Why mujoco_playground when we already have brax?

[IrvingF7]

Hi!

First of all, thanks for this open source project. Forgive me for this blunt question. I have been playing with brax and mujoco_playground for a bit and still a noob in this regard, but I can't help but wonder what differentiates mujoco_playground from brax.

I do see that brax does not come with as many envs as mujoco_playground, but is that the only thing? Any insight is appreciated. Thank you!

[kevinzakka]

1. Playground is primarily focused on robotics. The dm_control port was added as an extra contribution, but it is not the core focus of the project.
2. Playground does not depend on Brax for environment definitions. It only uses Brax’s RL library for training, and the environments themselves are independent. In principle, Playground can be integrated with any reinforcement learning library of your choice.