# Robotics_Tutorial: a 736,020-line Chinese robotics curriculum reviewed > Michael-Jetson/Robotics_Tutorial is a Markdown-only teaching corpus covering math, C++, SLAM, planning and control, and embodied AI. It is free for study under CC BY-NC 4.0 and useless to anyone who wants runnable code. **Michael-Jetson/Robotics_Tutorial** — A systematic robotics teaching document covering math fundamentals, advanced C++, planning and control, SLAM and more. - Repository: https://github.com/Michael-Jetson/Robotics_Tutorial - Website: https://robotics-tutorial-wiki.vercel.app - Stars: 743 · Forks: 81 - Language: Python - License: NOASSERTION - Published: 2026-09-20 · Updated: 2026-09-20 · Language: en - Canonical page: https://hysenlabs.com/projects/michael-jetson-robotics-tutorial ## What Robotics_Tutorial actually is, and who it is written for This is a documentation repository, not a library. There is no package to install, no importable module, no entry point. The top level holds numbered directories (01_数学, 02_C++基础与进阶, 03_SLAM, 04_移动机器人规控, 05_运动控制, 06_具身智能) plus 00_项目导航, LICENSE, README.md and SUMMARY.md. The primary language field says Python, but the content is Markdown teaching text; the Python label reflects tooling or incidental scripts rather than a runtime. The intended reader is someone who already knows they want to work on robot software and wants a syllabus rather than a quickstart. The README states the project covers the full knowledge chain from mathematical foundations to embodied intelligence. The C++ track is explicitly an engineering track: language core, concurrency, software engineering, general-purpose libraries, ROS2, and what the README calls the foundations of planning and control engineering. That is a different audience from a beginner who wants to blink an LED. The author is Pengfei Guo, affiliated with DAMIAO Technology. The README carries an unusually direct disclosure that the content was written with heavy AI assistance (Claude and Codex) combined with internet research, and that errors cannot be fully ruled out despite multiple rounds of independent review and fact-checking. Take that disclosure at face value: it is a statement about provenance, and it is the reason the review pipeline described later in the README exists at all. ## How the corpus is organised: numbered tracks and a line-count contract Navigation is by directory number, then by chapter number inside each directory. The README gives a per-direction table with file counts, total line counts, and a completion status. The totals it reports are 428 files and 736,020 lines across six directions. 01 数学基础 is 90 files and 159,718 lines; 02 C++基础与进阶 is 58 files and 115,733 lines; 03 SLAM is 14 files and 14,642 lines; 04 移动机器人规控 is 84 files and 145,154 lines; 05 运动控制 is 152 files and 231,320 lines; 06 具身智能 is 30 files and 69,453 lines. 05 运动控制 is broken down further into 公共基础, 足式, 机械臂, 复合, and 仿真. The legged-robot track is described as a complete 24-chapter teaching sequence, mapped across named chapters Ch47 through Ch70 and grouped into 基础篇, 模型与接触, 控制篇, 规划篇, 工程篇, RL 与感知, and 综合篇. The chapter topics are concrete: Pinocchio, CppAD, spatial algebra and QP/NLP in the foundations; WBC/TSID, DDP/Crocoddyl and OCS2 MPC in control; gait, state estimation and foothold planning in planning; real-time C++ and hardware stacks in engineering; then RL training and deployment, RL+MPC hybrids, perception, code close-reading, a capstone project and research directions. That ordering is the design. A reader is expected to move from the math track into the C++ track before touching the control material, because the control chapters assume the tooling named in the C++ track. The README does not provide a dependency graph beyond the numbering, so the sequence is the only guidance you get. ## The v5.0 writing standard and the two-round review pipeline The README claims all 193 teaching articles follow an internal v5.0 writing standard and that each went through at least two independent review rounds, with review comments archived in 00_项目导航/审核意见/. The stated rules are specific enough to check: a minimum of 2,000 lines per chapter, dropping to 1,500 for engineering-practice articles; 15 rules across three layers (content, expression, cognition); and a five-stage quality gate running G1 structure, G2 accuracy, G3 consistency, G4 completeness, G5 polish. The cognitive requirements are the most distinctive part. Each chapter is supposed to contain at least three analogies, at least three counterfactual arguments, and at least two essential insights. The standard also forbids skipping steps with phrases like 显然 or 容易看出, and requires troubleshooting manuals, cross-chapter bridges, and cumulative projects. Fact verification is described as internet-based, covering paper citations, API names, and numeric claims. Here is the honest reading of that. A line-count minimum is a proxy for depth, and proxies get gamed; a 2,000-line chapter can be padded. The counterfactual and analogy quotas are the same kind of proxy, and they push writers toward a house style that some readers will find repetitive. What the pipeline does buy you is traceability: the review notes are supposed to be in the repository, so a sceptical reader can open the archive for a given chapter and judge whether the two rounds happened and whether the fixes were real. That is a stronger claim than most tutorial repositories make, and it is also the claim most worth spot-checking. ## Cloning it and reading your first chapter There is nothing to build. The README's usage section is a single clone command, followed by the instruction that all documents are standalone Markdown files and should be read in chapter-number order within each direction. ```bash git clone https://github.com/Michael-Jetson/Robotics_Tutorial.git ``` After the clone completes you get the six numbered directories plus 00_项目导航 at the repository root. Open SUMMARY.md first; it is the top-level index and the fastest way to see how the tracks connect before you commit to one. If your goal is legged-robot control, the README points you at the 24-chapter sequence, which starts at Ch47 in 05_运动控制/10_足式/. The foundations chapters there cover Pinocchio, CppAD, spatial algebra and QP/NLP, so expect to be reading about those libraries before you reach anything resembling a walking controller. Read the chapters in numeric order; the README states the sequence is a complete teaching progression, and the cross-chapter bridges assume you did not skip. ```bash cd Robotics_Tutorial ls SUMMARY.md 00_项目导航 05_运动控制 ``` That listing is worth running early. It tells you whether the directory layout in your checkout matches the tree printed in the README, which