policy

UMI

Stanford / Columbia / TRI · PyTorch

or hover any field below to flag it

Overview

Name
UMI
Author
Stanford / Columbia / TRI
Framework
PyTorch
License
mit
Skill type
manipulation
Evidence level
verified
Task description
Universal Manipulation Interface for in-the-wild data collection and cross-robot policy transfer. Humans collect demos with handheld GoPro-equipped grippers using SLAM-based 3D tracking. Diffusion policy backbone with relative-trajectory action representation enables hardware-agnostic deployment without retraining.

Spaces

Action space
end-effector-pose · 7-dim · 10Hz
Observation space
  • type: multimodal
  • · wrist_gopro_fisheye_rgb (155deg)
  • · end_effector_trajectory
  • · gripper_width

Links

HuggingFace repo
null

Compatible robots

0+2 mentioned but not in catalog yet

No robots list UMI as compatible yet. Know of one? Flag it above.

Compatible environments

2

Datasets that reference this policy

0

No datasets reference UMI yet.