Analysis and Modeling of Traffic Participants Considering Interactions at Intersections Without Traffic Signals

To realize safe autonomous driving, verification of performance is one of the most important steps in the design process. Verification of autonomous driving using real vehicles, however, has many difficulties in terms of cost and safety. Therefore, it is important to exploit the verification based on simulation. Effective verification using simulation requires precise behavior models that can represent ambiguous human decisions and motions with a wide diversity. From these perspectives, this paper presents how to model the drivers' and pedestrians' decision-making and motions, particularly in the interactive situation. The targeted task is passing the road without traffic signals, where traffic participants have to check other vehicles for their decision-making and motion. To observe their interactive behaviors, experiments are executed by using an interactive simulator which consists of driving simulators and virtual reality (VR) devices for pedestrians. From the observed interactive data, drivers' and pedestrians' behavior models are constructed by using a logistic regression model and linear motion model. Since the human behavior varies from person to person, this paper discusses the relation between the obtained model parameter and individual characteristics. In addition, these behaviors are classified and evaluated.