Coordination in multiagent reinforcement learning: a Bayesian approach