Mujoco Playground
A simple and efficient library for training humanoid locomotion in MJX and MuJoCo.
Getting Started
Clone our Mujoco Playground repository:
git clone https://github.com/kscalelabs/mujoco_playground.git
cd mujoco_playgroundInstall the package:
pip install -e .Running ZBot experiments
You can train a standing policy within 20 min on RTX 4090
python playground/runner.py --env ZbotJoystickFlatTerrainYou can inpect the training performance through the reward plot or videos of evaluated policy.
If you have access to Google Colab you can try running the notebook:
ipython playground/.ipynbEvaluation (WIP)
You can inspect the training results through the output.mp4 file or running sim2sim evaluation through
python sim2sim.py --model_path Sim2Real pipeline WIP
Updated about 2 months ago