TeardexRobots/Parts/Training
MuJoCo
A physics simulator owned by Google DeepMind and free for anyone to use. It is known for modelling contact, such as a foot on the floor, accurately. Robot makers use it as a second opinion before trusting a controller on real hardware.
- 2 of 5
- robots in the database
- Google DeepMind
- made by
- Training
- brain layer
- Free
- listed price
Key numbers
- Owned by
- Google DeepMind
- Licence
- Apache 2.0: free to use, change and sell
Source: Google DeepMind on GitHub, MuJoCo (github.com)
Price guide
Source: Google DeepMind on GitHub, licence file (github.com)
Strengths and trade-offs
Our analysisStrengths
Trusted for accurate contact physics.
Runs on an ordinary computer with no special graphics card.
Free, open and not tied to any chip maker.
Trade-offs
Slower than graphics-card simulators for training thousands of robots at once.
Its pictures are plain, so it is weak for training vision.
Why robot makers choose it
Our readA controller that only works in the simulator it was trained in is fragile. Checking it in a second, different simulator catches that before a real robot falls over. MuJoCo is the usual choice for that check.
This is our reasoning from the public facts. No maker has confirmed it.
Robots using it
Unitree G1ConfirmedHumanoid, all versionsUsed as a second check before moving to the real robot.
Unitree Go2ConfirmedQuadruped, all versionsUsed as a second check before moving to the real robot.
This list covers only the robots researched so far. A robot missing here may still use this part.
Numbers and listed prices on this page come from the sources shown. Anything marked as our analysis, our read or our estimate is our own view and has not been confirmed by any company. Last checked 12 Oct 2026. How we verify