Intelligent traffic signal scheduling using artificial intelligence algorithms based on real-time traffic flows

Intelligent traffic management has become a critical research focus due to the rapid increase in urban populations and vehicle ownership, which has led to severe congestion and safety concerns worldwide. Traditional fixed-time traffic signal systems fail to adapt to dynamic traffic flows, while actuated-time methods, although more responsive, are still limited in handling real-time complexity. Recent advances in Artificial Intelligence (AI) and deep learning, particularly object detection algorithms such as YOLOv۸, provide new opportunities to enhance traffic signal scheduling through real-time vehicle recognition and flow estimation. In this paper, we propose an intelligent traffic signal scheduling model based on YOLOv۸ for vehicle detection, combined with a Multi-Layer Perceptron (MLP) neural network for computing optimal green-phase durations. The system accounts for real-time vehicle counts, lane occupancy ratios, and signal history, while embedding safety and operational constraints such as minimum/maximum green times, fixed yellow phases, and red-time derivation. Moreover, emergency vehicles are given higher priority weights to minimize their clearance time. Simulation results demonstrate that the proposed approach achieves higher detection accuracy (mAP@۰.۵ of ۵۰.۲%) compared to Faster R-CNN and YOLOv۷ baselines, and reduces average vehicle delay by up to ۲۳% compared to fixed-time control. Emergency vehicle clearance times are also reduced by approximately ۳۵%, confirming the robustness and applicability of the method under varying traffic volumes and environmental conditions.

کلیدواژه ها:

intelligent traffic control ، YOLOv۸ Detection ، Neural Network Scheduling ، Artificial intelligence ، Real-Time Traffic Flows ، Emergency Vehicle Priority ، Adaptive Signal Timing