Adaptive Traffic Signal Management System Using Machine Learning

Abstract

Traffic congestion is a very common and a big issue nowadays in different countries around the world. The intelligent/smart traffic signaling system is crucial for an effective flow of traffic and has got much attention from researchers. Over the last few years, some intelligent systems have been developed using different techniques. In Pakistan, the current traffic signaling is still handcrafted. Traffic congestion is also a cumbersome task for traffic wardens. They set the time for each signal manually. In this regard, this project aims at developing an adaptive traffic signaling system for urban areas that will dynamically adjust signal timings according to the density of traffic. Our intelligent system will take videos of each side of intersection and extract frames. It will generate vehicle on simulation that will provide the data further to a deep reinforcement learning model which assigns the time to a signal, depending upon the ratio of vehicles and their waiting time on each side of the intersections dynamically. In this way, traffic congestion at the intersection will be reduced in an effective and efficient manner.