DATA-DRIVEN UNDERSTANDING OF AI-BASED TRAFFIC SIGNAL CONTROL AND ITS IMPLICATIONS ON LOW-CARBON URBAN TRANSPORT SYSTEMS

Abstract

Emission by vehicles is one of the main sources causing the exceedance of the Air Pollution Index at roadside. Therefore, reducing emissions from vehicles in busy corridors has been a priority of the Government in improving roadside air quality. The existing pedestrian lights, which follow a predetermined signal timing for road crossings, are typical static traffic lights. The pushbutton has been added to improve the pedestrian crossing efficiency by rendering the green phase to the pedestrian as soon as possible. However, the current detection devices are not capable of streamlining the traffic flows and prioritizing pedestrians on the curb according to the actual needs. With the use of artificial intelligence (AI) techniques such as deep learning, this project can analyze the traffic conditions of a designated area, including the length of traffic queues and the number of pedestrians waiting at a crossroad. Our system can then manage traffic lights to facilitate efficient traffic flow between vehicles and pedestrians. In addition, the IoT technologies enable the system to collect traffic information for big data analysis and facilitate efficient traffic management. In the implementation stage of the project, a local field study of a crossroad with a school, bus stops, and shopping malls in the vicinity was conducted and relevant data for further analysis was collected. The results showed that the system could help to shorten traffic queues and pedestrian waiting time at the crossroad. The goals are to examine strategies in traffic signal control based on AI to reduce carbon footprint. The deliverables will assess how urban transport systems can reduce carbon emissions and adapt to climate change with a focus on traffic signal rationalization. It provides guidance, policy recommendations, and outlines research and information needs.