Research On The Service Migration And Placement Strategy In Iov Based On User Characteristics

In recent years,the number of cars has skyrocketed,and the corresponding road safety,traffic efficiency,and environmental pollution problems have become increasingly serious.People urgently need a more efficient,smarter,and greener intelligent traffic management system.The Internet of Vehicles is an important carrier of the system,and more and more vehicles will be connected to the Internet of Vehicles to enjoy various vehicular services.These services usually require large computing power and short service delay,but the local computing power of the vehicle cannot meet the computing power requirements,and the cloud server processing service cannot meet the delay requirements.Therefore,people proposed a vehicular edge network,placed services on the roadside units to achieve a good quality of experience.Vehicular services can be divided into two categories according to the number of serviced users: customized services and general services.For customized services,the high mobility of vehicles makes the distance between vehicle users and their dedicated customized servers continue to increase,and service delays continue to increase;for general services,the high mobility of vehicles causes large changes in the load of each roadside unit,and the computing resources of each roadside unit are very limited.We cannot allocate a lot of computing resources to all roadside units to place services.Therefore,it is very important to develop appropriate customized service migration strategies and general service placement strategies.On the other hand,different vehicle users have different characteristic information,such as driving speed,number of calculation tasks issued,etc.Considering the characteristic information of vehicle users will help the reasonable developing of the above-mentioned migration strategy and placement strategy.Therefore,this article focuses on the customized service migration strategy and the general service placement strategy under the predictable setting of vehicle mobility,which considers the difference of user characteristic information.This paper studies the customized service migration strategy under the setting of predictable vehicle mobility,which takes the difference of user characteristic information into account.This paper combines the user characteristics information such as vehicle speed,service type and service demand to derive the total service delay in the case of service migration and non-migration,and constructs the corresponding optimization problem,that is,to minimize the system loss by making appropriate service migration decisions.In order to solve the above problems,this paper derives the set expression of the migration target roadside units determined by the vehicle mobility prediction information,and proposes a service migration decision-making algorithm based on Qlearning.Besides considering the number of wired hops between vehicle users and corresponding dedicated customized servers,the vehicle speed,service demand in the last time slot,and the service type are also defined as status information in this algorithm specially to make a personalized service migration strategy that takes the user characteristic information into account.The simulation result shows that the algorithm can better balance the total service delay and migration cost.At the same time,this paper studies the general service placement strategy under the setting of predictable vehicle mobility that takes the difference of user characteristic information into account.This paper pays attention to the large differences between the service demands of different vehicle users especially.Therefore,combined with the service demands of vehicle users,the service delays using vehicular edge computing and cloud computing are derived respectively,and the corresponding service placement problem is constructed,that is,by making appropriate service placement decisions to maximize system rental utility.In order to solve the above problems,this paper designs an algorithm based on the idea of group learning.The algorithm learns the relationship between user’ characteristic information and user’ service demand to estimate the service demand of each vehicle user,and then realize the load estimation of each roadside unit,in order to flexibly allocate the limited computing resources required by the service placement.The simulation result shows that the algorithm can achieve high system rental utility under the limited service placement resource constraints.