parl_agents

This repository offers pytorch based Hierarchical RL agents extending stable-baselines3

Related Repos

we use parl_agent, parl_benchmark, parl_annotations together.

• parl_agents: Hierarchical RL agent codes
• parl_minigrid: Add-on to the minigrid environemtns
• adding different kinds of annotation to RL task, we extend parl_annotations
• adding new annotated RL environments, we addd new parl_benchmark such as parl_minigrid

Install

• first create a conda environment for installing parl_annotations, parl_agents, parl_minigrid.

$ conda create -n parl python=3.7

• install packages as editable library

pip install -r requirements
pip install -e .

Stable Baselines3

• To avoid stable-baselines version issues, this repo stores src from

pip install -e git+https://github.com/DLR-RM/stable-baselines3.git@v1.5.0#egg=stable_baselines3

• We can directly install stable_baselines3 using setup.py

$ cd src; pip install -e .

Usage

There are sample scripts for running hppo, ppo, and dqn agents under test_scripts.

Citations

• 2021 ICAPS PRL Workshop paper

@inproceedings{lee2021ai,
  title={AI Planning Annotation in Reinforcement Learning: Options and Beyond},
  author={Lee, Junkyu and Katz, Michael and Agravante, Don Joven and Liu, Miao and Klinger, Tim and Campbell, Murray and Sohrabi, Shirin and Tesauro, Gerald},
  booktitle={Planning and Reinforcement Learning PRL Workshop at ICAPS},
  year={2021}
}

• 2023 NEURIPS GenPlan Workshop paper

@inproceedings{lee2021ai,
  title={Hierarchical Reinforcement Learning with AI Planning Models},
  author={Lee, Junkyu and Katz, Michael and Agravante, Don Joven and Liu, Miao and Tasse, Geraud Nangue and Klinger, Tim and Sohrabi, Shirin},
  booktitle={Generalization in Planning GenPlan Workshop at NEURIPS},
  year={2023}
}