A Data-Driven Optimization-Based Approach for Freeway Traffic State Estimation Based on Heterogeneous Sensor Data Fusion
Jinyu Zhang,Di Huang,Zhiyuan Liu,Yifei Zheng,Yu Han,Pan Liu,Wei Huang
DOI: https://doi.org/10.1016/j.tre.2024.103656
2024-01-01
Abstract:Accurate estimation of freeway traffic states is crucial for designing effective traffic management and operational strategies. The integration of various sensor data, such as data from the Electronic Toll Collection (ETC) system and traffic detectors, can significantly enhance the granularity and coverage of traffic state estimation. This study introduces a data-driven optimization-based approach for estimating freeway traffic states, leveraging the fusion of ETC data with detector data. This methodology capitalizes on the broad coverage provided by ETC data and the fine granularity offered by detector data. The probabilistic interdependence between the traffic state of a segment and its upstream and downstream counterparts is captured from real-world traffic state data. Two optimization models, based on the maximum likelihood and maximin likelihood principles, are developed to accurately depict the distribution patterns of freeway traffic states. To address the computational challenges of large-scale scenarios, the study proposes both a decomposition algorithm and a heuristic algorithm. A case study utilizing real-world data from the G92 freeway in Zhejiang, China, is conducted. The findings indicate that the two optimization models exhibit commendable accuracy, with mean absolute percentage errors of 0.9% and 2.3% during peak hours, and 0.9% and 1.4% during off-peak hours, respectively.