Skip to content

lupusorina/robot_learning

Repository files navigation

Performance

Run the robot with the keyboard

cd robot_learning/src/jax/envs
mjpython biped_test.py

Note: on MacOS you will likely need to whitelist both the terminal and the mjpython executable for input monitoring. In System Settings, navigate to Privacy & Security -> Input Monitoring, then add your terminal and mjpython (cmd+shift+G in the file browser will let you add a path).

Install requirements

Tested on Python 3.12.

pip3 install -r requirements.txt

Install the repo

pip3 install -e .

Run biped learning code in Jax

cd src/jax
python3 train.py

Inference: test.ipynb

File structure

robot_learning
    └── src
            └── jax
                └── biped.py                   (Biped in Jax)
                └── train.py                   (Train PPO on Biped)
                └── test.ipynb                 (Jupyter notebook for testing)
                └── mjx_env.py                 (file taken from mujoco-playground and modified)
                └── wrapper.py                 (file taken from mujoco-playground and modified)
                └── randomize.py               (domain randomization)
                └── utils.py                   (utils)
            └── assets
                └── biped                      (biped)

    └── tests

This project uses/derives from MuJoCo Playground (Apache 2.0) by Google DeepMind.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 5