Advanced introduction to the theory of optimal control of time-varying and time-invariant linear systems; Solutions to the linear-quadratic regulator, optimal filtering, and linear-quadratic-gaussian problems; Robustness analysis and techniques to enhance robustness of controllers.
Why take this course?
This course provides a rigorous introduction to linear optimal control systems. The emphasis is on being able to derive the fundamental properties of these systems as well as to provide an introduction to important theoretical tools such as dynamic programming and least-squares optimization in function spaces.
ECE 5744 or ECE 5754 or ME 5544 or ME 5554 or AOE 5744 or AOE 5754.
Major Measurable Learning Objectives
Derive the solutions to the Linear-Quadratic Regulator problem and the Linear-Quadratic-Gaussian problem.
Use linear optimal control techniques to solve problems such as disturbance rejection and tracking.
Test the robustness of linear control systems to unstructured uncertainty and implement methods to improve robustness.
Derive the basic equations of the Kalman filter and describe the important stochastic properties of this filter.
Percentage of Course
1. Introduction and Review
2. Linear Quadratic Regulator (LQR)
a. Derivation from Dynamic Programming Theory
b. Derivation from Least Squares Theory
c. Penalty Matrix Selection
d. Application to Disturbance Rejection and Tracking
a. Singular Values and the Multivariable Nyquist Test
b. Gain and Phase Margin Properties of LQR
c. General Uncertainty Bounds
4. Kalman Filtering
a. Stochastic Dynamical Systems
b. Derivation as Linear, Minimum-Variance Estimator
5. Linear-Quadratic-Gaussian Control
a. Stochastic Dynamic Programming
b. Derivation of LQR with Additive or Multiplicative Noise
c. Separation Principle
d. Loss of Robustness and Loop-Transfer Recovery
e. Approaches to Robustness with Structured Uncertainty