FuRL

Environment Setup

Install the conda env via:

conda create --name furl python==3.11
conda activate furl
pip install -r requirements.txt

Training

Generating Expert Dataset

An optional setting in FuRL is to use a goal image to accelerate the exploration before we collected the first successful trajectory.

python main.py --config.env_name=door-open-v2-goal-hidden --config.exp_name=oracle

The oracle trajectory data will be saved in data/oracle.

Example on Fixed-goal Task

python main.py --config.env_name=door-open-v2-goal-hidden --config.exp_name=furl

Example on Random-goal Task

python main.py --config.env_name=door-open-v2-goal-observable --config.exp_name=furl

Paper

FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning

Yuwei Fu, Haichao Zhang, Di Wu, Wei Xu, Benoit Boulet

International Conference on Machine Learning (ICML), 2024

Cite

Please cite our work if you find it useful:

@InProceedings{fu2024,
  title = {FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning},
  author = {Yuwei Fu and Haichao Zhang and Di Wu and Wei Xu and Benoit Boulet},
  booktitle = {Proceedings of the 41st International Conference on Machine Learning},
  year = {2024}
}