juanwulu / sac_vsl

A variable speed limit control algorithm designed with the Soft Actor-Critic reinforcement learning.

Soft Actor-Critic for Variable Speed Limit Control

A variable speed limit control algorithm designed with the Soft Actor-Critic reinforcement learning.

This is a final project for the individual research at UC Berkeley. Please see my final report for more details about this project.

Installation

Use the package manager pip to install the dependencies locally.

git clone -b master --depth 1 https://github.com/ChocolateDave/sac_vsl.git
cd sac_vsl & pip install -r requirements.txt & pip install -e .

Usage

A script is provide for running our codes on different penetration rate settings.

mkdir logs/ & bash sac_vsl/scripts/train_sac_multi_pr.sh

Citation

If you use this source code, please cite it using bibtex as below.

@misc{Juanwu2022,
  author = {Juanwu Lu},
  title = {Reinforcement Learning for Freeway Variable Speed Limit Control: A Mixed Traffic Flow Case Study},
  year = {2022},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/ChocolateDave/sac_vsl}},
  commit = {2b675bac077bc695048ce0072f254de25c898050}
}

License

This project is licensed under the BSD 3-Clause License