TC-PAA: Deep Learning-Enabled QoS Enhancement Scheme for Cooperative Internet of Vehicles

Abstract

Quality of Service (QoS) plays a pivotal role in numerous delay-sensitive applications that range from general to specific such as the Internet of Medical Things (IoMT), Industrial Internet of Things (IIoT), Unnamed Aerial Vehicles (UAVs), Industrial Automation, and Cooperative Internet of Vehicles (C-IoV), etc. Every application has numerous contributions to human daily life activities, but here in this work, we focused on the C-IoV in the context of QoS metrics. Even though the literature suggested several techniques to address the QoS issues in this emerging technology, but we have not come across a single article that addresses this issue in a cooperative environment, considering the impact of communication congestion and contention by taking into account emergency vehicles and traditional vehicles. Given that, in this paper, we introduce a hybrid framework known as the Traffic Congestion and Priority-Aware Algorithm (TCPAA). This innovative paradigm leverages the capabilities of computer vision, Deep Neural Networks (DNN) and Dijkstra algorithm to strategically incorporate the transmission channels and network entities with an objective to improve the QoS metrics in emergency vehicles. Initially, we developed a dataset with computer vision algoritms “real-time (OpenCV ”Background Subtraction”) to evaluate and chose the best machine learning algorithms among random forest, support vector machine (SVM), k-means clustering, and DNN. Based on the result statistics, we select DNN, and classified vehicles into two classes: Emergency and traditional vehicles to train the model. Subsequently, we set standard for two type of communications such as regular and prioritized traffic. We incorporate a micro base station (μBS) in the network for prioritized traffic to facilitate congestion-free communication of emergency vehicles, while the Dijkstra algorithm is used to managed the communication of traditional vehicles. Considering the nature of operation of future autonomous vehicles, we managed most of the decisions processes at the client-side by categorizing the traffic based on the vehicle requirements. Through reliable client-side management, the high performance and accuracy of TC-PAA underscore its efficiency compared to established field-proven schemes. Adhering to reliability metrics such as latency, packet loss ratio, communication cost, data availability, and traffic priority, the proposed model improves QoS metrics in high-demanding areas of IoV networks.