Effects of En-Route Choices on Utility-Based Route Choice Modeling: an Empirical Analysis
Dawei Li,Tomio Miwa,Takayuki Morikawa
2013-01-01
Abstract:This paper discusses the effects of en-route choices on route choice modeling, with the utility maximizing behavior assumption. Route choice modeling is essential to traffic assignment and traffic demand prediction. In previous research, it is often assumed that drivers will only make pre-trip route choices, i.e. the route choices at the origin. However, drivers will also change their route choice decisions during the current trips(i.e. en-route choices), especially with the effect of real time information providing. At first, this paper discusses the potential effects of en-route choices on the prediction of route shares, under the framework of random utility theory. These effects are mainly caused by three factors: the dynamic feature of route characteristic perceptions, the probability of making en-route choices at each decision node, and the response taste heterogeneity which is choice situation specific. These effects are illustrated using a simple example. Then a hybrid model is proposed to considering en-route choices explicitly. In this model, whether making en-route choice or not at each decision node is treated as a latent variable, which is related to the dynamic perceptions of route characteristics. Both observed and unobserved taste heterogeneity are incorporated in a mixed-logit model. A simulated data set on a test network is generated for the empirical analysis. Using the synthetic data set, proposed choice model is estimated, along with several less general models, in which not all of the three factors are considered. The estimation results are compared with each other and also with the parameters for data generation. With the comparison, the potential biases because of not considering en-route choices are illustrated. At last, because the proposed model is too complicate for some practical applications, an approximate model is also proposed. In this model, instead of calculating the probability of making en-route choices at decision nodes, a correction term, which is related to the dynamic perceptions of route characteristics, is added in the specification of route utility. The approximated model is estimated and evaluated using the same synthetic data set.