HumDex
physical-superintelligence-lab
A humanoid dexterous manipulation stack combining teleoperation, data collection and policy learning.
What you can explore
Explore the documented G1 and Wuji Hand simulation, hardware controllers and demonstration workflows.
Before you use it
The stack combines distinct environments and external whole-body controllers. Follow the supported versions and hardware configuration; it is not a general controller for every humanoid.
Original contributors
Liang Heng, Yihe Tang, Jiajun Xu, Henghui Bao, Di Huang and Yue Wang
License / access: Code: MIT · check external data, models and hardware terms separately
Keep exploring
Wuji SDK
Device discovery, streaming, recording and hand-retargeting interfaces for Wuji devices.
ACT / Action Chunking with Transformers
An imitation-learning implementation that predicts chunks of robot actions from demonstrations.
Open-TeleVision
Immersive robot teleoperation with active stereo visual feedback and hand-motion tracking.
Editorial reference · checked 2026-09-22