Capturing Spatial-Temporal Traffic Patterns: A Dynamic Partitioning Strategy for Heterogeneous Traffic Networks

Macroscopic fundamental diagram (MFD) has become a popular model used in developing network traffic controls for a roughly homogeneous traffic network, encounters limitations when applied to the inherently heterogeneous nature of real-world transportation networks, affecting its predictive accuracy and applicability. To address these challenges, this paper proposes a dynamic network partitioning strategy for heterogeneous traffic networks. We devise a spatial-temporal dual-form graph to accurately represent the road network’s spatial-temporal traffic patterns. To solve the dynamic network partitioning problem, we employ a spectral theory technique known as RatioCut. Our evaluation of the partitioning methods’ effectiveness relies on fitting performance to the MFD and network modularity as metrics. A case study set in Yangzhou, China, executed with the Simulation of Urban Mobility (SUMO), demonstrates the performance of our dynamic partitioning approach. The results highlight our method’s ability to capture the spatial and temporal evolution of the congestion in the road network and cluster the heterogeneous network into multiple quasi-homogeneous regions. Moreover, it shows that speed is potentially a more accessible indicator for network partitioning as it performs similar to density and is easier to collect.