Intelligent Traffic Management System Using Real-Time Vehicle Detection and Dynamic Signal Control with Client-Server Architecture

This paper addresses one of the major problems that are present in modern society these days is traffic congestion, because it destroys productivity and creates frustration in the large crowds of commuters. Considering this prevalent issue in general populations, our project aims to use the new science developments to produce an effective solution for traffic management. Thus, we designed a smart traffic light system with respect to using multi-faceted approaches toward this critical problem. We began with implementing on the Raspberry Pi with YOLOv3. This was able to give decent results, but it had to be improved for better performance and scalability. So, we upgraded the architecture of the system to Jetson Nano, which grants more processing power and efficiency. Thus, we will have Jetson Nano acting as a client device, capturing the real-time video frames of traffic conditions in real time. These frames are then sent to a much larger server (PC) where they may be analyzed. That server will take in this data, apply advanced decision-making algorithms, and then provide back to the Jetson Nano the changes it needs to make in response in real-time. Our solution holds queuing of conventional traffic timers and balances the wait time among all the road users involved. Our smart traffic light system with machine learning applied to real-time data analysis presents a scalable and efficient framework to improve traffic flow, thus by consequence the general commuting experience. This paper discusses the methodologies applied, the results obtained, and the implications of our findings for future solutions in traffic management.