Traffic Network Partitioning for Hierarchical Macroscopic Fundamental Diagram Applications Based on Fusion of GPS Probe and Loop Detector Data

Abstract: Most network partitioning methods for Macroscopic Fundamental Diagram are mostly based on a normalized cut mechanism, which takes the traffic statistics of each link, e.g. link volume, speed or density, as input to calculate the degree of similarity between two links, and perform graph cut to divide a network into two sub-networks at each iteration when the traffic dynamics between links are dramatically different. These methods assume complete link-level traffic information over the entire network, e.g. the accurate measurement of the traffic conditions exist for every single link in the network, which makes them inapplicable when it comes to real-world setting. In this paper, we propose a method which, based on fusing Probe Vehicle Data and loop detector data, extracts the locally homogeneous subnetworks with a grid-level network approach instead of dealing with detailed link-level network. By fusing the two data sources, we take advantage of both better coverage from PVD and the full-size detection from LDD.