Research On Dynamic Task Scheduling In UAV-assisted Mobile Edge Computing Systems

The UAV-enabled Mobile Edge Computing(MEC)systems and Vehicular Ad-hoc Network(VANET)-supported applications are very popular topics these days.Internet of Vehicles has still faced great challenges in terms of security and reliability,due to its highly dynamic topology,variable network density,complex urban conditions and big data processing.With the development of applications related to the Internet of Vehicles,the intensity of task computation is gradually increasing.However,it is difficult to meet the requirements of low task delay locally,due to the limited computing capacity of vehicles.On this basis,the offloading problem of vehicle tasks in UAV-enabled MEC systems based on software-defined vehicular network is studied in this thesis.A dynamic greedy heuristic(DGH)algorithm is proposed for the problem under study.Multiple communication and energy consumption models are employed to formulate the problem.The objectives are to minimize the linear weight sum of total time delay and energy consumption.The main research contents of this thesis are as follows:(1)Aiming at the task offloading problem of Internet of Vehicles,a UAV-enabled dynamic task scheduling model is established for complex urban conditions.In this model,one UAV and one edge server are provisioned for the workload from the moving vehicles in the region.From the time that a vehicle enters the region,it would generate computing requests periodically until it leaves the region.Each request is taken as a computation task and can be offloaded locally on the vehicle,the UAV,or the edge server.A variety of communication and energy consumption models are designed to minimize the total time delay and energy consumption during dynamic traffic processes.(2)Aiming at the task offloading problem of Internet of Vehicles,a set of solutions of UAV-enabled dynamic task scheduling model are proposed and designed.Combined with the characteristics of VANET,UAV and edge server,a DGH algorithm is proposed to solve the biobjective optimization problem by taking the smallest linear weight sum of total time delay and energy consumption.In the simulation experiment,the proposed algorithm is compared with the UAV-assisted vehicular computation cost optimization algorithm based on game theory and simulated annealing algorithm under the conditions of dynamic traffic,different road distribution,building distribution and other factors.Experimental results show that the proposed algorithm is more effective than the comparison algorithm in the scenario presented in this thesis,and has better stability and robustness.(3)In view of the above research,a UAV-enabled dynamic task scheduling system is proposed and implemented.The problem scene studied is modeled in 3D to show a three-dimensional dynamic effect.The page displays 3D city traffic map according to the parameters which has been set by users on the front page.The Gantt chart of task execution is obtained through interaction between the front and back ends.