Machine learning techniques for road traffic prediction using vehicular communications data

• Main Author: Gonçalves, Francisco dos Santos
• Publication Date: 2023
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10773/41831
• topic: Machine learning; Deep learning; Traffic prediction; Time series; Intelligent Transportation Systems; Smart cities; Vehicular communications

By leveraging advanced data analytics, machine learning algorithms, and real-time monitoring, traffic prediction goes beyond traditional reactive approaches, allowing for a more dynamic and responsive system. One key advantage of traffic prediction in Intelligent Transportation Systems is its ability to forecast congestion, bottlenecks, and traffic incidents before they occur. This foresight enables traffic management authorities to implement proactive measures, such as optimizing signal timings, adjusting speed limits, or rerouting vehicles, to alleviate potential issues and maintain a smooth flow of traffic. In the realm of smart cities, traffic prediction plays a pivotal role in creating more sustainable and eco-friendly urban environments. By providing insights into traffic patterns and trends, cities can optimize public transportation routes, encourage the use of alternative modes of transportation, and reduce overall carbon emissions. This holistic approach to transportation planning contributes to the development of greener and more livable urban spaces. However, predicting road traffic is arduous because it frequently follows complicated nonlinear temporal patterns. As a result, traditional statistical prediction approaches fail to provide reliable/accurate predictions. Deep Learning techniques are employed to solve this issue due to the recent increase in computational power and available historical traffic data. This dissertation describes the implementation of a storage solution for vehicular communications data, that is being collected within the scope of a Cooperative Intelligent Transportation System (C-ITS) via an MQTT broker, and the subsequent usage of said data for the development of a deep learning algorithm for predicting road traffic flow. The system’s behaviour was validated by tests conducted on its various components.