SeDaTiVe: SDN-Enabled Deep Learning Architecture for Network Traffic Control in Vehicular Cyber-Physical Systems

Abstract

The rapid growth in the transportation sector has led to the emergence of smart vehicles that are equipped with ICT. These modern smart vehicles are connected to the Internet to access various services such as road condition information, infotainment, and energy management. This kind of scenario can be viewed as a vehicular cyber-physical system (VCPS) where the vehicles are at the physical layer and services are at the cyber layer. However, network traffic management is the biggest issue in the modern VCPS scenario as the mismanagement of network resources can degrade the quality of service (QoS) for end users. To deal with this issue, we propose a software defined networking (SDN)-enabled approach, named SeDaTiVe, which uses deep learning architecture to control the incoming traffic in the network in the VCPS environment. The advantage of using deep learning in network traffic control is that it learns the hidden patterns in data packets and creates an optimal route based on the learned features. Moreover, a virtual-controller-based scheme for flow management using SDN in VCPS is designed for effective resource utilization. The simulation scenario comprising 1000 vehicles seeking various services in the network is considered to generate the dataset using SUMO. The data obtained from the simulation study is evaluated using NS-2, and proves that the proposed scheme effectively handles real-time incoming requests in VCPS. The results also depict the improvement in performance on various evaluation metrics like delay, throughput, packet delivery ratio, and network load by using the proposed scheme over the traditional SDN and TCP/IP protocol suite.