How tianxingchen/embodied-ai-guide works
1. Repository type and purpose
- This is not an application runtime repo; it is a static knowledge base / curated guide for embodied AI in Chinese.
- The architecture is centered on documentation organization, not code execution: Markdown pages, PDFs, and images form the content layer.
2. Top-level entrypoint
- README.md is the primary landing page and navigation hub.
- It introduces the project, sets learning goals, points to major topic areas, and links to external learning resources and community channels.
- For an architecture diagram, treat README.md as the front door that routes readers into the rest of the knowledge tree.
3. Core content modules
- The main taxonomy lives under topics/ and is split into four subject areas:
- topics/algorithm.md
- topics/control.md
- topics/hardware.md
- topics/infrastructure.md
- These are the repository’s main domain subsystems, each covering a different layer of the embodied AI stack.
4. Algorithm knowledge layer
- topics/algorithm.md is the largest conceptual layer.
- It organizes the “capability stack” from foundational tools and vision models up through robot learning, LLMs for robotics, VLA models, computer vision, graphics, multimodal models, navigation, and embodied AI applied to other domains.
- This page is the repository’s main methodology and model-architecture index.
5. Control and robotics foundation layer
- topics/control.md covers the stability and dynamics side of embodied systems.
- Its structure emphasizes control theory, robotics foundations, kinematics/dynamics, SLAM, odometry, and engineering tools.
- This acts as the system-level grounding layer underneath learning-based methods.
6. Hardware layer
- topics/hardware.md organizes the physical embodiment stack: embedded systems, mechanical design, robot system design, sensors, tactile sensing, data-collection hardware, and company ecosystem references.
- This module represents the real-world interface boundary where algorithms meet sensors, actuators, and platform constraints.
7. Infrastructure layer
- topics/infrastructure.md focuses on the research and benchmarking substrate: simulators, benchmarks, and datasets.
- This is the repo’s evaluation and data foundation, connecting methodology to reproducible training and comparison.
- In an architecture diagram, this layer sits between the abstract algorithm layer and the practical data/evaluation workflow.
8. Learning workflow / guided path
- The README defines a recommended onboarding path, especially the “hands-on” RoboTwin 2.0 workflow.
- That workflow describes a sequence of simulation-based data collection → policy training → policy evaluation for a manipulation task.
- This is the only clearly defined end-to-end process in the repo, and it serves as the guide’s practical learning pipeline.
9. Static asset layer
- The files/ directory stores non-text resources:
- PDFs for foundational reading
- images for logos, community promotion, and sponsor visuals
- These assets support the documentation experience but do not introduce runtime behavior.
- They should be shown as a content asset store connected to the README and possibly referenced from topic pages.
10. External ecosystem dependencies
- The repo heavily links outward to external resources rather than embedding them:
- RoboTwin 2.0, SAPIEN, and ACT tutorials for hands-on practice
- community sites, paper lists, benchmarks, and related open-source projects
- social/community channels and lab references
- These are reference dependencies, not package dependencies; they extend the knowledge base into the wider embodied-AI ecosystem.
11. Audience and interaction model
- The primary users are learners, researchers, and community contributors.
- Interaction is read-oriented: users browse topic pages, follow external links, and may contribute via PRs or community contact.
- There is no backend, API, database, or build pipeline visible in this repository.
12. Governance and boundary conditions
- The repo is explicitly non-commercial and community-oriented.
- Licensing and citation instructions are part of the repository’s governance layer, which matters for reuse and redistribution.
- For a diagram, note that the content layer is bounded by a non-commercial license and community contribution workflow.
13. Suggested diagram framing
- Model the repository as a documentation hub with one entrypoint (README.md), four subject modules under topics/, an asset store under files/, and outbound links to external tools, papers, and community resources.
- The main flow is: reader enters via README → chooses a topic module or hands-on workflow → follows external references or downloads assets.
Main entry points: README.md README · files Assets · infrastructure.md Data & eval · algorithm.md Algorithm
Components 16
Entry & governance
Top-level navigation, contribution context, and licensing boundary.
- README · Landing page, navigation hub, and learning path guide.
README.md - License · Non-commercial reuse boundary and citation guidance.
LICENSE - RoboTwin path · Simulation-based data collection, policy training, and policy evaluation loop.
README.md
Topic modules
Core knowledge areas under the embodied AI taxonomy.
- Algorithm · Methodology, models, and capability stack for embodied AI.
topics/algorithm.md - Control · Dynamics, control theory, SLAM, and robotics foundations.
topics/control.md - Hardware · Physical embodiment stack, sensors, platforms, and data-collection hardware.
topics/hardware.md - Infrastructure · Simulators, benchmarks, and datasets for research and evaluation.
topics/infrastructure.md - Capability stack · Foundational tools, vision, learning, VLA, navigation, and cross-domain embodied AI.
topics/algorithm.md - Robotics basics · Kinematics, dynamics, odometry, SLAM, and control concepts.
topics/control.md - Embodiment stack · Embedded systems, mechanical design, sensors, tactile sensing, and robot integration.
topics/hardware.md - Data & eval · Simulator, benchmark, and dataset workflow that supports reproducible comparison.
topics/infrastructure.md
Content assets
Static PDFs and images that support the guide.
- Assets · Static PDFs, images, and supporting visual materials.
files - PDF readings · Foundational Chinese-language reference PDFs for background study.
files/机器人学简介.pdf - Route map · Embodied AI foundation roadmap PDF.
files/具身智能基础技术路线-YunlongDong.pdf
External ecosystem
Outward references, tools, benchmarks, and community resources.
- Community refs · Labs, paper lists, benchmarks, social channels, and related projects.
- Practice tools · RoboTwin 2.0, SAPIEN, and ACT tutorials used for hands-on learning.
Connections 19
- README → Algorithm: routes to · Primary navigation into the taxonomy.
- README → Control: routes to · Navigation into control and robotics foundations.
- README → Hardware: routes to · Navigation into the physical embodiment layer.
- README → Infrastructure: routes to · Navigation into simulation, benchmarks, and datasets.
- README → RoboTwin path: starts · Recommended onboarding workflow for applied learning.
- README → Assets: references · Landing page points readers to downloadable assets.
- README → License: bounds · License and reuse constraints frame the repository.
- Algorithm → Capability stack: organizes · Algorithm page expands into the major capability stack.
- Control → Robotics basics: organizes · Control page expands into foundational robotics concepts.
- Hardware → Embodiment stack: organizes · Hardware page expands into embodiment and platform layers.
- Infrastructure → Data & eval: organizes · Infrastructure page expands into evaluation and data workflow.
- RoboTwin path → Practice tools: uses · Workflow depends on external simulation and training tools.
- RoboTwin path → Data & eval: depends on · Hands-on path uses simulation, data, and evaluation substrate.
- Capability stack → Practice tools: cross-links · Method pages connect to practical tooling and tutorials.
- Embodiment stack → Community refs: cross-links · Hardware references extend to companies, labs, and ecosystem resources.
- Data & eval → Community refs: cross-links · Benchmarks and datasets reference external projects and lists.
- Assets → PDF readings: contains · Asset directory includes background reading PDFs.
- Assets → Route map: contains · Asset directory includes roadmap and overview PDFs.
- README → Community refs: points outward · README directs readers to community and reference resources.