Research On Computing Task Offloading Algorithms Between Vehicles In Vehicular Communication

With the advancement of information technology,delay-sensitive applications such as road environment augmented reality,human-vehicle dynamic interaction,and driving safety warnings are emerging in an endless stream,which poses a huge challenge to the low-latency and high-reliability data forwarding requirements of in-vehicle communication.In addition,traditional cloud computing can no longer meet the processing requirements of millisecond-level latency,and the application of mobile edge computing technology to the Internet of Vehicles environment is the development trend and direction.However,the number of edge servers deployed is limited,and some vehicles are also equipped with high-performance computing units but are not fully utilized.In this case,how to design a vehicle-to-vehicle computing task offloading algorithm suitable for the vehicle-connected edge network to make full use of high-performance vehicle idle resources,make up for the insufficient deployment of edge servers,and reduce vehicle local computing delays has become an urgent problem.This paper focuses on the vehicle-to-vehicle edge network environment,and focuses on the offloading of computing tasks between vehicles in the vehicle-oriented self-organizing network.The main research work of this thesis is as follows:(1)A static vehicle computing task offloading based on a multi-arm gambling machine.The scheme divides the driving vehicles into task vehicles and service vehicles.The task vehicle refers to the computing task initiator and the service vehicle refers to the computing task receiver;then the entropy method is used to establish a priority model to make it generate at the same time The computing tasks are arranged in an orderly manner;then the available computing and communication resources of the service vehicle are analyzed,and the computing task offloading problem is modeled by the multi-armed gambling machine problem,and the offloading strategy is solved by the adaptive UCB algorithm.The results show that compared with the original UCB(Upper Confidence Bound)and-greedy algorithm,this scheme reduces the average calculation delay by about 12%,reduces by about 22% compared with random offloading,and reduces the cumulative timeout probability by about 10%.(2)A dynamic vehicle computing task offloading based on vehicle clustering.This scheme first generates a list of neighbor nodes of each vehicle based on the angle of movement between vehicles;secondly,it describes the sample difference by the Mahalanobis distance between the vehicles,and defines the distance between the reachable distance magnified scatter points and the cluster center;Then use the Mahalanobis distance as the edge and the vehicle samples as the vertices to build an undirected graph,and then find the minimum spanning tree in the graph and convert it into a hierarchical clustering structure.At the same time,the clustering tree is compressed to remove scattered points;then automatically according to the cluster stability Extract clusters,and select cluster heads for cluster management according to vehicle mobility related parameters;finally,based on the clustering results,combined with the multi-arm gambling machine problem model of research content 1,add the cross-cluster unloading tolerance parameter to the UCB algorithm.Computing task offloading strategy is solved.The results show that the scheme has been optimized in terms of the average calculation of unloading delay and the return rate of calculation results compared with the comparison scheme,and the survival time of cluster clusters and cluster heads of the scheme has been prolonged by 10%～20%.