AI-DRIVEN ADAPTIVE TRAFFIC MANAGEMENT SYSTEM

Abstract

Traffic congestion has become a significant problem in urban areas as the number of vehicles continues to increase. Conventional traffic signals generally operate using predefined timing schedules and do not respond to the actual traffic conditions at an intersection. This can lead to unnecessary waiting, inefficient use of road infrastructure, increased fuel consumption, and congestion. This paper presents an AI-Driven Adaptive Traffic Management System that adjusts traffic signal timing according to the number of vehicles detected in real time. A camera captures the traffic scene, and the video frames are processed using OpenCV and the YOLOv8 object detection model. Vehicles detected within predefined Regions of Interest (ROIs) are counted to estimate the traffic density in different directions. Based on these counts, an adaptive decision module determines the appropriate green signal duration for each traffic phase. The generated control decisions are transmitted to ESP32 microcontrollers, which operate the physical miniature traffic signals. The proposed prototype combines computer vision, artificial intelligence, and embedded control to demonstrate a low-cost approach to adaptive traffic signal management. The system is intended to improve traffic flow by allocating signal time according to the observed traffic density.