Reinforcement Learning with Internal Reward for Multi-Agent Cooperation: A Theoretical Approach - EUDL