This is amazing. Really impressive scaling up. Will you do a post on your DeepRL work? I'm particularly interested in what Python simulator you're referring to. I come from a Games AI, not a Robotics background and am trying to bridge the gap myself.