Towards efficient long-horizon decision-making using automated structure search method of hierarchical reinforcement learning for edge artificial intelligence.