End-to-End Learning Framework for Transportation Network Equilibrium Modeling

End-to-End Learning Framework for Transportation Network Equilibrium Modeling

Headshot of Yafeng Yin. The link directs to their profile page on the CCAT website.
Yafeng Yin
The University of Michigan Logo. The link directs to the funded research led by this institution.

Principal Investigator(s):

Yafeng Yin, Donald Cleveland Collegiate Professor Of Engineering – The University of Michigan
Donald Malloure Department Chair Of Civil And Environmental Engineering – The University of Michigan

Project Abstract:
This project aims to outline a groundbreaking “end-to-end” transportation demand modeling framework, driven by deep learning techniques and empirical multi-source data. Unlike traditional models, which typically employ a multi-step process, this framework directly associates time-series observations of traffic patterns, urban land use, and socioeconomic features with prediction of future traffic flow distributions. The end-to-end modeling framework is designed to learn travelers travel and route choices while refining link performance functions that estimate travel time based on traffic flow. After calibration against empirical data, the proposed framework can recommend optimal policies or projects for enhancement, thereby facilitating informed decision-making. By utilizing passively collected trajectory data, this framework aims to significantly improve modeling accuracy and the realism of behavioral representation, without additional costs for data collection in the existing modeling system.

Institution(s): University of Michigan – Ann Arbor

Award Year: 2024

Research Focus: Mobility

Project Form(s):