ICML-98 Submission #184

Multiagent Reinforcement Learning: Theoretical Framework and an Algorithm

	Hu and Michael P. Wellman
	Artificial Intelligence Laboratory 
	University of Michigan 
	Ann Arbor, MI 48109-2110, USA 

Abstract

	In this paper, we propose adopting general-sum stochastic games as
	a framework for multiagent reinforcement learning. Our work extend
	previous work by Littman to a much broader framework. We design a
	multiagent Q-learning method under our framework, and prove that 
	it converges to the Nash equilibrium of the game.

Keywords: Reinforcement learning, multiagent learning

Email address of contact author: junling@umich.edu

Phone number of contact author: 734/763-1563