mujoco: MuJoCo: The Standard for Robotics and Biomechanics Simulation
google-deepmind/mujoco
About the project
It is a general-purpose physics engine that rapidly computes joint structures and contact dynamics. It supports research and development requiring precise simulation in the fields of robotics, biomechanics, graphics, and machine learning. Maintained by Google DeepMind, it optimizes performance by leveraging low-level data structures based on a C API.
The built-in XML compiler pre-allocates data to maximize the speed of runtime simulation modules. It provides interactive visualization on native desktops and web browsers and includes various utility functions for calculating physical quantities. Python bindings and a Unity game engine plugin are also provided.
Differentiable physics simulation is supported through the JAX-based MuJoCo XLA branch. This enables the automatic derivation of analytic gradients for learning walking policies. The ecosystem also includes bindings for various languages such as MATLAB, Swift, Java, Julia, and Rust, as well as model conversion tools for OpenSim and SDFormat.
It is released under the Apache-2.0 license and supports Linux, Windows, and macOS platforms. It can be run instantly as a WebAssembly version in a browser, or models can be manipulated via drag and drop using the MuJoCo Studio app locally. It is characterized by the ability to experience simulation directly on the web without a complex installation process.
google-deepmind/mujoco
Multi-Joint dynamics with Contact. A general purpose physics simulator.
C++
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.