ManipulationHardware: MaxRoadmap scaffold
Diffusion Policy
Robust manipulation via denoising diffusion over action sequences.
Lab runtime / ManipulationOK / VIS-01
Diffusion Policy runs against the same observable edge stack used for testing and deployment.
This entry is a roadmap scaffold, not a finished demo. Its description records intended scope; the linked repository is the source of truth for current code.
About this demo
Diffusion-policy implementation that generates action sequences instead of single actions, dramatically improving robustness on contact-rich and multi-modal tasks. Trained on real-to-sim captured environments.
Highlights
- →Action-chunking for robustness
- →Multi-modal task handling
- →Real-to-sim trained
- →ONNX export for edge inference
Supported robots
Franka PandaUR5e
Related demos
View all →Manipulation
LeRobot ACT policy
Hugging Face LeRobot framework — train and deploy with one config.
Open demo →ManipulationWarehouse pick-and-stow
Cluttered-bin manipulation with F/T wrist sensor on Max.
Open demo →TeleopCockPit
Predefined teleop dashboard with dual cameras, 3D model, map, and controls.
Open demo →