Resource Download
This is the official technical documentation for Global Humanoid Robot Challenge 2026. It describes a humanoid robot simulation platform built on the LeRobot framework and provides a complete workflow from physics simulation, data collection, and model training to deployment.
1. Project Overview
This documentation is intended for Global Humanoid Robot Challenge 2026 (GHRC 2026) participants and R&D teams. It provides a unified baseline implementation:
- Build a high-fidelity robot simulation environment based on NVIDIA Isaac Sim
- Record data through keyboard teleoperation and export it in the standardized LeRobotDataset V3.0 format
- Train and fine-tune models based on imitation learning algorithms, such as ACT and Pi0
- Quickly reproduce results and run comparative experiments with the official pretrained weights
2. Core Capabilities
| Capability | Description |
|---|---|
| Simulation Environment | High-fidelity Walker S2 robot simulation based on NVIDIA Isaac Sim, supporting a 20-dimensional state space: 14 arm joints, 4 gripper joints, and 2 gripper control commands |
| Data Collection | Supports keyboard teleoperation and exports data in LeRobotDataset V3.0 format |
| Model Training | Supports imitation learning algorithms such as ACT and Pi0 |
| Four-Camera Real-Time Display | Supports real-time preview from 4 RGB cameras: head_left, head_right, wrist_left, and wrist_right |
3. Resource Description
Some large files in this project are hosted on Hugging Face. Before first use, please complete the downloads first:
| Resource Type | Local Directory | Remote URL |
|---|---|---|
| Simulation environment and robot assets | assets/ (Git submodule) | UBTECH-Robotics/challenge2026_assets |
| Training dataset | datasets/ | UBTECH-Robotics/challenge2026_dataset |
3.1 Configuration Information
| Minimum Requirements | Recommended Configuration | Ideal Configuration | |
|---|---|---|---|
| Operating System | Ubuntu 22.04 / 24.04; Windows 10 / 11 | Ubuntu 22.04 / 24.04; Windows 10 / 11 | Ubuntu 22.04 / 24.04; Windows 10 / 11 |
| CPU | Intel Core i7 (7th Generation); AMD Ryzen 5 | Intel Core i7 (9th Generation); AMD Ryzen 7 | Intel Core i9, X-series or higher; AMD Ryzen 9, Threadripper or higher |
| Core Count | 4 | 8 | 16 |
| RAM | 32GB | 64GB | 64GB |
| Storage | 50GB SSD | 500GB SSD | 1TB NVMe SSD |
| GPU | GeForce RTX 4080 | GeForce RTX 5080 | RTX PRO 6000 Blackwell |
| VRAM | 16GB | 16GB | 48GB |
| Driver | Linux: 580.65.06; Windows: 580.88 | Linux: 580.65.06; Windows: 580.88 | Linux: 580.65.06; Windows: 580.88 |
A larger amount of RAM and VRAM is recommended, especially when model training is required. In addition, if driver version 595 is installed, Isaac Sim may crash later when running inside the Docker container. Therefore, driver version 580 is recommended.
3.2 Tool Requirements
| Tool | Version | Notes |
|---|---|---|
CUDA | 12.8 | Official Guide |
Docker | latest | Official Guide |
NVIDIA Container Toolkit | latest | Official Guide |
Hugging Face | latest | pip install huggingface-hub; huggingface-cli --help # Verify installation |
Git | latest | sudo apt update; sudo apt install git -y; git --version # Verify version |
Miniconda | latest | Official Guide (optional) |