Urban Traffic Congestion Reduction Through Dynamic Scheduling

Traffic congestion often leads to an ineffective use of time, air pollution, noise pollution, human health issues, and human frustration. Also, the delay of emergency vehicles like ambulances and fire trucks, which often get stuck in traffic due to congestion caused at the intersections, impacting their response to reach their critical destination on time. The study proposes a real- time dynamic traffic scheduling system leveraging deep learning techniques to process traffic images. By continuously monitoring traffic conditions through live videos captured from roadside cameras, the system can accurately determine the number of vehicles and their distribution across different lanes, divided into different priority zones based on their proximity to the junction. Based on this analysis, the traffic signal timings are dynamically adjusted to optimize flow and reduce congestion. By incorporating emergency detection, the scheduling system will be able to prioritize their passage, reducing response times. This is achieved by integrating multiple junctions to form a connected network to alert surrounding junctions of incoming emergency vehicles upon detection and make the necessary preemption upon arrival. Integrating multiple junctions enables a green corridor, coordinating consecutive junctions for uninterrupted emergency vehicle passage. The system will be able to detect anomalies and handle them appropriately. YOLOv8n was evaluated to have a 86% average precision (AP). The proposed scheduling algorithm was simulated and gave a result with 14.7% less waiting time for vehicles than the default pre-timed mechanism.