Never Explore Repeatedly in Multi-Agent Reinforcement Learning