TWIST (whole-body teleoperation)

TWIST is a whole-body teleoperation system that lets a Unitree G1 humanoid imitate human motion captured from a person.

Licenses

Code: MIT (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (MIT) at commit 42d8c13 (opens in a new tab), on 16 Oct 2026.
Commercial use: Unknown

It needs closed parts under their own terms: read them before any commercial use.

One part it needs is closed

Needs NVIDIA Isaac Gym: TWIST trains its policies in NVIDIA Isaac Gym, a closed NVIDIA download (README).

Trains a teacher policy with reinforcement learning, then a student policy with RL plus behavior cloning, from a provided motion dataset.
Exports the trained student policy to a TorchScript (jit) model for deployment.
Provides a pretrained general motion tracker checkpoint that can be deployed directly.
Splits control into a high-level motion sender (teleop) and a low-level RL controller, connected through Redis.
Supports simulation checks before running on hardware, then deployment on a physical Unitree G1 over Ethernet.
Real-time motion retargeting is handled by the separate GMR project; the included teleop script is a legacy version.

Written by Robotheca from the project's README (opens in a new tab) at commit 42d8c13.

Difficulty: Advanced
Hardware it needs, as the project states it: Training runs on a single Nvidia RTX 4090 with 24 GB of memory in 1 to 2 days. Real-robot deployment needs a Unitree G1 connected to a laptop by Ethernet cable, and a Redis server on the computer.
How to start: The project's website (opens in a new tab)

Checked on GitHub on 16 Oct 2026.

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.