CS 295-Z Robot Learning and Autonomy
From Brown University Robotics
CS295-Z Robot Learning and Autonomy
Spring Semester 2010
Instructor: Prof. Chad Jenkins
This course attempts to address the question "What are the driving applications of robotics?" How will robots move out of structured laboratory settings into real-world applications where a diversity of users, environments, and tasks abound? How will robots become the path of least resistance for managing physical environments, similar to how computing has become the preferred tool for digital environments? How should robots be programmed? Can robot learning be used to develop engineering solutions to practical problems, or is robot learning itself our primary motivation?
Towards this end, CS295-Z is a seminar course that covers current research topics related to perceiving and acting in the real world, emphasizing robot learning from demonstration (LfD) and physical object manipulation. These topics will be pursued through independent reading, class discussion, and project implementations. Papers covered will be drawn mostly from robotics and machine learning as well as related domains, such as computer vision, neuroscience, and animation. Special emphasis will be given to learning autonomous robot control from human demonstration and video game style interfaces.
Grading for individual enrolled students is broken down as follows:
Students are expected to attend all class meetings (unless an exception is given beforehand), actively participate in discussion, present (at least) 2 papers to the class, and significantly contribute towards the development and implemenation of a final project.
Students are expected to send summaries of papers they are not presenting to the course mailing list (cs295-z@list) prior to the corresponding class meeting.
For paper presentations, student presenters must have a rough draft prepared and consult with the instructor at least 2 days before the presentation date.
Previous versions of this course
Each class meeting will consist of 2 paper presentations given by students. This should take between 1-2 hours. The remaining time will be devoted to a collaborative hacking session to prototype, implement, and evaluate new ideas.
2/1 Course Overview
2/8 Surveys of Robot LfD and Manipulation
2/15 Mixed-initiative LfD
2/22 No class (Long weekend)
3/1 Reinforcement Learning with Human Guidance
3/8 Inverse Reinforcement Learning
3/15 Behavior-based and Layered Learning
3/22 Mobile Manipulation: Grasp Types and Planning
3/29 No class (Spring Recess)
4/5 Mobile Manipulation: Domestic Service Systems
4/12 Mobile Manipulation: Compliant Grasping and Assembly
4/19 Manipulation: Developmental Robotics
4/26 Experiments with Human Subjects and Signifance Testing (papers TBD)
5/3 Final project demos