Learning Robot Reading Group

organizer: Michael L. Littman
with support from the MetroBots robot soccer team.

Current meeting time: Friday 4:30-6:30
Current meeting place: CBIM lab

Future Meetings

April 25: Last meeting. Least Squares Policy Iteration (pdf). Questions.

May 2: No class. American Open .

Past Meetings

January 24: Introduction to reinforcement learning. Markov games as a framework for multi-agent reinforcement learning.

January 31: Discussion of Aibo software environment. CMU's RoboCup team description.

February 7: Generalization in reinforcement learning. Practical issues in temporal reinforcement learning , Generalization in reinforcement learning: Safely approximating the value function , Generalization in reinforcement learning: Successful examples using sparse coarse coding , Scaling reinforcement learning toward RoboCup soccer . Questions.

February 14: Partially observable environments. Acting optimally in partially observable stochastic domains , Acting under uncertainty: Discrete Bayesian models for mobile-robot navigation . Questions.

February 21: POMDP RL. Reinforcement learning with perceptual aliasing: The perceptual distinctions approach. Questions.

February 28: POMDP RL 2. Instance-based state identification for reinforcement learning . Questions.

March 7: Hierarchical learning State abstraction and discounting in hierarchical RL .

March 14: Large-scale learning Using machine learning techniques in complex multi-agent domains .

March 21: Spring break.

March 28: RL Theory. Paper by John Langford . Questions.

April 4: Continuous space Memory-based RL. Practical reinforcement learning in continuous spaces . Questions.

April 11: Continuous space Memory-based RL 2. Representations for learning control policies , Memory-based regression tutorial . Questions.

April 18: (Michael out of town).

I will be hosting weekly meetings this spring for people interested in studying the use of machine-learning algorithms in robotics. Each week we will discuss a research paper on topics such as vision, locomotion, route planning, localization, and manipulation. We will also use our legged Sony Aibo robots for class projects.

Interested people should send me email (mlittman@cs.rutgers.edu) and I will create a mailing list for discussing meeting times and topics. Students wishing to receive academic credit should contact me to discuss options (independent study or light seminar 198:500).

Hope you can join us!