Low Entropy Communication in Multi-Agent Reinforcement Learning