Research On Time-Dependent Computing-Intensive Task Offloading In Internet Of Vehicles

With the construction of 5th Generation Mobile Networks(5G)infrastructure and popularization of electric vehicles,the Internet of Vehicles(IoV)has obtained huge development opportunities.The development of big data and artificial intelligence has brought many applications.The Vehicle Cloud Computing(VCC)architecture based on the communication technology C-V2X(Cellular Vehicle to Everything)of the Internet of Vehicles has been proposed to solve the complex applications to improve the Quality of Service(QoS)of users.However,high mobility of vehicles causes rapid changes of network topology,which poses a huge challenge to the resource allocation of time-dependent subtasks.Therefore,based on the C-V2X Internet of Vehicles communication technology,this thesis makes reasonable resource allocation strategies for time-dependent tasks in single-user and multi-user Internet of Vehicles scenarios.On the one hand,for linked list structured task with one-to-one time dependence between subtasks in single-user scenario,this thesis optimizes the task completion delay under the constraints of the price budget.First,a static offloading strategy of linked list structure tasks based on branch and bound algorithm is proposed for the quasi-static scenario of small task offloading.The simulation shows that this strategy can obtain better benefits under different task lengths,vehicle scales and price budgets.Then,a dynamic offloading strategy of linked list structure tasks based on revenue discount is proposed for the highly dynamic scenario of large task offloading.The simulation shows that this strategy can obtain better benefits under different task lengths,vehicle reach rates and average vehicle speeds.The program running time increases linearly with the task length.On the other hand,for tasks represented by Directed Acyclic Graph(DAG)with one-to-one,one-to-many,and many-to-one time dependence between subtasks in multi-user scenario,this thesis optimizes the user’s completion time.In view of the different requirements of program running time and task completion time,firstly,we propose a multi-user DAG task offloading strategy based on the critical path to optimize the delay of the critical path with lower time complexity;Then,we propose a multi-user DAG task offloading strategy based on simulated annealing to search for the global optimal solution to optimize the user’s fastest completion delay.The simulation shows that the two strategies have better benefits under different vehicle scales,number of users and VM(Virtual Machine)computing resources.The delay of the latter are lower than the former,but the latter’s program running time is longer,so we need to adopt different solutions for different needs.