An Integrated Choice and Latent Variable Framework to Incorporate the Influence of Travel Time Variability on Truck Route Choice

Abstract

Route choice models (or path choice models) are useful for quantifying travellers’ preferences for or sensitivity to route attributes, predicting network-level traffic flows, examining the influence of information provided to travellers, and studying travellers’ adaptation to uncertainty in travel conditions. Among the various factors influencing route choice, variability in travel conditions is an influential one. Day-to-day and within-day variations in travel conditions influence route choice decisions in many geographical contexts. Empirical studies on values of time and reliability have concluded that travellers, besides being interested in minimizing their travel times, also wish to minimize their travel time variability. The influence of travel time variability on route choice becomes more important in the context of freight transportation and logistics where delays due to uncertainty translate to large financial losses. Therefore, it is useful to quantify variability in travel conditions and to understand the influence of such variability on freight route choice decisions. This thesis proposes an Integrated Choice and Latent Variable (ICLV) modelling framework that allows simultaneous estimation of route-level travel time variability and incorporation of the influence of such variability on travel route choice of freight-trucks. The proposed framework considers the travel time on a route as a latent (unobserved) variable and uses GPS data measurements of route-level travel time to identify the parameters of its statistical distribution. Since such measurements are not always available for all routes, the latent variable component of the ICLV framework helps impute or inform the travel time distribution for routes without travel time measurements. In this regard, simultaneous estimation of the measurement and choice components of the proposed model allows the use of partial measurement data for estimation of travel time variability as well as incorporation of the influence of travel time variability on route choice. Further, route-level travel time variability is viewed as a result of variability in travel conditions (e.g., variability of travel speeds on links, etc.) and is captured through random coefficients on the route attributes specified in the latent variable model. The proposed model is applied to an empirical data set on truck route choice using truck-GPS data collected in the Tampa Bay region of Florida, USA. The empirical parameter estimates suggest that the variability of travel time on a route depends on the network structure along the route, such as the lengths of different roadway types, largely due to differences in variability of travel speeds among different types of roadways. The empirical findings indicate a superior statistical model fit of the proposed ICLV model than the traditional choice models that do not consider the influence of travel time variability on route choice. Although the ICLV model in this study was applied to the empirical context of freight-truck route choice, the proposed framework is applicable to accommodate the influence of variability in travel conditions on other travel choices such as transit route choice and travel mode choice; thanks to the increasing ubiquity of passively collected data on travel time (such as GPS data).