State based potential games

There is a growing interest in the application of game theoretic methods to the design and control of multiagent systems. However, the existing game theoretic framework possesses inherent limitations with regards to these new prescriptive challenges. In this paper we propose a new framework, termed state based potential games, which introduces an underlying state space into the framework of potential games. This state space provides a system designer with an additional degree of freedom to help coordinate group behavior and overcome these limitations. Within the context of state based potential games, we characterize the limiting behavior of two learning algorithms termed finite memory better reply processes and log-linear learning. Lastly, we demonstrate the applicability of state based potential games on two cooperative control problems pertaining to distributed resource allocation and the design of local and distributed control laws. (C) 2012 Elsevier Ltd. All rights reserved.