CSE 571: AI-Robotics

Spring 2023

Tue & Thu 11:30-12:50 @ MEB 242

Instructor:

Abhishek Gupta
Office hours: 4-5 PM Fri
Gates 215

Teaching Assistants:

  • Yi Li
    Office hours: 3-4 PM Thu
  • yili18@cs.washington.edu
    Gates 152  
  • Srivatsa GS
    Office hours: 4-5 PM Mon
  • vatsa@uw.edu
    Gates 152  


Discussion:
Canvas: https://canvas.uw.edu/courses/1631624

Please access Zoom class lectures and recordings via Canvas.

Discussion board: https://edstem.org/us/courses/38159/discussion/

Use this board to discuss the content of the course. Feel free to discuss homeworks, projects, and any confusion over topics discussed in class. It is also acceptable to ask for clarifications about the statement of homework problems, but not about their solutions.


Textbook:
There is no required textbook for the class. Many of the lectures and homework assignments will have associated papers and chapters from:

Probabilistic Robotics, S. Thrun, W. Burgard, and D. Fox., MIT Press, Cambridge, MA, September 2005.


Homeworks:
There will be 3 homeworks with programming components done in Python.

Late policy: You have a total of 8 late days for the whole quarter. But each assignment may be handed in up to 5 days late, after which a penalty of 20% of the maximum grade is applied per day.

While we encourage students to discuss homeworks, each student must write up their own solution. It’s fine to use a source for generic algorithms (with attribution), but it is not allowed to copy solutions to the problems. Additionally, students may not post their code online. If we determine that a student posted their code online, they will get an automatic 50% reduction on the entire assignment (math + code) and if they copy code for the problems from another student or from online, they will get an automatic 0% for the entire assignment (and possibly reported to the college).

Projects:
The projects will be done in teams of 1-2 and will account for 50% of your grade.
Project can be investigating any question related to robotic learning, reinforcement learning, imitation learning.
Here we list several ideas as an example PDF.
We encourage ideas related to robotics and your own research.

Final projects will be presented in a poster session at CSE2 G04.

Grading:
You will submit your assignments (both writeups and code) to canvas. You have all been enrolled into the class. Please check your email.
  • Homeworks 45% (15%x3 covering estimation, control and planning, and end to end learning)
  • Participation (5%)
  • Final Project (50%)


Anonymous Feedback:
Please send your feedback via https://feedback.cs.washington.edu/


Course Outline

Lectures

Date Topic Slides & Recode Reading (textbook/papers) Homeworks & Projects
Section 1: Sensing the World - - -
Tue 28-Mar Introduction slides, record Elephants, End-to-end policy -
Thu 30-Mar Notation definitions and Probability review / Bayes filter slides, record Book chapter 2 -
Tue 4-Apr Kalman filtering and smoothing slides, record Book Chapters 3 and 7, Deep KF, BackProp KF
Thu 6-Apr Extended Kalman Filter / Unscented Kalman Filter slides, record Book Chapters 3 and 7, Comparison, New EKF, New UKF -
Tue 11-Apr Particle filtering slides, record Book Chapters 4 and 8, Adaptive PF, PF in Robotics, Realtime PF -
Thu 13-Apr Motion / Sensor models and Applications to Filtering slides, record Book Chapter 5 and 6, Landmark-based sensor model HW1 release
Tue 18-Apr Occupancy Mapping slides, record Book Chapter 9, OctoMap, An Example Using Grid Maps, Another Example -
Thu 20-Apr Simultaneous Localization and Mapping slides, record Book Chapter 10, EKF-SLAM, EKF-SLAM v.s. Optim-based SLAM Project proposal due
Tue 25-Apr SLAM II slides, record Book Chapter 11 and 12, Factor Graphs, GTSAM -
Section 2: Acting in the World - - -
Thu 27-Apr Deterministic Planning - Search Methods slides, record A* optimality proof, LifeLong Planning A*, D* Lite -
Tue 2-May Sampling Based Motion Planning slides, record Quality Analysis, BIT, RRT, Non-holonomic RRT HW1 due, HW2 release
Thu 4-May Linear Quadratic Control - LQR/DDP slides, record Iterative LQR, Guided Policy Search -
Tue 9-May Lyapunov Analysis slides, record Basic Lyapunov Theory, Notes -
Section 3: Joint Sensing and Acting - - -
Thu 11-May Markov Decision Processes AND Value Iteration slides, record LACTO, TrajOPT, TrajOpt Tutorial -
Tue 16-May Behavior Cloning and Policy Gradient slides, record No-Regret, ALVINN, Imitation Learning, Connectionist RL, Variance Reduction -
Thu 18-May Actor Critic slides, record Actor-Critic, Soft AC, Double DQN, Continuous DQN HW2 due, HW3 release
Tue 23-May Model Based Reinforcement Learning slides, record PILCO, MBMF, PDDM, PETS, MOReL, MPPI -
Section 4: Frontier Research - - -
Thu 25-May Guest Lecture: Lerrel Pinto - - -
Tue 30-May Guest Lecture: Guanya Shi - - -
Thu 1-Jun Guest Lecture: Pulkit Agrawal - - HW3 due
Tue 6-Jun Final Project Presentation at CSE2 G04 - - -