Real-Time Traffic Perception and Safety Alerting in Dilemma Zones Using a Connected IoT System

Abstract: Signalized intersections in the US have witnessed a concerning number of fatalities attributed to red-light running, particularly within what is termed the "dilemma zone." This zone, noted since the 1960s, poses decision-making challenges for drivers during the onset of the yellow light, influencing red-light running incidents. Although significant efforts have been made to address the physical constraints (Type I dilemma zone) of these intersections, the behavioral decisions influenced by individual drivers (Type II dilemma zone) continue to pose significant challenges. Existing methods for identifying and addressing these zones are plagued by issues like inaccurate vehicle detection due to limited sensing capabilities, an over-reliance on post-processing rather than real-time assessments, and a lack of real-time connectivity between drivers and infrastructure. Addressing these challenges, this research introduces an innovative Edge-AI-based dilemma zone sensing and warning system. The proposed system integrates real-time data processing, AI-enhanced vehicle detection, and a communication interface linking both infrastructure and road users. The system employs computer vision technology on edge devices to detect and track vehicles approaching intersections. By analyzing the speed and trajectory, the system judges if vehicles are within a dilemma zone. Upon identification, it disseminates alert notifications to drivers, transportation agencies, and transportation infrastructure such as signal controllers. By effectively bridging the objective intersection standards with the subjective behaviors of drivers, the system offers a promising approach to significantly enhance safety at signalized intersections.