They trained the model in a simulation environment using reinforcement learning until it had a broad understanding of how to articulate and move many parts of a robot body. They then implemented ...