HiLight

Hierarchically and Cooperatively Learning Traffic Signal Control

Introduction

This project HiLight has been published as the following conference paper in AAAI:

@inproceedings{hilight,
  author    = {Bingyu Xu and
               Yaowei Wang and
               Zhaozhi Wang and
               Huizhu Jia and
               Zongqing Lu},
  title     = {Hierarchically and Cooperatively Learning Traffic Signal Control},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  series = {AAAI '21},
  pages     = {669--677},
  publisher = {{AAAI} Press},
  year      = {2021},
}

Before running above codes, you may need to install following packages or environments:

• Python 3.6
• CityFlow 0.1 (Installation Guide: https://cityflow.readthedocs.io/en/latest/install.html)
• Keras 2.3.1

Start an experiment by:

python -O runexp.py

Here, -O option cannot be omitted unless debug is necessary. In the file runexp.py, the args can be changed.

• runexp.py

  Run the pipeline under different traffic flows. Specific traffic flow files as well as basic configuration can be assigned in this file. For details about config, please turn to config.py.

For most cases, you might only modify traffic files and config parameters in runexp.py.

To run different road networks, set the parameters in runexp.py, traffic_light_dqn.py, map_computer.py accordingly.

Dataset

• conf/

Configuration files.

• data/

Traffic flow files.

• synthetic data

  Traffic file and road networks can be found in data/4_4.

• real-world data

  Traffic file and road networks can be found in data/jinan && data/manhattan_16_3 && data/shenzhen (the last is our original dataset).

Structure

Other files follow the project structure of IntelliLight: https://github.com/wingsweihua/IntelliLight

@inproceedings{wei2018intellilight,
  title={IntelliLight: A Reinforcement Learning Approach for Intelligent Traffic Light Control},
  author={Wei, Hua and Zheng, Guanjie and Yao, Huaxiu and Li, Zhenhui},
  booktitle={Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining},
  pages={2496--2505},
  year={2018},
  organization={ACM}
}

The files are functioning like this:

• runexp.py

Load the experiment setting, and call traffic_light_agent.

• traffic_light_dqn.py

Simulate the reinforcement learning agent. This file simulates the timeline, get current state from sumo_agent, get current state from sumo_agent and call the agent to make decision.

• sumo_agent.py

Interact with map_computor to get state, reward from the map_computor, and convey action to map_computor.

• map_computor.py

Read data from CityFlow and operate CityFlow.

• agent.py

Abstract class of agent.

• network_agent.py

Abstract class of neural network based agent.

• deeplight_agent.py

Class of our method.

Baseline

CoLight: https://github.com/wingsweihua/colight

@inproceedings{colight,
 author = {Wei, Hua and Xu, Nan and Zhang, Huichu and Zheng, Guanjie and Zang, Xinshi and Chen, Chacha and Zhang, Weinan and Zhu, Yamin and Xu, Kai and Li, Zhenhui},
 title = {CoLight: Learning Network-level Cooperation for Traffic Signal Control},
 booktitle = {Proceedings of the 28th ACM International Conference on Information and Knowledge Management},
 series = {CIKM '19},
 year = {2019},
 location = {Beijing, China}
}