Application-Aware Offloading Policy Using SMDP in Vehicular Fog Computing Systems

The emergence of various applications in vehicular networks has greatly improved the traffic safety and transportation efficiency while providing a comfortable driving experience. Meanwhile, many of these applications, such as augmented reality (AR) techniques, etc., are time sensitive and require intensive computation loads, which pose new challenges in terms of computation and processing. Fog computing has emerged as a promising architecture to provide a distributed infrastructure brings the efficient computing capability of the cloud to the edge of the network thus providing higher performance. However, considering different delay requirements required for the applications and the variable computation resources in vehicular networks, it is difficult to achieve an efficient offloading. To solve this problem, in this paper, we propose the vehicular fog computing system (VFCS) to utilize the public service vehicles, such as buses, as the fog servers. To describe the variability feature of the available resources and the application-aware delay requirements, a priority queuing system is applied to model the VFCS. By the semi-Markov decision process (SMDP), we propose an application-aware offloading policy to obtain the optimal computation resource allocation scheme and maximize the long-term expected reward of the VFCS, where both the heterogeneous delay requirements of the applications and the dynamic topology of the vehicular networks are considered. Experimental results show that the proposed SMDP- based offloading policy can improve the offloading performance when compared to the conventional baseline policies.