A Public Traffic Demand Forecast Method Based on Computational Experiments

Abstract

The conventional mathematical models that are used for traffic distribution and traffic mode choice forecasts consider neither the individual heterogeneity on the micro level nor the changeable traffic scenes. This prompted us to propose a new forecast method composed of a traffic survey, an artificial transportation system (ATS), and computational experiments. We introduced a BDI modeling method in the agent-based ATS. This method considers an individual's psychological characteristics in combination with logical thinking, which was introduced to individual passenger agents, to deduce each passenger's decision-making process when choosing the traffic mode and route. A series of computational experiments were conducted on the ATS by using a school bus system as a case study to validate the feasibility and superiority of our method. Several computational experiments were conducted to predict the traffic distribution in normal and abnormal traffic scenarios and to analyze the extent to which each factor influences the travel modal split. Furthermore, the outcomes of various vehicle-scheduling plans were predicted and analyzed by using computational experiments to determine the optimal plan and support the establishment of transportation policies in the real world.