ML & DL for intelligent transportation management with related architecture based-on AWS cloud

• Main Authors: Page, Shridhar, Bhore, Nitin
• Format: Conference Proceeding
• Language: English
• Subjects: Architecture; Cloud computing; Deep learning; Industrial applications; Industrial development; Industry 4.0; Intelligent vehicles; Machine learning; Programming languages; Project management; Public domain; Python; Roads; Safety management; Traffic accidents; Traffic flow; Traffic management; Traffic safety; Transportation management; Transportation models; Transportation systems
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0238744

The dynamics of transportation in urban areas are experiencing a paradigm shift due to unprecedented growth in vehicles putting unmanageable burden on traffic authorities. Increasing number of traffic problems are badly affecting road safety, resulting in more road accidents cases. Getting emergency help becomes difficult though the situation is improving through traditional automation in traffic management. Thanks to ‘Digital India’ initiatives and especially projects like Smart City including Smart Transportation Systems across Bharat (India), in line with what has already happened in other countries like USA, UK, Germany, etc. Increasing number of smart vehicles are running on road now-a-days. Gradually, Bharat is moving ahead with Industry 4.0 ecosystems though still it is in its infancy stage. Therefore, there was a need felt to develop ML (Machine Learning) including DL (Deep Learning) models to check efficacy of these techniques in better traffic-flow prediction and smart transportation management. This research paper carries out comprehensive experiments of developing such ML and DL models and then comparing their results on various parameters of prediction quality. The main software-programming language used here is python with ML & DL related packages. However, availability of required data in public domain from Indian context was the limitation. Therefore, we have used publicly availably traffic-flow data about Northern Virginia/Washington DC capital region, for this academic experimentation purposes and for proof-of-concept (PoC). Towards the end of this paper, we propose AWS Cloud-based architecture for Real Time Traffic Flow Prediction, which we are working on. Through meticulous analysis of these datasets, the study uncovers critical challenges and biases inherent in real-world traffic scenarios. The insights presented through this study aim to empower other researchers, practitioners, and policymakers to navigate the intricate landscape of traffic flow prediction, contributing to further development of modern smart transportation systems. Also, AWS cloud-based architecture (as mentioned above), gives motivation for further research.