Learning to Run with Potential-Based Reward Shaping and Demonstrations from Video Data | IEEE Conference Publication | IEEE Xplore