roythuly / brhpo

BrHPO

Official implementation of

Bidirectional-Reachable Hierarchical Reinforcement Learning with Mutually Responsive Policies by

Yu Luo, Fuchun Sun, Tianying Ji and Xianyuan Zhan

Getting started

We provide examples on how to train and evaluate BrHPO agent.

Training & Visualization

To visualize the performance of BrHPO,

python render_result.py  logs/AntMaze/visualization 6000  # for AntMaze

python render_result.py  logs/AntPush/visualization 6000  # for AntPush

python render_result.py  logs/Reacher3D/visualization 4800  # for Reacher3D

To train BrHPO,

python main.py --env_name AntMaze

Citation

If you find our work useful, please consider citing our paper as follows:

@inproceedings{Luo2024BrHPO,
  title={Bidirectional-Reachable Hierarchical Reinforcement Learning with Mutually Responsive Policies}, 
  author={Yu Luo and Fuchun Sun and Tianjing Ji and Xianyuan Zhan},
  booktitle={Reinforcement Learning Conference},
  year={2024}
}