Abstract

Traffic congestion in urban areas is a big issue for cities around the world. Thus, studying congestion and respective counter measures is of high importance for the increasing urbanization of society. Congestion analysis and forecast is most of the times done either on a link-wise network or on a network-wide level. Though, due to bottlenecks in the infrastructure and similar commuting patterns by road users, usually the same parts of an urban traffic network get congested. The idea is to observe and investigate primarily these most vulnerable parts of the network, which are denoted as congestion clusters, as they are crucial to both, drivers and operators. A methodology for determining congestion clusters is described, which provides a significant amount of flexibility to be able to meet different needs for different applications or cities. Based on a five months set of Floating Car (FC) data, the suggested methodology is tested. First analyses are conducted to understand up to which degree these clusters are able to represent the congestion level of the entire network. Besides, correlations between the clusters are investigated on a statistical basis and conclusions are drawn. The results provide a basis for potential traffic estimation and forecast systems.