A game-theoretic approach to cooperation in multi-agent systems

Abstract

The development of semantic agent frameworks is strongly influenced by the means for representing and reasoning about knowledge. Nevertheless, when faced with decision-making in multi-agent environments, current approaches are limited in their ability to model strategic interactions. As a result, ontological approaches such as those based on OWL, offer only a reactive behavior for such interactions. We argue that a game-theoretic approach is more suitable when faced with modeling cooperation in Multi-Agent Systems. We introduce a new type of games, Boolean Games with Currency, which combine features of Boolean Games, with the transferable payoff setting. We also give a computational characterisation for a solution concept for coalitional stability: the core. We show that the core of a Boolean Game with Currency is always non-empty, and we prove the core membership problem to be co-NP complete.