G1 MOVES
60 motion capture clips for the Unitree G1 humanoid robot
1
Capture
2
Retarget
3
Train
4
Policy
60 motion capture clips for the Unitree G1 humanoid robot
G1 Moves is a dataset of 60 motion capture clips for the Unitree G1 humanoid robot (29 DOF, edition EDU). Each clip flows through a four-stage pipeline — from human performance to autonomous robot motion.
Primarily captured with MOVIN TRACIN markerless mocap (LiDAR + vision, 60 FPS, BVH output). Alternatively, video2robot extracts motion from any camera source — smartphone video, YouTube, or AI-generated footage.
Per-frame IK via MOVIN-SDK-Python maps human skeleton to G1 29-DOF joint limits. Motion-Player-ROS provides real-time preview in RViz.
mjlab PPO across 8,192 parallel MuJoCo-Warp envs. 4-layer MLP actor (160→512→256→128→29). 14-body tracking with adaptive early stopping.
RoboJuDo sim2real framework. MuJoCo sim2sim validation, then deployment to physical G1 via Unitree SDK over Ethernet. Hardware E-stop required.
The complete motion-to-policy pipeline uses six tools. Every trained policy is available as an ONNX model (160-dim input, 29-dim output) with baked-in observation normalization.
If you use G1 Moves in your research or project, please cite:
Download Whitepaper (PDF)
@misc{chaiet2026g1moves,
title = {G1 Moves: A Motion Capture Dataset for Unitree G1 Humanoid Robot},
author = {Chaiet, Mitch},
year = {2026},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/exptech/g1-moves},
note = {60 motion capture clips with trained RL policies for sim-to-real transfer}
}WARNING: Deploying learned locomotion policies to physical robots is inherently dangerous and can cause serious injury, death, or property damage. The policies provided in this dataset are trained in simulation and have not been validated for safe real-world operation. Sim-to-real transfer involves unpredictable failure modes including but not limited to: sudden loss of balance, uncontrolled high-speed limb movements, unexpected falls, collisions with people or objects, and hardware damage.
Do not deploy these policies to a physical robot unless you have: (1) advanced expertise in sim-to-real transfer, robot safety systems, and real-time control; (2) appropriate physical safety infrastructure including overhead tethers, safety enclosures, and emergency stop mechanisms; (3) completed thorough sim-to-sim validation; (4) conducted a formal risk assessment for your specific deployment environment; and (5) ensured all personnel in the vicinity are trained on emergency procedures.
LIMITATION OF LIABILITY: THE POLICIES, MODELS, DATA, AND SOFTWARE IN THIS DATASET ARE PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, SAFETY, OR NONINFRINGEMENT. IN NO EVENT SHALL EXPERIENTIAL TECHNOLOGIES, ITS AFFILIATES, CONTRIBUTORS, OR LICENSORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, CONSEQUENTIAL, OR PUNITIVE DAMAGES (INCLUDING BUT NOT LIMITED TO PERSONAL INJURY, DEATH, PROPERTY DAMAGE, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES, LOSS OF USE, DATA, OR PROFITS, OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS DATASET OR ITS DERIVATIVES, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
By downloading, using, or deploying any part of this dataset, you acknowledge that you assume all risk and full responsibility for any consequences arising from its use. You agree to indemnify and hold harmless Experiential Technologies, its officers, directors, employees, and agents from any claims, damages, or liabilities arising from your use of these materials.
