Uncovering the contribution of travel time reliability to dynamic route choice using real-time loop data

Travel time reliability has generally been surmised to be an important attribute of transportation systems. In this paper, we study the contribution of travel time reliability in travelers' route choice decisions. Traveler's route choice is formulated as a mixed-logit model, with the coefficients in the model representing individual traveler's preferences or tastes towards travel time, reliability and cost. Unlike the traditional approach involving the use of traveler surveys to estimate model coefficients and thereby uncover the contribution of travel time reliability, we instead apply the methodology to real-time loop detector data, and use genetic algorithm to identify the parameter set that results in the best match between the aggregated results from traveler's route choice model and the observed time-dependent traffic volume data from loop detectors. Based on freeway loop data from California State Route 91, we find that the estimated median value of travel-time reliability is significantly higher than that of travel-time, and that the estimated median value of degree of risk aversion indicates that travelers value a reduction in travel time variability more highly than a corresponding reduction in the travel time for that journey. Moreover, travelers' attitudes towards congestion are not homogeneous; substantial heterogeneity exists in travelers' preference of travel time and reliability. Our results validate results from previous studies involving the California State Route 91 value-pricing project that were based on traditional traveler surveys and demonstrate the applicability of the approach in travelers' behavioral studies.