Embodied-AI-Guide

Embodied-AI-Guide is a Chinese-language knowledge base and resource index for the field of embodied AI, which refers to AI systems that perceive their surroundings and act through a physical body, most commonly a robot. The project is maintained by the Lumina Embodied AI community and is structured like an encyclopedia, covering the major techniques, tools, papers, and learning paths in the field.

The guide is organized around several main areas. One section provides a hands-on tutorial designed to be completed in roughly a week, walking through a full workflow using the RoboTwin 2.0 simulation platform: collecting robot movement data, training a policy (the decision-making model), and evaluating how well the robot performs a task. This section targets beginners and requires a GPU with at least 16GB of video memory to run.

Another large section covers the algorithm landscape, organized from foundational building blocks up through higher-level decision systems. Topics include robot learning methods, visual foundation models, language models applied to robotics, and vision-language-action models that combine perception with physical action. Computer vision and graphics techniques relevant to robot sensing are also covered.

The guide also indexes hardware platforms, simulation environments, real-robot datasets, and benchmark collections that researchers use to develop and test embodied AI systems. Community resources are included: Chinese public accounts, research blogs, paper reading lists, and links to top academic conferences and journals in the area.

The project is aimed at Chinese-speaking newcomers who want to quickly build a mental map of embodied AI before diving into specific algorithms or engineering. The README is primarily in Chinese, though section titles and linked resources often include English.