is the quickest sanity check on whether you cloned the branch you think you did. ## Where Robotics_Tutorial fails you The licence is the first hard boundary. The README states CC BY-NC 4.0: share and adapt freely for study and research with attribution, no commercial use, and no commercial licence granted. That rules out internal training material at a company, paid courses, and anything embedded in a product without contacting the author. The repository's LICENSE file is present, but the metadata field for the repository reports NOASSERTION rather than a recognised identifier, so if licence terms matter to your organisation, read LICENSE directly instead of trusting the badge. The second boundary is language. The README, the directory names, the chapter tables and the review standard are all in Chinese. There is no stated English translation. If your team does not read Chinese, the 736,020 lines are inaccessible regardless of quality. The third is that this is prose, not code. The README describes cumulative projects and code close-reading chapters, but it does not describe a runnable reference implementation, a test suite, or a simulator you can launch. Someone who wants to compile and run a controller today will get nothing from a clone. The README also does not document versioning, a changelog, or rollback of any kind, and there are no retrieved releases, so there is no release channel to pin to. Fourth, the AI-assistance disclosure cuts both ways. The author is transparent that Claude and Codex contributed heavily and that errors remain possible. Multiple review rounds reduce that risk; they do not eliminate it. Treat numeric claims and paper citations in the text as leads to verify, not as settled facts. ## How it compares with Probabilistic Robotics and the Modern Robotics text The obvious alternatives are the standard textbooks. Probabilistic Robotics (Thrun, Burgard, Fox) is narrower and deeper on estimation and SLAM, and it is a published book with a stable, citable edition. Modern Robotics (Lynch and Park) covers kinematics, dynamics and control with an accompanying code library and a Coursera course, and it is written in English. Both are commercial products with editorial and legal accountability behind them; Robotics_Tutorial is a personal, AI-assisted corpus under a non-commercial licence. The difference in approach is scope versus depth per topic. A textbook commits to one subject and proves it. Robotics_Tutorial commits to the seam between subjects: it is the only one of the three that walks from differential geometry and Lie groups through C++ concurrency and ROS2 into WBC, OCS2 MPC, RL deployment and VLA models in a single numbered sequence. That breadth is the reason to open it, and it is also why the per-topic depth will not match a dedicated text. For legged robots specifically, the 24-chapter sequence is a curated reading path through Pinocchio, Crocoddyl and OCS2 that a beginner would otherwise have to assemble from scattered documentation. If you want a citable, English, commercially supported reference, take the textbook. If you want a free Chinese-language map of the whole field with an auditable review trail, this repository is the closer fit. ## Maintenance, upgrade cost and licence implications The repository is not archived, and the last push was on 2026-09-16, four days before this review. That is recent activity, and the README's completion tables describe every direction as finished, with the audit loop closed. Finished is not the same as frozen: the README invites issue reports for factual errors, citation drift and technical misstatements, so the expected maintenance pattern is corrections rather than new chapters. Upgrade cost is close to zero in the software sense. There is no dependency to bump, no API to track, no migration to plan. You pull, and you get whatever the current Markdown says. The real cost is re-reading: if a chapter you relied on is corrected, your notes are stale, and there is no changelog in the repository to tell you what changed between two pulls. Diffing the repository yourself is the only way to see it. Licence implications, stated plainly and without legal advice: CC BY-NC 4.0 permits copying, redistribution and adaptation for non-commercial purposes provided you attribute the source, link the licence, and indicate whether you modified the material. It does not permit commercial use, and the README says commercial licensing requires contacting the author. The README also asks that infringement or factual errors be reported through GitHub issues. If you plan to reuse figures, text or citations in anything commercial, that is a conversation with the author, not a reading of the licence badge. ## Conclusion Adopt it if you read Chinese and want a structured, audited syllabus to work through alongside your own code; skip it if you need a runnable starter kit, an English text, or anything you can ship commercially. Before committing, open 05_运动控制/10_足式/ and read one chapter end to end, then check 00_项目导航/审核意见/ to see whether the review notes for that chapter are actually there, because the README claims a closed audit loop for every direction and that claim is the one worth verifying first. ## FAQ ### How can I learn robotics on my own with Robotics_Tutorial? Clone the repository and read the numbered directories in order, starting with 01_数学 and 02_C++基础与进阶 before the control tracks, since the README presents the sequence as a complete progression and the cross-chapter bridges assume you did not skip. All documents are standalone Markdown files, so you can work through them without any build step. ### Does Robotics_Tutorial use C or C++? The repository is documentation, so it does not implement anything in either language. Its 02_C++基础与进阶 track is a C++ engineering curriculum covering the language core, concurrency, software engineering, general-purpose libraries and ROS2, and the control chapters assume that tooling. ### Is Robotics_Tutorial difficult to learn from? It assumes prior motivation rather than prior knowledge, but the sequence is long: the README reports 736,020 lines across 428 files, with a minimum of 2,000 lines per chapter. The legged-robot track alone is 24 chapters that begin with Pinocchio, CppAD, spatial algebra and QP/NLP before reaching controllers. ## Sources - [Issues](https://github.com/Michael-Jetson/Robotics_Tutorial/issues) - [Michael-Jetson/Robotics_Tutorial on GitHub](https://github.com/Michael-Jetson/Robotics_Tutorial) - [Project website](https://robotics-tutorial-wiki.vercel.app) - [README](https://github.com/Michael-Jetson/Robotics_Tutorial/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/michael-jetson-robotics-tutorial