Revealing a Hidden, Stable Spectral Structure of Urban Vehicular Traffic

A deeper understanding of urban vehicular traffic is important to enable better design and evaluation of future vehicular and cellular communication networks. In this paper, we study the presence of spectral structure in urban vehicular traffic. By analyzing publicly available sets of fleet vehicle mobility traces obtained from two real-world deployments that consist of more than 2,000 taxis in Shanghai and Beijing respectively, we reveal the existence of a stable, low-dimensional spectral structure in vehicular networks, which was often unnoticeable when using classic spatio-temporal data analysis. This stable spectral structure not only significantly simplifies the representation of high dimensional transportation data, but also offers interpretable insights into urban mobility patterns. Leveraging the stability of spectral structure, we demonstrate that the spectral structure analysis could effectively tackle practical problems in the field of transportation research, such as traffic anomaly detection.