Multiagent reinforcement learning and self-organization in a network of agents