In October 2021, we announced that we had acquired MuJoCo Physics Simulatorand made it freely available to everyone to support research everywhere. We also committed to developing and maintaining MuJoCo as a free, open, community-driven project with best-in-class capabilities. Today, we are ecstatic to announce that the open source is complete, and the entire code base is on GitHub!
In this article, we explain why MuJoCo is a great platform for open source collaboration and provide a preview of our future roadmap.
Collaboration platform
Physical simulators are key tools in state-of-the-art robotics research and can often be categorized into the following two categories:
- Commercial, closed-source software.
- Open source software, often created in an academic environment.
The first category is unclear to the user and while sometimes free to apply, it cannot be modified and is tough to understand. The second category often has a smaller user base and suffers when its creators and maintainers graduate.
MuJoCo is one of the few fully featured simulators backed by a respected company that is truly open source. As a research-driven organization, we see MuJoCo as a collaborative platform where roboticists and engineers can join us in developing one of the world’s best robot simulators.
Features that make MuJoCo particularly attractive to those looking to collaborate include:
- A full-featured simulator that can Model complex mechanisms.
- Readable, productive and portable code.
- Easily extensible code base.
- Detailed documentation: both user comments and code comments.
We hope that colleagues from academia and the OSS community will apply this platform and contribute to the code base, which will improve the quality of research for all.
Efficiency
As a C library without active memory allocation, MuJoCo is very brisk. Unfortunately, raw physics speed has historically been hampered by Python wrappers that make batch, multi-threaded operations unproductive due to the presence of a Global Interpreter Lock (GIL) and uncompiled code. We address this issue in the future in our roadmap below.
For now, we’d like to share benchmark results for two popular models. The results were obtained on a standard AMD Ryzen 9 5950X machine running Windows 10.
road map
Here’s our upcoming roadmap for MuJoCo:
- Discover the speed potential of MuJoCo with batch, multi-threaded simulation.
- Support for larger scenes thanks to improvements in internal memory management.
- Up-to-date incremental compiler with better model composition.
- Support for better rendering via Unity integration.
- Native support for physical derivatives, both analytic and finite difference.
Learn more
Useful resources for MuJoCo:
We are waiting for your suggestions!