Improving Reinforcement Learning Efficiency with Auxiliary Tasks in Non-Visual Environments: A Comparison