Video-to-Sim-to-Real Pipeline Trains Apollo 2 for Cornhole

In an internal test, Google DeepMind researchers used Gemini 4 Argon to analyze a video of a person playing cornhole, build a MuJoCo simulation, and calculate physics and whole-body control for Apptronik’s Apollo 2 humanoid. The article says the robot then made seven consecutive shots in the real world. DeepMind describes the pipeline as reducing modeling work from weeks to a few hours.