Dynamic decision and adjustment processes in commuter behavior under real-time information.

摘要

Advanced Traveler Information Systems (ATIS), by providing real-time traffic information, can assist trip-makers in selecting efficient travel choices, and aid the attainment of desirable system goals including reduced costs and increased efficiencies. The success of ATIS in achieving such goals critically depends on user behavior in response to information. This research focuses on investigating dynamic aspects in commuter behavior under real-time information. A dynamic interactive travel-behavior simulator, that enables a consistent representation of the nonlinear time-dependent interactions between network performance, trip-makers choices, and information, is used to observe trip-maker behavior. Using the simulator, interactive experiments are conducted where a range of experimental factors including network loading, day-to-day traffic evolution and ATIS information strategies are varied and the consequent trip-maker behavior is observed. Constituent models are proposed to analyze the choice dimensions of route, departure time, and compliance. The dynamic kernel logit (DKL) formulation is presented for analyzing these data and its theoretical and computational suitability established. The results confirm the significance of compliance and inertia as key mechanisms influencing route choice. Departure time adjustments appear to be based on a sequential heuristic search. Calibrated models also provide evidence of learning, adjustment, perception, judgment, and updating processes in trip-maker behavior. Empirical results indicate that real-time information and time-dependent network conditions are strong determinants of trip-maker behavior in a commuting context. The nature and quality of ATIS information (accuracy and reliability), the magnitude of network loading and its day-to-day evolution, and users' past traffic experience are important influences on how commuters select routes and departure times. At the unobserved level, general dynamic and stochastic patterns, including, heterogeneity, state-dependence, habit-persistence, and correlations are present in trip-makers' decisions. These substantive results have important implications for network state prediction, travel demand forecasting, design and evaluation of ATIS services and deployment of Intelligent Transportation System (ITS) programs. User behavior models developed here can be integrated with dynamic network traffic assignment models to obtain more accurate system performance modeling capabilities with considerable applications in tactical and strategic system planning and traffic operations.