101 Uses for Path-based Assignment



101 Uses for Path-based Assignment

Authors

KUPSIZEWSKA D and VAN VLIET D, University of Leeds, UK 121

Description

Traditionally traffic assignment models are based on flows per link as opposed to flows per individual 0-D pair although, in principle, the latter is the basic building block. The basic issue is the computer RAM required. This paper proposes a set of link

Abstract

Traditionally traffic assignment models are based on flows per link as opposed to flows per individual 0-D pair although, in principle, the latter is the basic building block. The basic issue is the computer RAM required. This paper proposes a set of linked traffic assignment algorithms based on path flows which, it is shown have many desirable properkes and are computationally feasible with current-day Pc's. Thus they enable faster convergence rates and more accurate estimates of small changes in flows to be achieved via "perturbation" techniques. They are also very useful in situations where the trip matrix to be assigned is variable, not fixed, since a good starting point for an assignment with an updated trip matrix may be generated by factoring path flows pro rata (which is not possible with pure link flows). This occurs with "elastic" assignment models linked to, e.g., distribution, modal split or time-period choice models ("peak- spreading"). Their efficiency is demonstrated using a set of real-life networks.

Finally, path flows easily and quickly provide useful information for the analyses of traffic flow patterns, for example to determine which precise trips are on a particular link or which paths a particular 0-D pair uses. They also provide essential linkages between aggregate assignment models and more detailed vehicle-by-vehicle simulation models. In such cases more efficient algorithms may be developed.

Publisher

Association for European Transport