Adaptive traffic signal timing by using machine learning techniques

Abstract

In the context of increasingly complex urban traffic and severe congestion in major cities, optimizing traffic signal control has become a crucial factor in minimizing waiting times and improving traffic efficiency. This study explores the application of machine learning and deep learning in simulating and optimizing the timing of intelligent traffic signal control. Specifically, the research is conducted in the Pham Van Bach Street area, Cau Giay, Hanoi, where high traffic density and frequent congestion occur. The study employs deep learning methods, particularly Reinforcement Learning, to develop an intelligent traffic signal control model. This model is trained using traffic data collected from surveillance cameras, including vehicle flow and stop frequency at signals, and subsequently utilizes this data to optimize signal timing according to real-time traffic conditions. The research findings indicate that the intelligent traffic signal control system can significantly reduce waiting times and improve vehicle flow compared to traditional methods. This study not only contributes to enhancing traffic efficiency in the Pham Van Bach area but also opens up the potential for the broader application of machine learning and deep learning techniques in intelligent transportation systems in large cities, ultimately helping to mitigate congestion and air pollution