Enhancing Intersection Traffic Safety Utilizing V2I Communications: Design and Evaluation of Machine Learning Based Framework

Abstract

In recent years, improving intersection traffic safety has become a major focus for researchers. However, it remains a significant concern due to the increasing number of vehicles and the introduction of autonomous and cooperative driving systems. Advancements in artificial intelligence (AI) and vehicle-to-everything (V2X) technologies offer promising solutions to reduce collisions between vehicles. As V2X technologies are slowly being integrated into traffic safety systems, questions arise about their impact on safety effectiveness. This research area is relatively unexplored but has potential to enhance intersection safety. In this paper, we introduce the Intersection Traffic Safety Framework (ITSF), a safety-oriented system devised to mitigate collisions at road intersections. This framework achieves this through the implementation of a collision avoidance mechanism that harnesses vehicle-to-infrastructure (V2I) communications. Additionally, it incorporates a machine learning model tasked with distinguishing between vehicles posing a risk and those that do not, as part of the collision avoidance process. Furthermore, this work assesses the performance of this framework by considering critical factors like the penetration rate of V2X-enabled vehicles, end-to-end latency, and the responsiveness of drivers to safety alerts. The proposed approach proves highly effective in ensuring road users’ safety with a penetration rate of 60% or higher, resulting in a significant reduction in collisions and nearly 98% accuracy in classifying risky vehicles. Additionally, the algorithm remains successful even when some drivers neglect warnings, showcasing its robustness in minimizing collision rates.

Keywords

• Intersection traffic safety
• traffic collision avoidance
• V2I communications
• penetration rate
• machine learning