Scholarlink

An efficient and intelligent traffic control system is essential for reducing unnecessary congestion and fuel wastage. This study presents a smart traffic control system that dynamically adjusts green light duration according to real-time traffic density. Three object detection models - YOLOv5, Faster R-CNN, and SSD - were implemented and evaluated on a traffic image dataset from Kaggle. YOLOv5 showed superior performance with the lowest inference times (72.59s for training, 3.21s for testing, and 6.80s for validation), outperforming Faster RCNN and SSD. Due to its speed, YOLOv5 was selected for further processing. The vehicle count detected by this approach was transmitted to an Arduino-based system simulated in SimulIDE, through serial communication. This approach focuses on reducing the wait time and enhancing traffic flow efficiency.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive Traffic Control Using Deep Learning and Arduino Simulation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Riya Kumari,
  • Sneha Suman,
  • Jasdeep Kaur Dhanoa

摘要

An efficient and intelligent traffic control system is essential for reducing unnecessary congestion and fuel wastage. This study presents a smart traffic control system that dynamically adjusts green light duration according to real-time traffic density. Three object detection models - YOLOv5, Faster R-CNN, and SSD - were implemented and evaluated on a traffic image dataset from Kaggle. YOLOv5 showed superior performance with the lowest inference times (72.59s for training, 3.21s for testing, and 6.80s for validation), outperforming Faster RCNN and SSD. Due to its speed, YOLOv5 was selected for further processing. The vehicle count detected by this approach was transmitted to an Arduino-based system simulated in SimulIDE, through serial communication. This approach focuses on reducing the wait time and enhancing traffic flow efficiency.