Deep Learning-Based Content Centric Data Dissemination Scheme for Internet of Vehicles

摘要

With the evolution of Internet of Things (IoT), there has been an overwhelming increase in the number of connected devices in recent years. Due to this, generation of massive amounts of data is inevitable from these enormous number of devices in IoT environment, especially in Internet of Vehicles (IoV). In such an environment, there is a need of a paradigm shift from traditional host-centric approach to a more flexible content-centric networking approach. The existing TCP/IP-based congestion control mechanisms can not be directly applied in IoV environment as there is a requirement of content sharing among vehicles with reduced delay and high throughput which most of the existing TCP variants (Tahoe, Reno, NewReno and TCP Vegas) may not be able to provide. So, in this paper, a deep learning- based content centric data dissemination approach for IoV is presented by taking into account the mobility of vehicles and type of content shared among vehicles. The proposed scheme works in three phases: 1) In the first phase, an energy estimation scheme is designed to identify the vehicles which can participate in data dissemination. 2) In the second phase, connection probability of these vehicles is computed to identify stable and reliable connections using Weiner process model. 3) In the last phase, a convolutional neural network (CNN)-based scheme for estimating the social relationship score among vehicle-to-vehicle pairs is designed. CNN is used to identify the ideal vehicle pairs, which can share data to ensure minimum delay and high data availability. The proposed scheme is evaluated on a highway topology using extensive simulations. The results obtained proves the efficacy of the proposed scheme with respect to performance metrics such as-content disseminated, energy, and social score.