TianxingChen/Embodied-AI-Guide

What is it?

Key features

• 01Encyclopedia-style sections covering algorithms, infrastructure, control and hardware
• 02End-to-end RoboTwin 2.0 tutorial from environment setup to policy evaluation
• 03Pre-collected dataset download for 50 dual-arm simulation tasks
• 04Curated lists of papers, conferences, newsletters and research labs
• 05Glossary of about 20 embodied AI terms and a learning path by background

Use cases

• A developer new to robot learning who wants a structured map of the field
• Running a first simulated manipulation policy training and evaluation without a physical robot
• Finding recent VLA papers and related open-source repositories
• Choosing a hardware or control learning path based on a software or robotics background

Quick start

1. 1

   Clone the RoboTwin repo with its XPolicyLab submodule

   git clone --recurse-submodules https://github.com/RoboTwin-Platform/RoboTwin.git
2. 2

   Download the pre-collected data for the beat_block_hammer task

   bash scripts/download_xpolicylab_data.sh beat_block_hammer
3. 3

   Install the ACT policy runtime environment from the XPolicyLab policy folder

   bash install.sh
4. 4

   Follow the official install guide to set up the RoboTwin platform environment

Things to know

• !Training the ACT tutorial needs a GPU with at least 16GB of VRAM.
• !Official guidance advises avoiding A, H, and V series GPUs for data collection and policy evaluation.
• !The project is licensed for non-commercial use only.

Topics

FAQ

What is Embodied-AI-Guide?

A Chinese-language embodied AI knowledge base and curated index covering algorithms, infrastructure, control, and hardware, plus a hands-on RoboTwin 2.0 tutorial. Embodied-AI-Guide is a Chinese-language knowledge base for embodied AI, the field of robots and other physical agents that learn to perceive and act in the real world. It is organized as an encyclopedia-style index: it explains the main technical areas (VLA models, reinforcement and imitation learning, simulators, datasets, benchmarks, whole-body control, sensors and tactile hardware) and points to papers, courses, labs, newsletters and awesome-lists for each.

Is Embodied-AI-Guide free to use? What license does it have?

Embodied-AI-Guide is open source under the Other license. Check the license terms before using it in commercial products.

Is Embodied-AI-Guide still maintained?

Actively maintained. The last commit to Embodied-AI-Guide was on Oct 14, 2026.

What can I use Embodied-AI-Guide for?

A developer new to robot learning who wants a structured map of the field Running a first simulated manipulation policy training and evaluation without a physical robot Finding recent VLA papers and related open-source repositories Choosing a hardware or control learning path based on a software or robotics background

How do I install Embodied-AI-Guide?

Clone the RoboTwin repo with its XPolicyLab submodule: git clone --recurse-submodules https://github.com/RoboTwin-Platform/RoboTwin.git Download the pre-collected data for the beat_block_hammer task: bash scripts/download_xpolicylab_data.sh beat_block_hammer Install the ACT policy runtime environment from the XPolicyLab policy folder: bash install.sh Follow the official install guide to set up the RoboTwin platform environment

Where can I get Embodied-AI-Guide?

The source code and installation instructions are on GitHub: https://github.com/TianxingChen/Embodied-AI-Guide