Research On Random Task Offloading Strategy In Mobile Edge Computin

The growing demand for computing resources in the Internet of Things(Io T)has led to the development of mobile edge computing(MEC)technology.However,the limitations of edge servers in base stations have made MEC assisted by unmanned aerial vehicles(UAVs)an increasingly popular solution.Nevertheless,many tasks that arrive in MEC networks are stochastic and continuous.Therefore,this thesis focuses on the stochastic task offloading problem in UAV-assisted MEC networks,and two aspects are investigated:(1)Considering the scenario where the base station cannot provide assistance to MEC due to communication interruption or ground facilities damage,a two-layer UAV-assisted MEC model is proposed in this thesis.To address the stochastic and continuous characteristics of task arrivals,a queueing model is constructed,and a system-wide energy optimization problem is proposed to ensure long-term performance.To solve this problem,the Lyapunov stochastic optimization theory is adopted to transform the problem into a deterministic one.The approximate optimal deployment of UAVs is determined using the differential evolution algorithm,and resource allocation is optimized using the deep deterministic policy gradient algorithm.Numerical results show that the proposed scheme has a significant effect on reducing the total energy consumption of the system.(2)With the surge in the number of mobile intelligent terminals,the limited computing and communication resources of the base station have led to communication overload and task backlog.To address this issue,this thesis proposes a multi-UAV-assisted MEC model.Building upon the first part of our work,this thesis considers the imbalance between the complex task data volume and computation,and propose a dual-queue model.To ensure long-term performance and user experience,this thesis formulates a weighted energy and latency stochastic optimization problem,which is processed using Lyapunov stochastic optimization theory.Then clustering algorithm is used to determine the initial deployment of UAVs and optimize the UAV trajectory,task offloading strategy,and resource allocation using a proximal policy optimization algorithm to solve the optimization problem.Numerical results show that the proposed approach can effectively reduce the total system energy consumption and average latency.