[{"content":"Hi, I\u0026rsquo;m Suzhongyu (苏中玉). Online I usually go by 某不科学的虚.\nI work at the intersection of mechanics and robotics: how tendon-driven transmissions shape what a robot hand can do, how to model that faithfully, and how to make the simulated hand behave like the real one. My background is engineering mechanics, and I care about the gap between a clean equation and a robot that actually moves.\nFor education, research experience, and skills, see my CV.\nResearch interests Tendon-driven robotic hands — routing, friction, wrapping, transmission efficiency Soft and continuum robotics — constitutive models, static modeling, fabrication-aware design Simulation and control — MuJoCo models, tendon actuators, joint-space control loops Embodied intelligence — vision-language-action models, and what manipulation policies need from hardware Mechanics-informed learning — using structure instead of brute force What I\u0026rsquo;m working on Building and validating a tendon-driven biomimetic hand in MuJoCo, including tendon routing and an angle outer-loop controller. Re-implementing papers on dexterous manipulation with tendon-driven hands. Exploring VLA models for manipulation and thinking about the sim-to-real gap. Elsewhere GitHub: @jokerxuxu Google Scholar: YOUR_SCHOLAR_ID Email: suzhongyu@zju.edu.cn ","permalink":"https://jokerxuxu.github.io/about/","summary":"Suzhongyu — engineering mechanics, tendon-driven soft robotics, and simulation.","title":"About"},{"content":"Why another personal site I spend most of my time reading papers, breaking simulations, and fixing the same class of bug twice. Most of that never leaves my notes app. This site is an attempt to make the useful part public: short technical notes that I would have wanted to find six months ago.\nIt is also a place to keep a project record that is more honest than a CV line — what I tried, what failed, and what I would do differently.\nWhat you will find here Tendon-driven hands and soft robotics — routing, transmission, friction, and control. Simulation notes — MuJoCo modeling details that are easy to get wrong. Paper summaries — short, with the assumptions I think matter. Occasional mechanics — derivations I do not want to redo. I will not write about unpublished results from my group, and I will not turn this into a diary. If a month has no post, that is fine.\nA formatting check This post also doubles as a template check, so here is a small piece of the mechanics I keep coming back to. For a cable crossing a joint, the tendon length and joint torque are\n$$ L(q) = L_0 + \\sum_i r_i(q_i)\\, q_i, \\qquad \\tau_i = r_i(q_i)\\, T, $$ where $T$ is the tendon tension and $r_i$ is the moment arm. The annoying part is that $r_i$ is not a constant: it changes with the wrapping geometry, and so does the friction loss across the guides.\nAnd a code block, to confirm syntax highlighting and the copy button:\ndef tendon_length(q, moment_arms, rest_length=0.0): \u0026#34;\u0026#34;\u0026#34;Tendon length for a serial chain, ignoring wrapping dynamics.\u0026#34;\u0026#34;\u0026#34; return rest_length + sum(r * qi for r, qi in zip(moment_arms, q)) If the equation above is typeset rather than shown as raw $$ ... $$, KaTeX is working.\nHow the site is built Hugo with the PaperMod theme, deployed to GitHub Pages by a GitHub Actions workflow. Content is plain Markdown, math is KaTeX, and the source is on GitHub.\nWhat\u0026rsquo;s next A proper write-up of the tendon routing model — including the parts that did not work.\n","permalink":"https://jokerxuxu.github.io/posts/hello-world/","summary":"Why this site exists, what I plan to write about, and how it is built.","title":"Hello, world"},{"content":"Why Tendon-driven hands move the actuators out of the hand and into the forearm, then route force through cables. That buys compactness and low distal inertia, and it costs you: tendon length depends on every joint the cable crosses, friction and wrapping are hard to model, and the joints are coupled even when the mechanism looks simple.\nI wanted a simulation model where I could change the routing, break the transmission, and see what happens — before committing to hardware.\nWhat I built A parametric hand model in MuJoCo, with tendon actuators whose length is computed from joint coordinates rather than hard-coded. A joint-space outer loop that tracks reference angles, plus the low-level tendon force allocation underneath it. Scripts to sweep routing parameters (moment arms, pretension, friction) and log tracking error. What I\u0026rsquo;m figuring out How much tracking error comes from friction and wrapping versus model error in the moment arms. Whether a simple Coulomb-plus-viscous friction model is enough for pick-and-place, or whether hysteresis has to be modeled explicitly. How to validate the simulated transmission against a real tendon-driven finger, not just against itself. Status In progress. Code and results will be linked here once they are presentable.