Federated Vehicular Tailgating Detection Utilizing Distance and Speed Data Analytics for Intelligent Transport System

Connected and Automated Vehicles (CAVs) are designed with sensors, cameras, and other computing devices to collect data to support cooperative Intelligent Transportation Systems (ITS). Many potential applications offered by CAVs exist for researchers and road safety stakeholders. Vehicle tailgating is one such application to analyze risky driving behavior. Reliable detection of such behavior is hard and expensive. This paper focuses on addressing the tailgating issue in CAVs using the dataset generated using Simulation Of Urban Mobility (SUMO) for Indian roads. The proposed model of this study uses the Relative speed and Euclidean distance of the preceding and rear vehicle as parameters on the distributed Federated Learning (FL) framework to incorporate security for sensitive vehicular data. We verified our prediction model in the TensorFlow Federated framework, and our results show an accuracy of 96 percent.