Digital Twin-based Blind Spot Visualization and Real-Time Hazard Notification in AR Glasses: Design and Proof-of-Concept

This paper proposes a digital twin (DT)-based system to enhance pedestrian safety by visualizing blind-spot hazards through AR glasses. The system leverages roadside units (RSUs) with LiDAR and edge servers for real-time object detection, trajectory prediction, and collision prediction. Potential risks are classified into three levels, and corresponding warnings are displayed in AR to prompt timely evasive actions. To achieve accurate AR visualization, the system also integrates DT-enhanced self-localization for centimeter-level accuracy and an end-to-end latency compensation mechanism. A prototype system was developed, and its effectiveness and practicality were validated through proof-of-concept experiments.