\n","permalink":"https://jokerxuxu.github.io/projects/tendon-driven-hand/","summary":"Simulation-first study of a tendon-driven anthropomorphic hand: routing, transmission, and joint-space control.","title":"A Tendon-Driven Biomimetic Hand in MuJoCo"},{"content":"Why this note exists A tendon is not a rigid link. It wraps around pulleys and guides, its tension drops across every contact, and the moment arm it produces is a function of joint angle — not a constant. Most quick models quietly assume otherwise, and then the controller spends its life compensating for the assumption.\nThese are my working notes on the static side of that problem.\nThe pieces I keep coming back to Moment arm as a function of configuration. For a cable crossing joint $i$, the torque is $r_i(q)\\,T$, where $r_i$ changes as the cable wraps. Capstan friction. Tension across a wrap follows $T_{\\text{out}} = T_{\\text{in}}\\,e^{-\\mu\\theta}$; a hand with several guides can lose a large fraction of the commanded force before it reaches the fingertip. Routing as a design variable. Where the cable enters and how many guides it crosses changes both the moment arm profile and the friction loss — the two cannot be tuned independently. Open questions When is a constant-moment-arm approximation acceptable for grasp force control? How much hysteresis is needed before a rate-independent friction model stops being useful? Status Notes in progress; no polished write-up yet.\n","permalink":"https://jokerxuxu.github.io/projects/tendon-routing-statics/","summary":"Working notes on how a cable\u0026rsquo;s path through a hand changes the force a joint actually sees.","title":"Tendon Routing: Statics, Friction, and Moment Arms"},{"content":"Why I\u0026rsquo;m reading this VLA models are usually evaluated on arms with parallel grippers. A tendon-driven hand has compliance, coupling, and friction that a position-controlled gripper does not — and I want to know which of those the policy is implicitly relying on.\nWhat I\u0026rsquo;m collecting Action spaces. Joint positions, end-effector poses, or latent actions — each one hides a different assumption about the controller. Data. How much of the success is the policy, and how much is a well-tuned impedance controller underneath it? Hardware-aware policies. Work that conditions on hand morphology or tendon state instead of treating the hand as a black box. Status Ongoing reading notes. I will publish short summaries as I finish each paper.\n","permalink":"https://jokerxuxu.github.io/projects/vla-reading-notes/","summary":"Notes from reading VLA work, with an eye on what it assumes about the robot body underneath.","title":"Vision-Language-Action Models: Reading Notes"},{"content":" A PDF version is coming soon. Education B.Eng. in Engineering Mechanics — [Zhejiang University], 2023 – 2027 (expected) Research experience Undergraduate researcher — [Lab / group], 2026 – present Modeling and simulating a tendon-driven biomimetic hand in MuJoCo. Implementing joint-space control with tendon-level force allocation. Building tools for parameter sweeps and tracking-error analysis. Projects See the projects page for details.\nSkills Programming: Python, JAX, Bash Simulation: MuJoCo Mechanics: statics and dynamics, continuum mechanics, tendon transmission modeling Tools: Git, Linux, LaTeX Contact GitHub: @jokerxuxu Email: suzhongyu@zju.edu.cn ","permalink":"https://jokerxuxu.github.io/cv/","summary":"\u003cblockquote\u003e\n\u003cp\u003eA PDF version is coming soon. \u003c!-- TODO: put cv.pdf into static/, then replace this line with: \u003e [Download PDF](/cv.pdf) --\u003e\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch2 id=\"education\"\u003eEducation\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eB.Eng. in Engineering Mechanics\u003c/strong\u003e — [Zhejiang University], 2023 – 2027 (expected) \u003c!-- TODO --\u003e\u003c/p\u003e\n\u003ch2 id=\"research-experience\"\u003eResearch experience\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eUndergraduate researcher\u003c/strong\u003e — [Lab / group], 2026 – present \u003c!-- TODO --\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eModeling and simulating a tendon-driven biomimetic hand in MuJoCo.\u003c/li\u003e\n\u003cli\u003eImplementing joint-space control with tendon-level force allocation.\u003c/li\u003e\n\u003cli\u003eBuilding tools for parameter sweeps and tracking-error analysis.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"projects\"\u003eProjects\u003c/h2\u003e\n\u003cp\u003eSee the \u003ca href=\"/projects/\"\u003eprojects page\u003c/a\u003e for details.\u003c/p\u003e","title":"CV"},{"content":"I am at an early stage of my research career, so this list is short for now. Papers and preprints will appear here as they become public.\n","permalink":"https://jokerxuxu.github.io/publications/","summary":"\u003cp\u003eI am at an early stage of my research career, so this list is short for now. Papers and preprints will appear here as they become public.\u003c/p\u003e\n\u003c!--\nWhen you have an entry, copy this pattern and delete the comment markers.\n\n### Paper title\n\n**Suzhongyu**, Coauthor A, Coauthor B. *Venue*, Year.\n\n[PDF](#) · [Code](#) · [BibTeX](#)\n\n---\n--\u003e\n\u003c!-- TODO: when you have a manuscript in preparation, add it here:\n## Manuscripts in preparation\n\n- Title — *in preparation*.\n--\u003e","title":"Publications"}]