Estimating MFD model parameters from sparse license plate recognition data: The role of path reconstruction and regionalization

Abstract

The Macroscopic Fundamental Diagram (MFD) provides a convenient and computationally efficient tool for urban traffic monitoring and control. However, the accurate estimation of the model parameters is a prerequisite for the MFD models to be practically applicable. The significant deficiencies in commonly used data sources underscore the importance of utilizing new data sources to estimate the MFD model parameters. This study proposes a comprehensive framework to estimate critical parameters for multi-region MFD models based solely on License Plate Recognition (LPR) data. An efficient link-based path reconstruction algorithm is introduced to transform observed vehicle trajectories into continuous paths. Considering the inherent incompleteness of observed trajectories, this study presents a novel regionalization framework to match initial or final observation points to specific subregional networks. An adaptive large neighborhood search (ALNS) algorithm is proposed to improve the flow correlation within these subregions. Subsequently, an OD inference model using importance sampling is introduced to estimate the true OD of trips within the identified subregions. The resulting complete vehicle trips can be used to estimate trip length distributions and path flow coefficients. In addition, a method for estimating the MFD shape, tailored for application to sparse LPR data and based on reconstructed vehicle trajectories, is proposed. The effectiveness of these methods is validated through both simulation experiments and empirical studies. This framework is expected to facilitate the practical application of MFD models to realistic road networks.