Research On Task Offloading And Resource Allocation In Vehicular Edge Computing System

With the rapid development of Vehicle-to-Everything(V2X)communication technology,many new and computationally intensive vehicle applications have emerged.Vehicular Edge Computing(VEC)can make full use of the computing resources of edge nodes by offloading computing tasks generated by intelligent networked vehicles to Road Side Units(RSU)or nearby vehicles that are idle and willing to share resources,effectively reducing the transmission pressure on the public network and reducing the task processing delay.However,the highly dynamic vehicle environment,how to motivate vehicles to share idle resources,and the uncertainty of vehicle shared resources have brought many challenges to the design of task offloading and resource allocation methods for VEC system.The main research contents and innovations are summarized as follows:1)In Vehicle-Vehicle(V-V)offloading mode,the rapidly changing network topology,the uncertainty of the wireless channel state and the amount of computing resources allocated to the task all bring great challenges to the task offloading between vehicles.Therefore,an online learning task offloading(SWOLTO)algorithm with sliding window is designed based on the theory of Multi-Armed Bandit(MAB).SW-OLTO works in a distributed manner,learning the latency performance of each shared-resource vehicle while offloading tasks,without requiring frequent exchange of state information between vehicles.In order to better adapt to the dynamic environment,the algorithm is conscious of occurrence and tracks the changes of the optimal shared-resource vehicle through the sliding window.The learning regret upper bound of SW-OLTO is theoretically analyzed,and conducts experiments in the python synthetic scene and the real highway scene based on Veins.The results show that compared with other existing upper confidence bound based learning algorithm,the SW-OLTO algorithm proposed in this paper can improve the latency performance of task processing.2)In the hybrid Vehicle-RSU(V-R)offloading and Vehicle-RSU-Vehicle(V-R-V)offloading modes,computing resource leasing is considered to motivate vehicles to share idle resources.In order to maximize the interests of the offloading-task vehicle,RSU and shared-resource vehicle,an adaptive type selection(ALTS)algorithm is designed for the shared-resource vehicle based on MAB theory,and then the interaction of the three parties is modeled as a multi-stage Stackelberg game,a computing resource lease contract was developed.Experimental results demonstrate that the proposed ALTS algorithm outperforms other existing learning algorithms,illustrate the effectiveness of lease contracts and tripartite transaction mechanisms,indicate the impact of energy costs on energy consumption and offloading decisions,and the comparison shows that the service provided by integrating RSU and vehicle idle resources is better than pure edge servers and pure shared-resource vehicles.