Frictional Q-Learning (FQL)

Frictional Q-Learning (FQL) is a batch deep reinforcement learning algorithm that learn off-policy with expanded constraints from static friction.

Commands

cd cleanrl
python run.py

cd fql
python run.py --cuda_id 0
python plot.py

Most of the evaluation results presented in the paper are provided in runs.zip

BibTeX

@misc{kim2025frictionalqlearning,
      title={Frictional Q-Learning}, 
      author={Hyunwoo Kim and Hyo Kyung Lee},
      year={2025},
      eprint={2509.19771},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2509.19771}, 
}