Hierarchically and Cooperatively Learning Traffic Signal Control

Authors: Bingyu Xu, Yaowei Wang, Zhaozhi Wang, Huizhu Jia, Zongqing Lu669-677

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we demonstrate that Hi Light outperforms state-of-the-art RL methods for traffic signal control in real road networks with real traffic.
Researcher Affiliation Collaboration Bingyu Xu1, Yaowei Wang1, Zhaozhi Wang2, Huizhu Jia2, Zongqing Lu2 1Peng Cheng Laboratory 2Peking University
Pseudocode Yes Algorithm 1 Hi Light training
Open Source Code No The paper mentions using 'City Flow (Zhang et al. 2019), an open-source simulator' and using 'open-source implementations' for baselines, but does not provide a link or state that the authors are releasing the source code for their own method, Hi Light.
Open Datasets Yes The road networks and traffic flows of Jinan, Hangzhou, and New York City are the public datasets1. 1https://traffic-signal-control.github.io/
Dataset Splits No The paper mentions training and evaluating performance, but does not explicitly provide specific train/validation/test dataset splits, percentages, or sample counts needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments, such as GPU models, CPU types, or cloud computing instance specifications.
Software Dependencies No The paper mentions using City Flow simulator, and algorithms like DQN and PPO, but does not provide specific version numbers for any software dependencies needed to replicate the experiment.
Experiment Setup Yes Hyperparameters Table 1 summarizes the hyperparameters of Hi Light, Co Light, and Press Light. For Co Light and Press Light, we use their open-source implementations. For fair comparison, we also use their default parameter settings which perform better than other settings as verified by experiments.