Modeling the Behavior of Pedestrians in a Group at Signalized Intersections Using Dynamic Bayesian Method

With the development of autonomous driving and connected vehicles, active safety systems are increasingly being used for traffic assessment, especially for collision avoidance. These systems identify potential collisions between pedestrians and vehicles by analyzing the pedestrian’s walking behavior. This paper mainly focuses on the crossing behavior and path predicition of pedestrians in group at signalized intersections. First, the group behavior and the intersection context information are integrated by using a dynamic Bayesian network approach. Then a crossing decision model and movement model are proposed to predict the group trajectories in the next few seconds. Finally, real traffic data were collected to calibrate the model parameters, and simulations are applied to verify the model reliability. This model is also conducive for use in autonomous vehicles to drive smoothly and to effectively avoid pedestrians.