Unitree Sim Web — Next-Gen Unitree G1 Humanoid Simulation Platform

Welcome to Unitree Sim Web, the most advanced browser-based simulation platform for the Unitree G1 humanoid robot, developed by STrike Robot. Built on MuJoCo WASM, this project brings professional-grade physical simulation directly to your browser without the need for complex GPU installations or heavy backend software.

---

Key Features

• High Performance: Leverages WebAssembly to run MuJoCo physics directly in the browser with ultra-low latency.
• Real-time Joint Control: Intuitive interface for controlling individual joint groups (Limb-based control).
• Keyframe Recording: Record pose sequences, play them back, and export high-quality .webm videos.
• No Backend GPU Required: All physics calculations are performed on the client-side for seamless accessibility.
• Developer Ready: Includes built-in user guides and pose-tips for the Unitree G1 model.

---

REVEAL: Humanoid Policy Integration

A major breakthrough we are excited to showcase is the platform's support for Humanoid Policy Integration. This feature allows researchers and developers to load pre-trained control policies (e.g., from Reinforcement Learning) and observe their performance in a realistic physics environment.

Take a look at the demo video below to see the humanoid policy in action:

Video: Demo of Unitree G1 Humanoid Policy Integration (Click to watch on YouTube)

This feature bridges the gap between algorithm development and physical experimentation, providing a powerful sandbox for robot control testing.

---

Quick Start — Local Development

Prerequisites

Tool	Version
Node.js	≥ 18
Python	≥ 3.10

Installation Steps

Follow these steps to set up and run the simulation locally:

1. Install dependencies:

   npm install

2. Build the frontend bundle:

   npm run build

3. Start the secure HTTP server:

   python3 run_server.py

Once the server is running, open http://localhost:8000 in your browser (Chrome or Edge recommended for optimal performance).

Important

Technical Note: You must use run_server.py instead of the standard python -m http.server. This server is specifically configured with CORP and COEP security headers required for SharedArrayBuffer (essential for MuJoCo WASM).

---

Deploy with Docker (Recommended)

For a consistent and stable environment, we recommend using Docker Compose:

docker compose up -d

The application will be available at port 8000.

---

Project Structure

Unitree_Sim_web/
├── src/                # Core logic (Three.js scene, MuJoCo Utils)
├── assets/             # Robot models (XML, STL, Textures)
├── images/             # Media assets (Images and Videos)
├── run_server.py       # Secure HTTP server implementation
├── Dockerfile          # Container configuration
└── start.sh            # Quick startup script

---

Useful Keyboard Shortcuts

Key	Action
Space	Pause / Resume simulation
Backspace	Reset robot to initial state
Ctrl + L	Reload the active XML scene
K	Save Keyframe (during recording session)

---

Contributors

Thanks to everyone who has contributed to this project!

Avatar	Name	Role
	bendsc-cmd	Maintainer
	cuongtdm	Contributor
	thanhnguyencanh	Contributor

Want to contribute? Check out the Quick Start guide and open a Pull Request!

---

Developed with ❤️ by STrike Robot. Feel free to explore and share your feedback! 🚀