Research Project

Essex Robotics Logo

Scaling Reinforcement Learning to Multi-Robot Systems

Members:

Dongbing Gu (PI)
E. Yang

Project Description:

This research addresses an investigation on reinforcement learning in multi-robot systems under the framework of stochastic games. The range involves a variety of areas, including multi-robot systems, reinforcement learning, behaviour-based robotics, fuzzy logic classifier, and game theory. The research is to develop Q-learning algorithms for co-ordination games with better understanding of their theoretical and practical aspects. It will provide new insights for multi-robot systems learning in the fundamentals of efficacy of co-learning and emergence of co-operative behaviours. It is also expected to extend the outcomes of handling large learning spaces and uncertainties to the learning in general agent-based systems. Through implementation of the research, it will lay the foundation to further study on this area and to more complex applications.

Funding:

This project is funded by EPSRC (92k) (2003-2005)

Related Groups:

Michael L. Littman's Homepage
Craig Boutilier's Web Page
Michael Bowling's Homepage
The Multi-Agent Systems Laboratory at U.Mass
Prof. Ron Sun's Homepage
Dr. Junling Hu
Michael Wellman Home Page
Peter Stone's Robot Soccer Page at The University of Texas at Austin
University of Michigan AI Lab Home
Kevin M. Passino's Home Page
Sandip Sen's Home Page

Useful Links:

Reinforcement Learning Warehouse
Game Theory
The UT Austin Villa Robot Soccer Teams
The RoboCup Soccer Simulator
Global and Local Optimization
Convex Optimization
Multi-Agent Systems Research Group: Selected Publications
American Association for Artificial Intelligence (AAAI)
Learning Robot Reading Group
Machine Learning, Tom Mitchell, McGraw Hill, 1997
The Machine Learning Network Online Information Service
Reinforcement Learning Repository at UMass, Amherst
The Reinforcement Learning Toolbox
Tekkotsu Homepage, An open source project created and maintained at Carnegie Mellon University
Open-R SDK for SONY Robots
Markov Decision Process (MDP) Toolbox for Matlab

Journals:

Machine Learning
Autonomous Agents and Multi-Agent Systems, Kluwer Academic Publishers
Cognitive Systems Research, published by Elsevier
Applied Intelligence
Robotics and Autonomous Systems, published by Elsevier
Journal of Robotic Systems
Journal of Artificial Intelligence Research (JAIR)
Systems, Man and Cybernetics, Part C, IEEE Transactions on
Fuzzy Systems, IEEE Transactions on
Neural Networks
Optimal Control - Applications and Methods
Automatica
IEEE Control Systems Magazine
IEEE Transactions on Automatic Control
IEEE Transactions on Control Systems Technology
IEE Proceedings - Control Theory and Applications
International Journal of Control
International Journal of Robust and Nonlinear Control
Journal of Dynamic Systems, Measurement and Control
Systems & Control Letters
Control Engineering Practice