This is a moderate to advanced course on model-based and model-free methods for optimal control of dynamical systems. Prior knowledge of linear dynamical systems is required. Topics include parametric and non-parametric methods of system identification, model predictive control, and model-free methods for (optimal) controller design.
Course Objective: : The overall objective of this course is to provide the students with a theoretical understanding and practical toolbox that they can use to design controllers for complex real-world systems from data. The specific learning objectives for the student are to:
● Understand the different approaches to the (optimal) control of dynamical systems using data
● Learn the fundamentals of linear system identification theory
● Learn the fundamentals of model predictive control
● Learn the fundamentals of behavioral systems theory
● Learn the fundamentals of Koopman theory
Instructor: Erfan Nozari (erfan.nozari@ucr.edu)
Lecture Times & Location: M & W 9:30 - 10:50am Pacific Time via Zoom (link announced on iLearn)
Office Hours: W 11 - 11:50am Pacific Time via Zoom (same link as lectures)
Textbooks (optional):
● Lennart Ljung, “System Identification, Theory for the User”, 2nd Ed.
● Grune & Pannek, “Nonlinear Model Predictive Control, Theory and Algorithms”
● Mauroy et. al., “The Koopman Operator in Systems and Control”
● Markovsky & Dörfler, “Behavioral systems theory in data-driven analysis, signal processing, and control”, 2021
Grading: homeworks (20%), final project (80%)
Homeworks: 8 sets, about one per week; due on Wednesdays at midnight via iLearn
A syllabus with full schedule can be found here.