Q-Decomposition for Reinforcement Learning Agents Russell Zimdars 2003

This is a paper provided by Russell as an example of 'good' multiagent work. It is full yanalyzable, and ultimately reducable to an central optimal controller. I didn't read this paper.