. Tomlin, 2012 (manuscript).

Click the blue links to access these books/manuscripts through the University of Michigan
library or authors’ websites.

Grading Policy
Homework (40%): Collaboration on homework assignments is encouraged. You may
consult outside reference materials, other students, or the instructor. You must properly
cite any outside reference that you use and list the students you discussed the solutions
with. All solutions that are handed should reflect your understanding of the subject matter

3
at the time of writing/implementation. You have one 48 hour grace period that you could
use during the semester; that is you can submit one of your problem sets up to 48 hours
late with no penalty. Note that it does not matter if you submit a problem set 30 minutes
late or 48 hours late, it will count as using the grace period. There will be a second 48
hour “diminishing” late submission period with up to 50% penalty. If you are submitting a
problem set late for the second time (i.e., you already used up your grace period), then the
grade you receive from the problem set will be scaled by 100 − (submission time−due
48
time)50
%. If
you submit more than 48 hours late, you will get zero points.

Paper presentation and critiques (12%+8%): Critically reading a research article is a


crucial skill for a graduate student. It is also important for engineers who want to keep
up with the state-of-the-art in their field. We will devote approximately ten lectures for
understanding and discussing relatively recent, high-impact research papers in the area of
hybrid systems. These papers will be chosen mostly from control and robotics journals
(IEEE TAC, Automatica, IJRR, IEEE T-RO, etc.) and conferences (HSCC, CDC, ACC,
ICRA, EMSOFT, ICML, RSS etc.). A team of 3 (or 4 depending on the class size) will
be responsible for each paper. They will prepare a presentation roughly with the following
parts:

• summarize the preliminaries and definitions required to understand the paper,

• present the concise problem statement (something of the form “given A, B, C, com-
pute/prove/achieve D”),

• present the main ideas (theory or algorithms),

• present the experimental/numerical results.

It can be a tag-team presentation or each student can be responsible for one of the above
parts. They will also lead the discussion after the presentation. Each student will present
once through the term. This presentation will constitute 12% of the grade. The rest of
the class are also expected to read the paper and we will use Perusall (or a similar tool)
for annotating the paper to be discussed. In addition, for a total of four papers among the
ones presented, you will need to prepare a 1-2 page(s) review including a summary of the
paper contributions, paper strengths, paper weaknesses, and extensions/applications you
can think of. The reviews are due at the beginning of the class the paper will be discussed.
If you are part of the presenting team that week, you do not need to write a review. You
might find it useful to read “How to read a research paper” by Michael Mitzenmacher
before writing your reviews. The reviews will constitute 8% of your final grade.

4
Term project (35%): The projects can either be in the form of a review of an area of
the literature or, preferably, involve the exploration of original applications or research
ideas. The length of the project report can be inversely proportional to its originality. Joint
project proposals (with groups of 2 or 3) are encouraged. The project should be chosen in
consultation with the instructor. We will cancel one lecture during the semester and I will
have individual meetings with project groups instead. The schedule is as follows:

• Project proposal (2 page summary) due around October 9

• Project update presentations: December 2 and 7

• Project report (10-12 pages) due final exams week

Some project ideas:


Investigation of a class of hybrid systems: evaluation and benchmarking of synthesis or
verification tools/techniques for a subclass of hybrid systems.
Applications: robotics, (semi)autonomous driving, aircraft electric power systems, air-
craft environmental control systems, car engine cooling, verification of neural networks.
Software development: parallel implementations of abstraction algorithms for hybrid
systems, code generation for low level constrained reachability controllers.
Open-ended problems: hierarchical (nested) abstractions of dynamical systems, dis-
tributed control protocols and cooperation in multi-player games, synthesis with partial in-
formation, temporal logic fragments for human-machine interaction, temporal logic based
reactive synthesis for security and privacy.

Commitment to equal opportunity


U-M is committed to a policy of equal opportunity for all persons, and it does not discrim-
inate on the basis of race, color, national origin, age, marital status, sex, sexual orientation,
gender identity, gender expression, disability, religion, height, weight, or veteran status.
In this class, I aim to treat everyone with fairness and respect, and I expect you to do
the same. I aim to create an environment where we can learn together, freely ask questions,
and help each other achieve better, while following the Engineering Honor Code. Please
feel free to contact me with any problem, concern, or suggestion.

Accommodations for students with disabilities


U-M is committed to providing equal opportunity for participation in all programs, services
and activities. Request for accommodations by persons with disabilities may be made by

5
contacting the Services for Students with Disabilities (SSD) Office, G-664 Haven Hall,
763-3000, . After your eligibility for an accommodation has been determined, the instruc-
tor will be automatically notified. If you have additional questions regarding the accom-
modations, feel free to let the instructor know. Any information you provide is private and
confidential and will be treated as such.

Student mental health and well-being


Graduate school can be stressful and it is critical to care for your emotional and mental well-
being. Resources and support are available for students through the Michigan Engineering
C.A.R.E. Center, Dean of Students Office or Counseling and Psychological Services. The
University also offers SilverCloud, an online mental health tool that offers self-guided pro-
grams for anxiety, depression, stress, resilience, or insomnia.

Resources to report sexual and gender-based misconduct


Here is some information about how to report sexual and gender-based misconduct. People
in certain roles are considered “Individuals with Reporting Obligations” (IROs) and are
required to report suspected Prohibited Conduct to the Equity, Civil Rights and Title IX
Office at the University of Michigan. I AM an IRO.
You can use the following link to notify the Equity, Civil Rights and Title IX Office at
the University of Michigan (ECRT) about assault or harassment: ECRT. Please note that
Title IX offices often distinguish between making a “report,” which does not launch an
investigation, and filing a “complaint,” which does.

You might also like