This project provides a structured, hands-on learning path for Embodied AI, targeting individuals from Python beginners to experienced AI engineers. It aims to demystify the field by offering runnable examples, practical tutorials, and a progression from foundational robotics to cutting-edge research, fostering an open-source ecosystem for embodied intelligence.

How It Works

The project employs a "learn-by-doing" methodology, guiding users through three stages: foundational robotics and hardware, core AI techniques (CV, RL, planning), and advanced topics like Large Models (VLA/VLN) and Sim2Real deployment. It leverages popular simulation platforms like MuJoCo and NVIDIA Isaac Sim, alongside real-world hardware integrations, to bridge the gap between theory and practice. The approach prioritizes reproducibility and extensibility, allowing users to build practical skills and contribute to the field.

Quick Start & Requirements

• Primary install/run command:

  git clone https://github.com/datawhalechina/every-embodied.git
  cd every-embodied
  conda create -n embodied python=3.8
  conda activate embodied
  pip install mujoco
  # Optional: pip install ruckig
  python examples/01_hello_every_embodied_mujoco.py
  

• Non-default prerequisites: Python 3.8+, NVIDIA GPU (RTX 3060+ recommended for Isaac Sim/RL), Ubuntu 20.04/22.04 (recommended). Core libraries include MuJoCo and Isaac Sim.
• Links: Detailed setup in examples/README.md. Video tutorials available for Habitat navigation, GenieSim, and Mujoco experiments.

Highlighted Details

• Comprehensive learning map covering robotics fundamentals, CV, RL, planning, and advanced VLA/VLN models.
• Integration with state-of-the-art research, including LLM-based drone navigation and large-scale VLA benchmarks.
• Support for multiple simulation environments: MuJoCo, NVIDIA Isaac Sim, and GenieSim.
• Recent updates include GenieSim one-click deployment, Habitat navigation, and integration with LeRobot SO101 robotic arms.

Maintenance & Community

The project is led by Ethan-Chen-plus and involves contributions from multiple university students. Community engagement is facilitated through a Datawhale WeChat group (scan QR code in README) and a Feishu knowledge base. Recent news indicates active development throughout early 2025.

Licensing & Compatibility

The project documentation and tutorials are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). This license permits free sharing and adaptation, provided attribution is given. Compatibility for commercial use should be assessed against the full CC BY 4.0 terms.

Limitations & Caveats

Certain modules, such as ROS/ROS2 integration and advanced manipulation (e.g., dexterous hands), are marked as under development (status '🚧'). Specific sub-project requirements may vary, necessitating consultation of individual README files.