Presentation
When Do They Cross? Temporal Dynamics of Pedestrian Intention Prediction and Crossing Actions
DescriptionModern autonomous vehicles (AVs) that imitate human drivers in predicting pedestrian intentions have the potential to improve safety and reduce accidents. However, accurately predicting pedestrian intentions remains challenging due to behavioral variability, situational uncertainty, and subtle motion cues. The temporal relationship between predicted intentions and actual pedestrian actions is still not well understood. This study investigates whether aligning the timing of intention predictions with future actions can enhance AV decision-making by providing earlier cues for proactive responses. Using time-series analysis of the PSI dataset, we examine the temporal lag between intention labels and pedestrian actions across four behavioral classes. Results indicate that intention predictions may precede or lag behind actions depending on behavior type. Notably, crossing and non-crossing pedestrians exhibit distinct lag patterns, and observable behaviors such as looking significantly enhance early intention recognition.
Event Type
Lecture
TimeTuesday, October 14th4:30pm - 5pm CDT
LocationGrand Hall J
Surface Transportation
Similar Presentations