Cooperation in Multi-Agent Reinforcement Learning