|
Professor: Rajesh Rao email: rao at cs.washington.edu office hours: By appointment |
TA:
Mike Chung email: mjyc at cs.washington.edu office hours: Tuesdays 1:30-3:30pm |
Your mission is to program a humanoid robot
to imitate human actions and learn new skills from human demonstration using
a Kinect RGB+depth camera. You will work as
part of a team to tackle the various sub-problems of (1) human motion capture
and interpretation from video, (2) control of a humanoid robot, and
(3) application of probabilistic reasoning and machine learning algorithms to
the problem of learning from human demonstration. There will three teams of 4-5 students each. Two warm-up
projects will help prepare you for a final course project of your choice.
Administrative
Email List: The
class mailing list is cse481c_au11@uw.edu.
You should be subscribed if you are registered. We
will use this list to make official class-related announcements.
Course wiki is here.
Grading
Syllabus
and Course Goals
The overall goal is to design and
implement a robotic system that can learn from human demonstration. This will involve
learning to write software for controlling a humanoid robot (the NAO) using a
Kinect RGB+depth camera. You will learn to use the
Kinect SDK as well as the NAO SDK. You will gain experience in applying machine
learning and probabilistic reasoning algorithms to concrete problems in 3D
vision and robotics. Specific course topics include:
Inverse kinematics for robot joint
control
Dimensionality reduction techniques
(e.g., PCA)
Regression and function
approximation
Classification of gestures and
actions
Probabilistic inference and learning
Last modified: September, 2011