Skip to content

honghaow/FORK

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FORK

Author's PyTorch implementation of FORK: A Forward-Looking Actor For Model-Free Reinforcement Learning. The paper can be found here.

We proposed a new type of Actor, named forward-looking Actor or FORK for short, for Actor-Critic algorithms.

FORK can be easily integrated into a model-free Actor-Critic algorithm.

Usage

TD3-FORK and SAC-FORK are tested on Mujoco continuous control tasks in OpenAI gym.

Neural Networks are trained using Pytorch 1.4 and Python 3.7

The results in the paper can be reproduced by running:

./run_td3.sh
./run_td3_fork.sh
./run_sac.sh
./run_sac_fork.sh

BipedalWalkerHardcore

BipedalWalkerHardcore is a advanced version of BipedalWalker with ladders, stumps, pitfalls.

TD3-FORK can slove the task with as few as four hours by using the defaulat GPU setting provided by Google Colab.

You can view the performance on Youtube.

Bibtex

If you use our code or data please cite the paper.

@article{wei2020fork,
  title={FORK: A Forward-Looking Actor For Model-Free Reinforcement Learning},
  author={Wei, Honghao and Ying, Lei},
  booktitle={2021 IEEE 60th Annual Conference on Decision and Control (CDC)},
  year={2021}
}

Acknowledgement

The TD3 code was based on TD3 The SAC code was based on SAC1 and SAC2.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published