Pedestrian Crossing Intent Prediction via Psychological Features and Transformer Fusion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ashayer, Sima, Nguyen, Hoang H., Liang, Yu, Sartipi, Mina
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910060691587072
author Ashayer, Sima
Nguyen, Hoang H.
Liang, Yu
Sartipi, Mina
author_facet Ashayer, Sima
Nguyen, Hoang H.
Liang, Yu
Sartipi, Mina
contents Pedestrian intention prediction needs to be accurate for autonomous vehicles to navigate safely in urban environments. We present a lightweight, socially informed architecture for pedestrian intention prediction. It fuses four behavioral streams (attention, position, situation, and interaction) using highway encoders, a compact 4-token Transformer, and global self-attention pooling. To quantify uncertainty, we incorporate two complementary heads: a variational bottleneck whose KL divergence captures epistemic uncertainty, and a Mahalanobis distance detector that identifies distributional shift. Together, these components yield calibrated probabilities and actionable risk scores without compromising efficiency. On the PSI 1.0 benchmark, our model outperforms recent vision language models by achieving 0.9 F1, 0.94 AUC-ROC, and 0.78 MCC by using only structured, interpretable features. On the more diverse PSI 2.0 dataset, where, to the best of our knowledge, no prior results exist, we establish a strong initial baseline of 0.78 F1 and 0.79 AUC-ROC. Selective prediction based on Mahalanobis scores increases test accuracy by up to 0.4 percentage points at 80% coverage. Qualitative attention heatmaps further show how the model shifts its cross-stream focus under ambiguity. The proposed approach is modality-agnostic, easy to integrate with vision language pipelines, and suitable for risk-aware intent prediction on resource-constrained platforms.
format Preprint
id arxiv_https___arxiv_org_abs_2603_19533
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Pedestrian Crossing Intent Prediction via Psychological Features and Transformer Fusion
Ashayer, Sima
Nguyen, Hoang H.
Liang, Yu
Sartipi, Mina
Computer Vision and Pattern Recognition
Robotics
Pedestrian intention prediction needs to be accurate for autonomous vehicles to navigate safely in urban environments. We present a lightweight, socially informed architecture for pedestrian intention prediction. It fuses four behavioral streams (attention, position, situation, and interaction) using highway encoders, a compact 4-token Transformer, and global self-attention pooling. To quantify uncertainty, we incorporate two complementary heads: a variational bottleneck whose KL divergence captures epistemic uncertainty, and a Mahalanobis distance detector that identifies distributional shift. Together, these components yield calibrated probabilities and actionable risk scores without compromising efficiency. On the PSI 1.0 benchmark, our model outperforms recent vision language models by achieving 0.9 F1, 0.94 AUC-ROC, and 0.78 MCC by using only structured, interpretable features. On the more diverse PSI 2.0 dataset, where, to the best of our knowledge, no prior results exist, we establish a strong initial baseline of 0.78 F1 and 0.79 AUC-ROC. Selective prediction based on Mahalanobis scores increases test accuracy by up to 0.4 percentage points at 80% coverage. Qualitative attention heatmaps further show how the model shifts its cross-stream focus under ambiguity. The proposed approach is modality-agnostic, easy to integrate with vision language pipelines, and suitable for risk-aware intent prediction on resource-constrained platforms.
title Pedestrian Crossing Intent Prediction via Psychological Features and Transformer Fusion
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2603.19533