Innovative Real-Time Adaptive Traffic Signal System Enhances Safety for Vulnerable Road Users

Current pedestrian crossing signals are fixed and do not adapt to the behavior of vulnerable road users (VRUs), including the elderly and disabled, potentially leaving them stranded. The No Pedestrian Left Behind (NPLB) system has been developed to monitor VRUs in crosswalks and automatically extend signal timing when necessary. Evaluations of five leading object detection models on the BGVP dataset revealed that YOLOv12 achieved the highest mean Average Precision at 50% (mAP@0.5) of 0.756. NPLB combines the fine-tuned YOLOv12 with ByteTrack multi-object tracking and an adaptive controller, improving VRU safety by 71.4% through 10,000 Monte Carlo simulations, while reducing stranding rates from 9.10% to 2.60% with signal extensions required in only 12.1% of crossing cycles.