Using agent-based of driver behavior in the context of car park optimization

Models based on activity of population have been chosen for a realistic representation of travel demand. In this paper, firstly we describe special features of a multimodal urban traffic simulator based on multi agent technology. By binding numerical simulation and behavioral aspects, we are developing a hybrid simulator designed as a decision aid tool; the user-player can test different scenarios by immersion like in a video game. Secondly, we focus on the description of an approach to model realistic choices (itinerary, transport mode or car park) of a driver agent. We show that the agent behavior can be defined with a standard mathematical formulation using real data resulting from particular questionnaires. We adapted the design of experiments method for this new approach. The paper suggests further contributions: the behavioral models take into account tangible (e.g. distance, cost) and intangible (e.g. weather, accessibility) variables. They can be adapted according to particular conditions of traffic of any city, generate the memory effect of individual agents and reduce needed data for each agent in order to improve the simulator performance. We illustrate the approach by using an example on the choice of car park. The behavior of a driver who does not find an available place to park is modeled by coupling the "design of experiments" method and the belief theory