A review of (dis)aggregation and decomposition methods in traffic assignment

Abstract

In this study we provide a comprehensive review of the existing literature on (dis)aggregation and decomposition methods in traffic assignment. The study takes on two different perspectives. First, we explore existing methods and relate them to their impact on one or more of the components involved in traffic assignment. It is found that there exists a clear separation between a demand modelling point of view, i.e. travel demand and (geographical) zoning, and supply modelling oriented methods, i.e. network topology and network loading. At the same time, the interface between demand and supply, i.e. connector and centroid placement which is to be considered a special type of aggregation, has received conspicuously little attention in this context, even though it is shown to be of significant impact on modelling results. The second perspective in this study places the discussed aggregation methodologies in the broader perspective of clustering procedures. It is shown that most existing methods can be classified as supervised – or classification based – clustering procedures while relatively few studies explore other known approaches such as semi-supervised or unsupervised clustering techniques, at least from a purely traffic assignment model point of view. Lastly, we discuss how aggregation techniques can be deployed to construct multi-scale modelling environments. There is however a lack of methodology to construct such models consistently. It is hoped that this work provides a first step in the direction of such developments by providing an objective classification framework for existing (dis)aggregation and decomposition methods.