“John Schulman developed the TRPO and GAE algorithms during his PhD as part of a research goal to achieve 3D humanoid locomotion using reinforcement learning.”