Skip navigation
JUST IN: Google DeepMind just released MuJoCo 4.0.
0:00 / 0:37

JUST IN: Google DeepMind just released MuJoCo 4.0.

Lukas Ziegler

@lukas_m_ziegler66K followers

Visit on X
535 likesDownloadOriginal post

29,207 views

JUST IN: Google DeepMind just released MuJoCo 4.0. The first major version of the world’s most widely used robotics physics engine since 2023. The problem it solves is subtle but fundamental. Collision detection checks object positions only at discrete timesteps. A thin or fast object can move further than its own thickness in a single step and pass straight through another. The usual fixes are a smaller timestep or an inflated margin. Both cost you speed or accuracy. MuJoCo 4.0’s headline feature, speculative contacts, takes a different approach. Each fast-moving geom is swept forward through the step to find when it would hit something. The solver gets a contact like any other that stops the motion at the surface. No sub-stepping required. What else is new: → Deformables solved implicitly together with rigid bodies, joints and tendons, keeping cables, cloth and shells stable at 5-10x larger timesteps → Penetration-free cloth and elastic rods with bending and twist → Plasticity: wires, clay, folded paper and metal foil now keep their dents and creases → MuJoCo Studio: a new viewer SDK with physically based rendering, running in a browser tab with nothing to install → MuJoCo Playground 1.0 ships alongside The simulators are catching up with the real world! 😮‍💨 🔗 Check it out here: Congrats Carolina Parada, Yuval Tassa and team! ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

12 Comments

Comments from the original post on X (Twitter)

  • clankr's profile picture

    @clankrmedia

    Pretty big deal. Still best option for lightweight physics simulation for robots.

    2 likes

  • Don Aborah's profile picture

    @antifragiledon

    @6carty

    1 like

  • Timi's profile picture

    @iitiimii

    Thankss!!!

    1 like

  • Scott Evernden's profile picture

    @yay3d

    do I have to learn this too !?

    1 like

  • Atomik's profile picture

    @poatomik

    Tunneling is the old bug where a bullet-fast object skips through a wall because the sim only checks each frame. Sweeping forward to find the hit time fixes it without tiny timesteps. Cloth that keeps creases matters too: robots folding laundry need that.

  • Utsav Mathur's profile picture

    @mathurutsav

    Speculative contacts is the kind of unglamorous fix that quietly saves a thousand sim-to-real projects. MuJoCo 4.0 is a big one.

  • The SI Maximalist's profile picture

    @thesimaximalist

    A robot policy can learn to exploit missed collisions. A “successful” grasp is much less useful when it depends on the fingers passing through a thin object between simulation steps

  • Pablo V's profile picture

    @pvargasx

    Nice! I believe speculative contact is the right call for thin fast stuff, but I believe game engines that use it tend to get ghost collisions though, like objects braking near geometry they’d never actually touch. curious how the solver avoids that here

  • 🛸#𝕀𝕟𝕥𝕖𝕣𝕟𝕖𝕥𝕠𝕗𝔸𝕘𝕖𝕟𝕥𝕤's profile picture

    @yaelmendez

    @asimovinc 🤳

  • NotaDEV * fireply.ai's profile picture

    @leononrails

    plasticity for wires and clay is the part i want to try. cables that actually stay bent

  • AMAN VAIBHAV's profile picture

    @amanvaibhav

    A knife slipping through a timestep is the gap that only shows up once the policy meets hardware. CCD should make contact-rich sim a lot less leaky.

  • ask.charity's profile picture

    @askdotcharity

    @grok is this for cpu users or those rich gpu kids?