Design And Simulation Of Multi-constraint-based Cooperative Offload Strategy For Computing Tasks In VEC

Based on the development of Internet of Vehicle industry and communication technology,there has emerged a large number of intelligent application related to the artificial intelligence at present,such as automatic driving,voice interaction,traffic prediction,greatly improving people's driving experience,optimizing the road driving conditions,but the characteristics of such applications usually require a lot of computing resources and storage resources.The traditional Cloud Computing paradigm can provide computing power expansion for mobile devices through the network.However,due to the long distance between vehicles and cloud computing centers and high communication latency,the delay requirements of computing tasks are often unable to be satisfied.Hence there appears the Vehicular Edge Computing(VEC)system,in which a vehicle offloads computing tasks to the nearby roadside unit or to the cloud computing center for execution.On the one hand,the rich computing resources of the roadside unit and cloud computing center can effectively reduce the computing delay.On the other hand,the close distance between the vehicle and roadside unit can reduce the communication delay.Therefore,The VEC system can make up for the shortcomings of cloud computing and effectively improve the overall performance of computing tasks.In the VEC system,the vehicle needs to offload the computing task to the execution unit through communication link,and then the execution unit returns the computing result to the vehicle.Therefore,the offloading strategy is to select the appropriate execution unit and communication mode,in which vehicle characteristics,communication network and computing task characteristics all affect the offloading strategy.So this article puts forward a two-stage computing tasks collaborative strategy based on machine learning which is adaptive to VEC system.This strategy will perform the decision making problems related to execution unit and communication mode summarized as three kinds of offloading ways,then respectively train a classification model and a regression model for three kinds of offloading ways.Then the influencing factors are taken as input into the offloading decision,the final offloading way is obtained by combining the classification and regression prediction results of the two stages.This strategy considers many factors such as network condition,task condition,vehicle mobility,system load and so on,which can get the optimal offloading way quickly and effectively.In this paper,Edge Cloud Sim is used to simulate the VEC system and the real road environment,which is the basis of offloading process simulation.And the proposed cooperative offloading strategy of two-stage computing tasks based on machine learning was tested and evaluated.Compared with other four existing offloading strategies,it can effectively reduce the overall computing task failure rate and improve the experience quality of computing task.