Estimation of discrete choice models for route choice from data collected using the rapp-up smartphone application

Abstract

A smartphone-based route choice stated and revealed preference application called RAPP-UP was used to collect route choice preference data from a small sample of Gauteng car commuters for their trip to work to demonstrate the application’s proof of concept. The application presented choice sets with two route alternatives to survey participants on their smartphone using real-time values of travel time and cost. The route choice preference data was used to estimate discrete choice models to simulate route choice in congested urban conditions as well as willingness to pay (WTP) measures such as the non-work related value of travel time (VTT). The utility attributes included travel time (disaggregated into free-flow, slowed-down and stop-start travel time); trip petrol cost; trip toll cost (if any); and the probability of arriving at the trip destination on time. This paper presents the results of the discrete choice models that were estimated based on this data set. Confirming international evidence, the models showed that the sample of commuters perceived statistically significant differences between the attribute coefficients of the three trip time categories and their associated values of travel time. The trip petrol cost, toll cost and on-time arrival attribute coefficients were also significant. When a Toll Road Quality Bonus (TRQB) was included in the utility equation as a dummy attribute the coefficient was statistically significant but had a negative sign. The sample of survey participants thus perceived the use of Gauteng tolled freeways negatively for the unobserved factors of utility. The route choice models demonstrated the proof of concept of the RAPP-UP application as a convenient, practical, low-cost tool for collecting route choice preference data on congested urban road networks and provides the basis for its application with a larger sample of car commuters.