Offloading of Interdependent Computing Subtasks Based on VANET in the Harsh Environment

Abstract

Inter-vehicle communication is crucial in many severe weather or natural disaster situations. However, the wireless link quality is worse than the wired in that the network topology between vehicles changes frequently as vehicles move at high speeds. This also seriously affects the quality of inter-vehicle communication and corresponding services. Emerging computing-intensive service tasks like autonomous driving are extremely time-dependent, which poses large issues for resource allocation between vehicles. In particular, vehicles driving will face greater obstacles when severe weather causes a serious shortage of roadside computing resources. Therefore, we develop appropriate offloading strategies for interdependent computation tasks in the harsh environment. To reduce the overall task completion time within the cost budget limitations, we study a dynamic offloading strategy based on vehicle distribution probability and revenue discount factor, where subtasks are offloaded to the appropriate vehicle by cellular vehicle-to-everything communication. Simultaneously, for small tasks with smaller amounts of calculation, we propose a quasi-static offloading strategy based on branch and bound. Simulation results for varying amounts of calculation, vehicle arrival rates, and average speeds show that the suggested strategies can enhance revenues while getting lower time complexity. Furthermore, our solutions are more adaptable to the extremely dynamic vehicular ad hoc network environment.