Chinese researchers developed a reinforcement learning framework that enables humanoid robots to stand up in diverse real-world scenarios