Agent-based simulation including traffic participants’ personal characteristics in pedestrian–vehicle mixed space

In order to ensure the safety and the travel efficiency of traffic participants, it is essential to verify driving systems and to design traffic environments using simulations. In local areas where vehicles and pedestrians who are vulnerable road users are mixed, it is thought that the behavior of traffic participants varies depending on their personal characteristics, but these personal characteristics are not taken into account in conventional simulations. Therefore, this paper proposes a new method to evaluate the personal characteristics of traffic participants’ decision-making in a pedestrian–vehicle mixed space using probabilistic models, and proposes new simulations that incorporate these personal characteristics. First, we created decision-making models of pedestrians and vehicle drivers using the data collected from the behavior observation experiments in a virtual environment. We then evaluated the aggressiveness of traffic participants’ decision-making using the created models, and showed that the models were consistent with the observed data and valid from the perspective of personal characteristics. Next, we conducted traffic simulations using the decision-making models of traffic participants whose decision-making was aggressive and conservative, and showed that these personal characteristics were indeed reflected in the simulations by examining traffic flow smoothness and danger risk levels. Finally, we conducted the simulations while changing the traffic design in the pedestrian–vehicle mixed areas, and showed that the range of simulation applications can be expanded by including the personal characteristics of traffic participants in the simulation.