ALL ABOUT CONGESTION: MODELING DEPARTURE TIME DYNAMICS AND ITS INTEGRATION WITH TRAFFIC MODELS

Abstract

This thesis comprehensively studies departure time choice models, and analyzes the consequent system-level peak spreading effects. In modeling, the school of discrete choice models successfully reveals the user heterogeneity. A mixture logit model and a latent class model based on the notion of carpooling preference have been estimated. Then a novel positive approach has been developed, which avoids the assumptions of rationality and focuses on how individuals actually make departure time decisions. Following this positive theory, we specify Bayesian learning, empirically estimate search start and stopping conditions that vary among users, and empirically derive search and decision rules from a joint reveal/stated-preference survey dataset. This innovative behavioral model is integrated with a traffic simulation model for a real-world study. Findings from this application reveal the potential of the proposed model to capture network dynamics and behavioral reactions. This integrated framework also provides a valuable tool for the evaluation of new transportation infrastructures, policies, and operation strategies.