Exploiting Semantic Epsilon Greedy Exploration Strategy in Multi-Agent Reinforcement Learning