Fog Connectivity Clustering and MDP Modeling for Software-defined Vehicular Networks

Abstract

Intelligent and networked vehicles cooperate to create a mobile Cloud through vehicular Fog computing (VFC). Such clouds rely heavily on the underlying vehicular networks, so estimating communication resilience allows to address the problems caused by intermittent vehicle connectivity for data transfers. Individually estimating the communication stability of vehicles, nevertheless, undergoes incorrect predictions due to their particular mobility patterns. Therefore, we provide a region-oriented fog management model based on the connectivity through vehicular heterogeneous network environment via V2X and C-V2X. A fog management strategy dynamically monitors nearby vehicles to determine distinct regions in urban centres. The model enables a software-defined vehicular network (\Gls{SDVN}) controller to coordinate data flows. The vehicular connectivity described by our model assesses the potential for vehicle communication and conducts dynamic vehicle clustering. From the stochasticity of the environment, our model is based on Markov Decision Process (MDP), tracking the status of vehicle clusters and their potential for provisioning services. The model for vehicular clustering is supported by 5G and DSRC heterogeneous networks. Simulated analyses have shown the capability of our proposed model to estimate cluster reliability in real-time urban scenarios and support effective vehicular fog management.