A review of cooperative multi-agent deep reinforcement learning