Liu, Dong (Nankai University), Lan, Enfan (Nankai University), Song, Peili (Nankai University), Yang, Yifan (Nankai University), Liu, Jingtai (Nankai University)

Deciphering Aggressive and Courteous Vehicle Behaviors: Acceleration-Enhanced Cross-Modal Transformer for Pedestrian Intention Prediction

Scheduled for presentation during the Video Session "On-Demand Video Presentations" (VP-VP), Saturday, November 22, 2025, 08:00−18:00, On-Demand Platform

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on April 2, 2026

Abstract

Pedestrian intention prediction is critical for the safe deployment of autonomous driving systems in urban traffic environments. Recent studies have employed attention mechanisms and recurrent neural networks to fuse pedestrian trajectories with ego-vehicle velocity, aiming to capture the interactions between pedestrians and the vehicle. However, these approaches struggle to fully characterize the aggressive and courteous vehicle behaviors that can significantly influence pedestrian intention. To address this limitation, a Cross-Modal Transformer module is employed, which incorporates ego-vehicle acceleration alongside pedestrian trajectories and ego-vehicle velocity, enabling fine-grained modeling of the complex, dynamic interactions between pedestrians and vehicles. In addition, to more effectively exploit multi modal feature information, a Mixture-of-Experts module is employed to dynamically assign feature weights. Extensive experiments on the PIE and JAAD datasets demonstrate that the proposed model achieves state-of-the-art performance, with AUC scores of 0.94 on PIE, 0.88 on JAAD_all, and 0.72 on JAAD_beh.