Zhou, Junle (Northwestern Polytechnical University), Shen, DAGang (Northwestern Polytechnical University), Ren, Yechen (Northwestern Polytechnical University), Xie, Yaxuan (Northwestern Polytechnical University), ZHANG, Kailong (Northwestern Polytechnical University), Nguyen, Thi Mai Trang (Sorbonne University - LIP6)

Multi-Mode Traffic Flow State Estimation Based Adaptive Signal Phase Decision and Control Mechanism for Service-Oriented C-ITS

Scheduled for presentation during the Regular Session "Traffic signal control" (WeAT9), Wednesday, September 25, 2024, 11:10−11:30, Salon 17

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Traditional traffic signals employ the fixed-time control mechanism, which is incapable of adapting to the dynamic changes in traffic conditions and cannot provide more reasonable scheduling for vehicles of different service urgencies, such as firefighting and public security. This significantly impacts the traffic efficiency and quality at intersections. In response to these issues, this paper studies and proposes an adaptive traffic signal phase decision and control framework based on multi-mode traffic flow state estimation. Initially, we constructed networked vehicle model with service priority attributes, as well as models for traffic objects such as roads and intersections. Subsequently, we researched and proposed a multi-mode, dynamic estimation method for traffic flow situations that integrates vision, communication, and multi-priority. Building upon this foundation, we designed and implemented a signal phase decision and control method that combines logical rules and deep reinforcement learning to effectively handle various intersection scenarios. Finally, we developed a prototype intelligent signal controller system and integrated a semi-physical simulation environment using SUMO, CARLA, and other tools. We comprehensively validated the designed methods based on typical scenarios.