DeepMind, an Alphabet subsidiary and AI research lab, acquired the MuJoCo physics engine for robotics research and development in October 2021. The simulator was to be open-sourced and maintained as a free, open source, community-driven project. DeepMind claims that the open sourcing is now complete, with the entire codebase available on GitHub.
MuJoCo, which stands for Multi-Joint Dynamics with Contact, is a physics engine designed to aid research and development in robotics, biomechanics, graphics and animation, and other fields that require fast and accurate simulation. MuJoCo can be used to implement model-based computations for machine learning applications such as control synthesis, state estimation, system identification, mechanism design, data analysis through inverse dynamics, and parallel sampling. It can also be used as a standard simulator, such as for gaming and interactive virtual environments.
According to DeepMind, the following are some of the features that make MuJoCo appealing for collaboration:
- Comprehensive simulator capable of simulating complex mechanisms
- Readable, performant, portable code
- Codebase that is easily extensible
- Extensive documentation, including both user-facing and code comments – We hope that colleagues from academia and the OSS community will use this platform and contribute to the codebase, thereby improving research for all.
DeepMind has more to say:
“As a C library with no dynamic memory allocation, MuJoCo is very fast. Unfortunately, raw physics speed has historically been hindered by Python wrappers, which made batched, multi-threaded operations non-performant due to the presence of the Global Interpreter Lock (GIL) and non-compiled code. In our roadmap below, we address this issue going forward.
“For now, we’d like to share some benchmarking results for two common models. The results were obtained on a standard AMD Ryzen 9 5950X machine, running Windows 10.”