Create Your First Project
Start adding your projects to your portfolio. Click on "Manage Projects" to get started
AI Traffic Light Management Systems
Project type
Date
This project examined how artificial intelligence could improve traffic-light control by using real-time vehicle detection to adjust signal timing. I focused on the YOLOv4 object-detection system, which uses neural networks and image processing to identify vehicles, pedestrians, bicycles, buses, and trucks while estimating traffic density at an intersection.
I researched and compared three major approaches to traffic control: manual operation, sensor-based systems, and AI-based control. The analysis considered efficiency, safety, installation and maintenance costs, adaptability, system reliability, environmental impact, and performance in dense traffic. AI-based control showed the greatest potential for responding dynamically to changing traffic conditions, reducing unnecessary idle time, and improving vehicle throughput.
The proposed system would calculate green-light duration based on vehicle density, vehicle type, and average passing time. High-traffic lanes could receive additional green-light time, while low-traffic lanes would avoid remaining active unnecessarily. The research also explored the underlying YOLOv4 architecture, including convolutional neural networks, object classification, bounding-box processing, Python, OpenCV, Darknet, and simulated traffic environments.
Published test results reviewed in the project reported detection accuracy of approximately 80% to 85%, depending on vehicle density, along with improved vehicle throughput across multiple lanes. I also evaluated the system’s limitations, including high initial equipment costs, dependence on cameras and computing hardware, privacy concerns, maintenance complexity, and its current focus on four-way intersections in densely populated areas.
I completed the research independently and created a 30-slide technical presentation explaining the problem, existing traffic-control methods, YOLOv4 architecture, simulation methods, published performance results, sustainability benefits, limitations, and opportunities for future development.