google-deepmind/mujoco

MuJoCo (Multi-Joint dynamics with Contact) is a general-purpose physics engine maintained by Google DeepMind. It is designed for scenarios requiring fast and accurate simulation of articulated structures and environmental contacts, such as robotics, biomechanics, graphics animation, and machine learning, offering a C API, Python bindings, and a Unity plugin. You can directly drag and drop your own models to view them in the WebAssembly version of MuJoCo Studio at mujoco.org or live.mujoco.org; Colab tutorials cover getting started, procedural modeling, multi-threaded rollouts, LQR control, the JAX-based MJX, and training locomotion policies using differentiable physics.

MuJoCo (Multi-Joint dynamics with Contact) is a general-purpose physics engine maintained by Google DeepMind. It is designed for scenarios requiring fast and accurate simulation of articulated structures and environmental contacts, such as robotics, biomechanics, graphics animation, and machine learning, offering a C API, Python bindings, and a Unity plugin. You can directly drag and drop your own models to view them in the WebAssembly version of MuJoCo Studio at mujoco.org or live.mujoco.org; Colab tutorials cover getting started, procedural modeling, multi-threaded rollouts, LQR control, the JAX-based MJX, and training locomotion policies using differentiable physics.

Why it matters Training robot policies requires fast and accurate simulation; now you can drag and drop models to try them out directly in the browser.

Stars 16916 (+574 / 24h)

Reason

  • 574 new stars in 24 hours (indicating attention only)

Usage

  • Download pre-compiled versions from Releases (Linux, Windows, macOS); Python users can start with the Colab tutorials

Risks

  • The web version runs on WebAssembly and is slower than the native C version
  • Aimed at researchers and developers, requiring some background in physics modeling to get started

Open

Trendshift(GitHub Trending 快照)

Explore the latest AI Radar updates