Learning multiagent coordination in the absence of communication channels

Aaron Goodman

In this work, we develop a reinforcement learning protocol for a multiagent coordination task in a discrete state and action space: an iterated prisoner's dilemma game extended int…