Predictions of the Intention and Future Trajectory of the Pedestrian

Abstract

In this chapter, we will focus on pedestrian intention and its trajectory prediction. Pedestrians are important traffic participants, and the dynamic changes of their walking intentions and motion trajectories bring major challenges to intelligent vehicle decision-making and planning. For example, pedestrians may decide to cross the road when they were waiting on the side of the road, appearing in any position of the vehicle. Therefore, research on intention and trajectory prediction of pedestrians is of great significance for reducing the probability of collisions between vehicles and pedestrians, improving vehicle safety, and promoting the development of intelligent vehicle technology. Most of the previous studies only use single human posture features such as skeletons to predict pedestrian intentions. Considering the head orientation feature is one of the important clues of intention projection of the pedestrian, merging it with pedestrian skeleton features is expected to improve the accuracy of pedestrian intention prediction. The introduction of intention features can enhance the learning of pedestrian dynamic changes by the trajectory prediction network and improve the accuracy of pedestrian trajectory prediction.