Multiagent Deep Reinforcement Learning: Challenges and Directions Towards Human-Like Approaches