Incremental Traffic Assignment: A Perturbation Approach

Abstract

This paper develops a new algorithmic approach to equilibrium road traffic assignment which, by directly estimating differences, can more accurately estimate the impact of (relatively) small traffic schemes or changes in the demand pattern. Comparing the outputs of two independent traffic assignments to “with” and “without” scheme networks very often masks the effect of the scheme due to the “noise” in the resulting solutions. By contrast an incremental approach attempts to directly estimate the changes in link flows - and hence costs - resulting from (relatively) small perturbations to the network and/or trip matrix. The algorithms are based firstly on “route flows” as opposed to “link flows“, and secondly, they use a variant of the standard Frank-Wolfe algorithm known as “Social Pressure” which gives a greater weight to those O-D path flows whose costs are well above the minimum costs as opposed to those which are already at or near minimum. Tests on a set of five “real” networks demonstrate that the Social Pressure Algorithm is marginally better than Frank-Wolfe for single assignments but is very much faster and more accurate in predicting the impact of small network changes.

Citation

Kupiszewska, D. and Van Vliet, D. (1998), "Incremental Traffic Assignment: A Perturbation Approach", Griffiths, J.D. (Ed.) Mathematics in Transport Planning and Control, Emerald Group Publishing Limited, Leeds, pp. 155-165. https://doi.org/10.1108/9780585474182-015