Global games with Poisson observations: Bio-inspired distributed coordination of multi-agent systems

Abstract: Global games are a class of incomplete information games where the payoffs exhibit strategic complementarity leading to an incentive for the agents to coordinate their actions. Such games have been used to model scenarios in many socioeconomic phenomena, where the private signals available to the agents are typically assumed to be Gaussian. We study an instance of a global game where the agents observe Poisson random variables, which are inspired by applications in microbiology where information signals are disseminated via discrete molecular signals rather than continuous. Although this observation model violates the essential technical assumptions present in the Gaussian case, we present preliminary results on the existence of Bayesian Nash equilibria in pure threshold policies in two variants of the underlying random state-of-the-world: an arbitrarily distributed discrete binary state and a continuous state with uniform distribution.