Reinforcement-Learning-based Control for a Two-Wheeled-Legged Balancing Robot using a Simple Linear Model
This repo contains the code for our paper of the same name. You can install the requirements using the requirements.txt file or build a Singularity container based on the sk8o-rl.def file.
After installing all the dependencies, you can start the training e.g. by
python training/train.py optimizer.lr=1e-4 optimizer.total_timesteps=2e6
The code supports logging via Weights & Biases (use wandb=true to enable) and configuration via hydra (see training/config).
The models used for training and verification are available via PyPI.