SR-Traffic

This repository contains the code used to produce the results of the paper SR-Traffic: Discovering Macroscopic Traffic Flow Models with Symbolic Regression

Installation

The dependencies are collected in environment.yaml and can be installed, after cloning the repository, using mamba:

$ mamba env create -f environment.yaml

Once the environment is installed and activated, install the library using

$ pip install -e .

Usage

To reproduce the results of the paper just run

$ python src/sr_traffic/fund_diagrams/fund_diagrams_results.py --road_name {road_name} --task {task_name}

where {road_name} is either US101 or US80, and {road_name} is either prediction or reconstruction.

To re-calibrate a given fundamental diagram, run

$ python src/sr_traffic/fund_diagrams/fund_diagrams_calibration.py  --config src/sr_traffic/fund_diagrams/configs/{fnd_name}.yaml

where{fnd_name} is either greenshields, triangular, Weidmann, del_castillo, or idm.

Finally, to perform a run of SR-Traffic, run

$ python src/sr_traffic/learning/stgp_traffic.py

You can change the parameters of the algorithm modifying stgp_traffic.yaml.

Citing

@article{mantisr,
  title={{SR}-{T}raffic: {D}iscovering {M}acroscopic {T}raffic {F}low {M}odels with {S}ymbolic {R}egression},
  author={Manti, S. and Mohammadian, S. and Treiber, M. and Lucantonio, A.},
  journal={Neural Information Processing Systems, ML4PS Workshop},
  year={2025}
}