CS/EE 146 Control and optimization of networks (2019 Fall)

Units

9 (3-3-3)

Lectures

MWF 1:00 - 1:55pm, Rm 243 Annenberg

Prerequisites

Ma 2, Ma 3 (or instructor permission)

URL

http://courses.cms.caltech.edu/cs146

Instructors

Steven Low <slow@caltech.edu>, CMS/EE, x6767, Rm 219 Annenberg
Guannan Qu <gqu@caltech.edu>, CMS, Annenberg

Admin. Assistant

Christine Ortega <cortega@caltech.edu>, 245 Annenberg

TAs

Office Hours

Instructors: by email appointment

Course Description

This course is research oriented. We will cover some control and optimization issues in cyber-physical networks, in particular congestion control on the Internet, power flow optimization on power networks, and learning in networks. For each of the three topics, the instructors will cover some basic theory, and then we will read and discuss research papers. Each of us will present papers and lead discussions in class, and your presentations will form a major part of your grade.

Here is the tentative plan, which can be tailored to the students' interest (see a more detailed schedule below):

Projects

Each student should design her/his project, in consultation with the instructor.

Grading

This is tentative and subject to change:

  • Class participation/presentations: 40%
  • Project: 60%

Course Schedule (tentative)

Week

Topic

Notes (continuously updated)

10/2

Internet congestion control: background

10/7

Internet congestion control: equilibrium structure

10/14

Internet congestion control: global stability

10/21

Internet congestion control: local delayed stability

10/28

Optimal power flow: background

11/4

Optimal power flow: exact relaxation

11/11

OPF: relaxation

Learning: background

  • Nov 11 Mon [lead: Sari Kerckhove]: Lingwen Gan, Na Li, Ufuk Topcu and Steven H. Low. Exact convex relaxation of optimal power flow in radial networks. IEEE Transactions on Automatic Control, 60(1):72-87, January 2015
    or Lecture Notes 12.5.8 (Proof of Thm 12.5)
  • Nov 13 Wed, 15 Fri [background: GQ]: background theory on learning in networks (linear regresson, LQR, model-predictive control)

11/18

Learning: from data to model (system id)

11/25

Learning: from data to control (data-driven control)

12/2

Project presentations

  • Schedule TBD (Rona on Dec 6 Fri).