VPLight: A Reinforcement Learning Approach for Traffic Signal Control with Pedestrian Dynamics

Abstract

This paper proposes VPLight, a novel reinforcement learning approach for traffic signal control that explicitly incorporates pedestrian dynamics. By considering the interactions between vehicles and pedestrians at intersections, VPLight achieves more effective and safe traffic signal control policies.

Key Contributions

• Proposed VPLight, a reinforcement learning framework for traffic signal control with pedestrian dynamics
• Developed a comprehensive state representation that captures both vehicle and pedestrian information
• Demonstrated superior performance in balancing vehicle and pedestrian effectiveness
• Validated effectiveness through extensive experiments on real-world traffic scenarios

Publication Details

• Journal: IEEE Transactions on Knowledge and Data Engineering (TKDE)
• Ranking: CCF-A, JCR Q1
• Year: 2025
• DOI: 10.1109/TKDE.2025.3641213

Authors

Xinyu Zhang, Zuohan Wu, Chen Jason Zhang, Libin Zheng, Peng Cheng, Jian Yin, Cyrus Shahabi

BibTeX

@article{zhang2025vplight,
  author       = {Xinyu Zhang and
                  Zuohan Wu and
                  Chen Jason Zhang and
                  Libin Zheng and
                  Peng Cheng and
                  Jian Yin and
                  Cyrus Shahabi},
  title        = {VPLight: A Reinforcement Learning Approach for Traffic Signal Control with Pedestrian Dynamics},
  journal      = {IEEE Transactions on Knowledge and Data Engineering},
  year         = {2025},
  doi          = {10.1109/TKDE.2025.3641213}
}