Revealing pedestrian behaviors to support the decision-making of autonomous vehicles

Before the widespread use of autonomous vehicles, it is needed to reveal the travelers' intentions and behavior patterns; with the conclusions gained appropriate response can be provided by the vehicle in a conflict situation. This paper focuses on vehicle-pedestrian conflict points. The aim was to elaborate a method to prevent pedestrians from being hit by an autonomous vehicle. Typical pedestrian movements have been identified by an on-site measurement and a questionnaire survey. Pedestrian crossing patterns were categorized based on the on-site measurement, whereas general habits and feelings behind a crossing were identified by the questionnaire survey. We found that the typical approach angle of a crossing point is 90 degrees. Moreover, the movements approaching a crossing point are strongly influenced by the environment (e.g. location of public transportation stop). Additionally, crossing behavior is strongly influenced by gender. The results support the prediction of pedestrians' position, which can be used in the software development of autonomous vehicles.