Ornstein–Uhlenbeck Process Saw this at Dyna Robotics. And a bunch of papers use this. For example: What Matters for Batch Online Reinforcement Learning in Robotics HuB Learning Extreme Humanoid Balance