Spatiotemporal clustering of urban networks: Real case scenario in London

Abstract

The problem of modeling traffic in urban areas is especially relevant because of the urgent need for managing congestion in big cities. In recent years, the macroscopic fundamental diagram (MFD) has been proposed as a macroscopic description of urban traffic. The MFD has been proved useful in the design of control strategies. However, the MFD cannot be properly defined if congestion across the city is heterogeneously distributed and if its characteristics change with time. This situation is common in large cities, and a method for identifying homogeneous areas in space and time is needed. This paper proposes a spatiotemporal clustering method with which to detect homogeneous areas over the network and over time. In each of these areas, an MFD can be defined. In particular, a segmentation method is proposed that can divide the time intervals into segments optimizing the clustering results within each of them. The algorithm is applied to data collected in the city of London, where a control system called SCOOT (split-cycle offset optimization technique) is in operation.