Markov Games as a Framework for Multi-Agent Reinforcement Learning Export

@inproceedings{Littman:1994, abstract = {In the Markov decision process (MDP) formalization of reinforcement learning, a single adaptive agent interacts with an environment defined by a probabilistic transition function. In this solipsistic view, secondary agents can only be part of the environment and are therefore fixed in their behavior. The framework of Markov games allows us to widen this view to include multiple adaptive agents with interacting or competing goals. This paper considers a step in this direction in which exactly two agents with diametrically opposed goals share an environment. It describes a Q-learning-like algorithm for finding optimal policies and demonstrates its application to a simple two-player game in which the optimal policy is probabilistic. 1 INTRODUCTION No agent lives in a vacuum; it must interact with other agents to achieve its goals. Reinforcement learning is a promising technique for creating agents that co-exist [Tan, 1993, Yanco and Stein, 1993] , but the mathematical framework that just...}, author = {Littman, Michael L.}, booktitle = {In Proceedings of the Eleventh International Conference on Machine Learning}, citeulike-article-id = {5722046}, citeulike-linkout-0 = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.48.8623}, keywords = {abm, article, cas, learning, markov, multiagent, reinforcement}, pages = {157--163}, posted-at = {2009-10-20 08:27:06}, priority = {2}, title = {Markov Games as a Framework for Multi-Agent Reinforcement Learning}, url = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.48.8623}, year = {1994} }

TY - CONF ID - Littman:1994 L3 - citeulike-article-id:5722046 N2 - In the Markov decision process (MDP) formalization of reinforcement learning, a single adaptive agent interacts with an environment defined by a probabilistic transition function. In this solipsistic view, secondary agents can only be part of the environment and are therefore fixed in their behavior. The framework of Markov games allows us to widen this view to include multiple adaptive agents with interacting or competing goals. This paper considers a step in this direction in which exactly two agents with diametrically opposed goals share an environment. It describes a Q-learning-like algorithm for finding optimal policies and demonstrates its application to a simple two-player game in which the optimal policy is probabilistic. 1 INTRODUCTION No agent lives in a vacuum; it must interact with other agents to achieve its goals. Reinforcement learning is a promising technique for creating agents that co-exist [Tan, 1993, Yanco and Stein, 1993] , but the mathematical framework that just... EP - 163 TI - Markov Games as a Framework for Multi-Agent Reinforcement Learning T2 - In Proceedings of the Eleventh International Conference on Machine Learning SP - 157 KW - abm KW - article KW - cas KW - learning KW - markov KW - multiagent KW - reinforcement AU - Littman, Michael PY - 1994/// UR - http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.48.8623 ER -