Joint Strategy Synthesis for Multi-Agent Games of Imperfect Information against Nature

Abstract: In this paper, we investigate how agents of a multi-agent game of imperfect information may coordinate their individual strategies beforehand in order to create a better strategy for the team. The work is in the area of game theory and focuses on the previously-introduced idea of epistemic expansion. Epistemic expansion is a tool for creating simplified models of games such that strategies can be synthesised in these models. In this paper, we develop and present a new, modified definition of epistemic expansion that allows for strategies to be translated from the simplified models back to the original games. Finally, we prove that epistemic expansion satisfies the important property that any winning strategy in the simplified model can be translated into a winning strategy in the original game.