16-714 Advanced Control for Robotics

Time: Monday and Wednesday 3:30 pm — 4:50 pm
Location: PH A21
Instructor: Changliu Liu (cliu6@andrew.cmu.edu)
Instructor Office Hours: Wednesday 2-3pm, NSH 4525
Canvas: https://canvas.cmu.edu/courses/42681
Piazza: https://piazza.com/class/m05auttsnij1v5

Course Material

Slides and Notes

Code Examples

Selected Student Projects

Course Description

This course discusses advanced control algorithms that enable robots to behave more intelligently. It is designed for students, primarily graduate-level but open to talented undergraduates, interested in advanced control.

Key Topics

  • Model predictive control
  • Adaptive control
  • Learning control
  • Lyapunov theory

Prerequisites

  • Familiarity with basic control theory and robotics concepts
  • Proficiency in Matlab for examples and assignments
  • Creativity and enthusiasm are essential

Course Goals

Familiarize students with advanced control algorithms and their applications in solving robotics problems.

Assessment Structure

  • Homework: 80%
  • Presentation: 15%
  • Participation: 5%

Policies

Class Attendance

Participation and attendance are critical components of the course. Notify the instructor ahead of time if you must miss a class.

Use of Canvas

Lecture slides, supplementary materials, homeworks, and solutions will be posted on Canvas. Students should review slides prior to lectures.

Homeworks

  • Number of Assignments: 5, each worth 16 points.
  • Submission: Typed in Word or LaTeX, submitted in PDF format via Canvas.
  • Self-Grading: Students will self-grade using provided solutions and grading instructions. Random checks will be conducted by the instructor.
  • Late Submission: Deadlines are firm. Contact the instructor early in case of emergencies.

Presentation

In the final two lectures, students will present an advanced control method as part of a team project. Topics include:

  • Safe control
  • Koopman control

Weekly Schedule

Week 1

  • 8/26: Lecture 1 - State-Space Models
  • 8/28: Lecture 2 - Kinematic and Dynamic Models (Vehicles, Manipulators)

Week 2

  • 9/2: Labor Day (No Class)
  • 9/4: Lecture 3 - Matlab Tutorial

Week 3

  • 9/9: Lecture 4 - System Properties
  • 9/11: Lecture 5 - Dynamic Programming
  • 9/13: Homework 1 Released (Optimal Control)

Week 4

  • 9/16: Lecture 6 - Maximum Principle
  • 9/18: Lecture 7 - Linear Quadratic Regulator
  • 9/27: Homework 1 Due; Homework 2 Released (MPC)

Week 5

  • 9/23: Lecture 8 - Infinite Horizon Optimal Control
  • 9/25: Lecture 9 - Model Predictive Control

Week 6

  • 9/30: Lecture 10 - MPC Feasibility
  • 10/2: Lecture 11 - MPC Stability

Week 7

  • 10/7: Lecture 12 - Iterative Learning Control (Frequency Domain)
  • 10/9: Lecture 13 - Iterative Learning Control (Time Domain)
  • 10/11: Homework 2 Due; Homework 3 Released (ILC & KF)

Week 8

  • 10/14-10/16: Fall Break (No Classes)

Week 9

  • 10/21: Lecture 14 - Probability and Least Square
  • 10/23: Lecture 15 - Kalman Filter

Week 10

  • 10/28: Lecture 16 - EKF and UKF
  • 10/30: Lecture 17 - Recursive Least Square
  • 11/1: Homework 3 Due; Homework 4 Released (RLS & LQG)

Week 11

  • 11/4: Lecture 18 - Separation Principle
  • 11/6: Lecture 19 - Model Reference Adaptive Control

Week 12

  • 11/11: Lecture 20 - Dual Control
  • 11/13: Lecture 21 - Value Approximation
  • 11/15: Project Proposal Due; Homework 4 Due; Homework 5 Released (Policy Gradient)

Week 13

  • 11/18: Lecture 22 - Policy Gradient
  • 11/20: Lecture 23 - Actor Critic

Week 14

  • 11/25: Lecture 24 - Lyapunov Methods
  • 11/27: Thanksgiving Break (No Class)

Week 15

  • 12/2: Advanced Topics: Safe Control (Safe Control with Learned Certificates: A Survey of Neural Lyapunov, Barrier, and Contraction Methods for Robotics and Control)
  • 12/4: Advanced Topics: Koopman Control (Koopman Operators for Estimation and Control of Dynamical Systems)
  • 12/6: Homework 5 Due

Additional Policies and Resources

For more information, refer to the full course policies and syllabus.