SR-Traffic::Discovering Macroscopic Traffic Flow Models with Symbolic Regression
Year: 2025
Type: conference-paper
Abstract: Traffic flows are complex systems that can be studied from a macroscopic perspective. In particular, first-order models are tractable but oversimplified, while higher-order models capture richer dynamics at the cost of complexity. Here, we introduce SR-Traffic, a data-driven, physics-informed framework that uses symbolic regression to learn effective phenomenological relations directly from experimental data while embedding them into an efficient, first-order PDE formulation. Our approach balances accuracy and interpretability, ensures physical consistency, and shows good generalization, overcoming the limitations of purely data-driven models. Overall, our findings could support the design of digital solutions for improving mobility systems.
Authors: Manti, Simone; id_orcid 0000-0002-4060-0620, Mohammadian, Saeed, Treiber, Martin, Lucantonio, Alessandro
Institutions: The University of Queensland, Technische Universität Dresden
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Awards: G753438505, G8356784893
OpenAlex ID: W7112896084
FWCI:
Cites:
Cited by:
Related to: