SDM5008 (F24) Advanced Control for Robotics

Course Objectives

  • Develop a solid foundation in robot modeling, control and learning to conduct cutting edge research in robotics
  • Math: Probability, optimization, stochastic systems
  • Modeling: Advanced rigid body kinematics and dynamics
  • Control: Differential IK, Model predictive control
  • Learning: MDP, Deep Reinforcement learning

Course Information

  • Course ID: SDM5008
  • Course Name: Advanced Control for Robotics
  • Credit: 3.0
  • Credit Hours: 48.0
  • Classroom: Classroom 524, Teaching Building One
  • Time: Week 1-16, Wednesday 19:00 pm - 22:00 pm;
  • Teacher: Wei Zhang(zhangw3@sustech.edu.cn)
  • TA: Jiaqi Song(12433008@mail.sustech.edu.cn), Haokai Su(12433009@mail.sustech.edu.cn)

Target Students

  • PhD students in robotics with a strong need of advanced control
  • Master students in control and robotics with a strong desire to pursue PhD degree
  • Students who can devote substantial time to read and learn outside classroom

Tentative Outline

  1. Advanced Kinematics and Dynamics
    1. Rigid body configuration and velocity
    2. Exponential coordinate of rigid body motion
    3. Kinematics of open chain
    4. Velocity Kinematics
  2. Basic Robot Control
    1. Basics of optimization
    2. Differential IK
    3. Introductory optimal control
    4. Model Predictive Control
  3. Reinforcement learning
    1. Probability review
    2. Markov Decision Process
    3. Basics of Neural Networks
    4. Value estimation via sampling
    5. Introduction to Policy gradient
    6. Policy gradient with baseline
    7. Advanced policy gradient
  4. Advanced topics (if time permits)
    1. Transformer
    2. robot dynamics

Prerequisite

  • Undergraduate introductory robotics course
    • Solid background in rigid body kinematics and dynamics in the level described in standard beginner textbooks, such as:
      • “Introduction to Robotics: Mechanics and Control”, J. Craig
      • “Robot Modeling and Control”, M. Spong, S. Hutchinton, and M. Vidyasagar
  • Undergraduate class in control
  • Maturity in math, solid understanding in linear algebra (see tutorial notes in linear algebra), good at abstract reasoning
  • Programming skills: Python

Lecture Notes

  • Lecture Note 1: Rigid body configuration [PDF] [Notes]
    Rigid Body Configuration, Rigid Body Velocity(Twist), Geometric Aspect of Twist: Screw Motion
  • Lecture Note 2: Operator view of rigid body operation [PDF] [Notes]
    Matrix exponential, rotation operator, rigid body operator, rigid body operator for screw axis
  • Lecture Note 3: Exponential coordinate of rigid rody configuration [PDF] [Notes]
    Exponential coordinate of SO(3), Euler Angles and Euler-Like Parameterizations, Exponential coordinate of SE(3)
  • Lecture Note 4: Instantaneous Velocity of Moving Frames [PDF] [Notes]
    Instantaneous Velocity of Rotating Frames, Instantaneous Velocity of Moving Frames
  • Lecture Note 5: Product of Exponential and Kinematics of Open Chain [PDF] [Notes]
    Motivating Example, Product of Exponential Formula Derivations, Practice Example
  • Lecture Note 6: Velocity Kinematics: Geometric and Analytic Jacobian of Open Chain [PDF] [Notes]
    Geometric Jacobian Derivations, Analyic Jacobian
  • Lecture Note 7: Rigid Body Dynamics [PDF] [Notes]
    Spatial Acceleration, Spatial Force, Spatial Momentum
  • Lecture Note 8: Mujoco Tutorial [PDF] [Notes]
    Short introduction to simulation, Introduction to Mujoco, Python Example
  • Lecture Note 9: Probability Review for Reinforcement Learning [PDF] [Notes]
    Probability and Conditional Probability, Random Variables and Random Vectors, Jointly Distributed Random Vectors and Conditional Expectation
  • Lecture Note 10: Markov Decision Process for Reinforcement Learning [PDF]
    Markov chain, Markov decision process, Bellman equations, Simulations

Homework

MISC

Background Reading

  • “Linear Algebra Review and Reference”, Zico Kolter
  • “The Matrix Cook Book” - Kaare Brandt Petersen, Michael Syskind Pedersen
  • Linear algebra
  • Matrix exponential

Reference