Abstract
We propose a method for predicting pedestrian behavior and predicting trajectories with the results of behavior prediction. To enable automated vehicles to move safely and efficiently in urban areas with mixed pedestrian traffic, it is important to understand the behavior of pedestrians whose paths intersect with the vehicle and to predict their future movements. In this study, we first predict the extent to which the pedestrians intend to yield the right-of-way to the vehicle, based on past movements of the pedestrian and the vehicle. Using publicly available datasets, we evaluated the usefulness of the proposed model and found that it improves trajectory prediction performance compared to the case where pedestrian behavior is not considered.
