Cloud-based Computation Offloading In Vehicular Networks

This paper studies computation offloading in cloudbased vehicular networks. Cloud-based vehicular networks are a promising paradigm to improve vehicular services through vehicle-to-infrastructure (V2I) communication mode, when mobile users in vehicles that are not equipped with powerful computing devices, they can offload time-sensitive and computationintensive tasks to remote cloud center or edge servers. However, high mobility of the vehicles may cause degradation of the offloading efficiency and unstable communication links between vehicles and edge servers. To solve this problem and reduce the transmission cost of the computation offloading, we propose a cloud-based offloading scheme in vehicular networks, which consider the mobility of the vehicles and the cooperation of the edge servers, the objective is to effectively utilize the computation resources of the edge servers, in this scheme, vehicles can offload tasks to servers out of its own communication range through predictive relay transmissions. Due to high mobility of vehicles, a dynamic programming-based predictive search algorithm (DPA) is proposed to find a set of the edge servers to offload the tasks, and this algorithm minimize the total energy consumption while ensuring that communication link is alive during the computation offloading process. Illustrative results are shown the performance of our proposed scheme.