An extension of Dial's algorithm utilizing a model of tripmakers' perceptions

Abstract

Dial has developed a probabilistic multipath traffic assignment algorithm which is very efficient. However, the assignment of trips to paths is based on a logit model which has some undesirable properties. The purpose of this paper is to extend Dial's algorithm by utilizing a simple model of tripmakers' perceptions and behavior. An important advantage these models have is that they can be calibrated through the study of tripmakers' behavior on a limited set of trips rather than through the study of origin and destination patterns and link flows on a large network. Three traffic assignment models are presented which are computationally equivalent to Dial's model and which can be implemented by making minor changes in his algorithm. One of the new models, based on path likelihoods, contains Dial's model as a special case. The assumptions required about tripmakers' perceptions to obtain this special case seem unreasonable and this explains some of the undesirable properties of Dial's model. In its general form, the new model eliminates some of these undesirable properties. However, like Dial's model, it generally allocates a disproportionate number of trips to portions of a road network in which there is a large number of choices of minor variations in route. The other two models, based on arrival likelihoods and turn likelihoods, both overcome this shortcoming. These two models have different interpretations in terms of tripmaker behavior. The model based on turn likelihoods is the most intuitive and contains Gunnarsson's multipath traffic assignment model as a special case.