Skip to content

MuJoCo MPC

  • Software
  • Humanoids and robot learning
  • Tool

An interactive application and framework for real-time predictive control of robots with MuJoCo, using shooting-based planners such as iLQG, Gradient Descent and Predictive Sampling.

  • Stale

Made by Google DeepMind.

Source on GitHub: github.com/google-deepmind/mujoco_mpc (opens in a new tab). Facts read at commit ff572a2 (opens in a new tab).

Licenses

Code
Apache-2.0 (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit ff572a2 (opens in a new tab), on 16 Oct 2026.

Commercial use
Yes

Labels

  • StaleNo commit in 12 months: the newest is from 27 May 2025.

What it does

  • Lets users author and solve complex robotics tasks interactively with MuJoCo models.
  • Supports multiple shooting-based planners, including derivative-based methods (iLQG, Gradient Descent) and the derivative-free Predictive Sampling.
  • Provides a graphical user interface for running and inspecting tasks.
  • Offers an experimental Python API for building and running MJPC tasks alongside MuJoCo models.
  • Includes example tasks such as a quadruped, bimanual manipulation, Rubik's cube unscrambling and humanoid motion-capture tracking.
  • Is described by its authors as a research prototype rather than production-quality software.

Written by Robotheca from the project's README (opens in a new tab) at commit ff572a2.

Getting started

Difficulty
Advanced
Hardware it needs, as the project states it
Tested on Ubuntu 20.04 and macOS-12; other versions and Windows are said to likely work but are untested. Requires building with CMake and clang-12 on Ubuntu, and Xcode on macOS.

Maintenance

Status
Stale: no commit in 12 months
Newest commit
27 May 2025 on main
Latest release
V0.1.0 (23 Feb 2024)

Checked on GitHub on 16 Oct 2026.

Robots that can run it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.