Kbot-Joystick

Train and deploy your own humanoid robot controller in 1k lines of Python

Documentation · K-Sim Examples

kbotvid2_noaudio.mp4

kbotvid1_noaudio.mp4

Getting Started

You can use this repository as a starting point

1. Clone this repository:

git clone https://www.github.com/kscalelabs/kbot-joystick
cd kbot-joystick

2. Create a new Python environment (we require Python 3.11 or later)
3. Install the package with its dependencies:

pip install -r requirements.lock
pip install 'jax[cuda12]'  # If using GPU machine, install Jax CUDA libraries

4. Train a policy: Policy converges in about 1hr, but reward keeps going up for about 24hrs

python -m train

5. Convert the checkpoint to a kinfer model:

python -m convert /path/to/ckpt.bin /path/to/model.kinfer

6. Visualize the converted model:

kinfer-sim assets/model.kinfer kbot-headless

Troubleshooting

If you encounter issues, please consult the ksim documentation or reach out to us on Discord.

Tips and Tricks

To see all the available command line arguments, use the command:

python -m train --help

To visualize training your model without using kscale-mujoco-viewer, use the command:

python -m train run_mode=view