Deployable hybrid probabilistic-possibilistic driver route choice models for real -time network control using information provision strategies
Abstract
The commonly proposed dynamic traffic assignment models for the real-time operational control of large-scale vehicular traffic networks under information systems are behaviorally restrictive and limited in their ability to model driver response to information. This research seeks to robustly predict driver route choice behavior under real-time information provision using observable field data so as to develop effective and deployable information provision strategies. The problem is characterized by linguistic variables, subjective interpretation, heterogeneity, and situational factors. Fuzzy modeling provides a synergistic mechanism to capture subjectively interpreted and/or linguistically labeled qualitative variables. This motivates the development of a hybrid probabilistic-possibilistic route choice model that combines quantitative and qualitative variables in a single framework to more robustly predict driver routing decisions under information provision. The approach uses transparent if-then rules to represent driver behavioral tendencies and membership functions to characterize qualitative phenomena that govern driver route choices under information provision. This enables the hybrid model to provide a single consistent framework to capture the day-to-day evolution and the within-day dynamics of driver behavior. Insights from simulation experiments suggest that the hybrid model can reflect the evolution of driver behavior over time, and adapt to the within-day variability in ambient driving conditions. From a deployment perspective, the hybrid model is amenable to seeking operational consistency with the real-time traffic flow measurements. This leads to a behavior-based consistency-seeking paradigm to deploy real-time information provision strategies to influence the traffic system performance. The behavior-based consistency-seeking model explicitly considers heterogeneous driver behavior classes vis-à-vis driver route choice behavior and seeks the driver behavior class fractions that minimize the difference between the predicted network states based on driver route choices and the actual conditions unfolding in real-time. From a computational standpoint, the approach is real-time deployable. Thus, behavior-based consistency-seeking models serve as alternatives to dynamic traffic assignment models to develop deployable information supply strategies to control system performance. Simulation experiments suggest that a realistic modeling of driver behavior and information characteristics is essential to realizing the benefits of information provision vis-à-vis real-world deployment. Behavior-based consistency-seeking models ensure greater realism in this context compared to dynamic traffic assignment models.
Degree
Ph.D.
Advisors
Peeta, Purdue University.
Subject Area
Civil engineering|Transportation
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.