// readme

🦾 Embodied-SuperMimic

Let robotic arms learn in a world model, and verify in simulation.

🌐 Open the online experiment showcase

🎬 Watch the arm execute Β· πŸ”Ž Switch embodiments and policies Β· 🧠 Learn about the online world model Β· πŸ“– Check results and protocol

No installation needed, just open it and watch. Select Panda, ARX5, PiPER, UR5e, or xArm7, pause, drag the progress bar, and observe motions and contacts.

πŸŽ₯ Watch recordings Β· πŸš€ Run locally Β· πŸ“Š Experiment results Β· πŸ§ͺ Reproduction notes

An independent robot manipulation research project. Starting from each robot’s own task experts, it explores how online world models and policies can improve each other.

✨ What can you see here?

  • Real saved simulation recordings: hanging a mug and placing a lemon; five robots and eight policy demonstrations, filterable by embodiment and task with one click.
  • Actions of different policies: Teacher, BC initialization, and online Dyna are…