Research On Computation Offloading In Resource-constrained Mobile-edge Computing Systems

With the emergence of augmented reality/virtual reality,online games and other emerging applications with computing-intensive and delay-sensitive needs,the limited computing capacity of mobile devices has been greatly challenged.As one of the core technologies of 5G,Mobile edge computing(MEC)can provide computing,storage,content/wireless sensing near the edge of wireless network.Computation offloading is a key technology in MEC.By offloading tasks to MEC servers,the problem of delay and energy consumption caused by terminals'insufficient computing capacity is greatly solved.However,compared with the increasing computing demand,the limitation of MEC servers' computing resources is becoming obvious.Therefore,how to design a reasonable computation offloading strategy in resource-constrained MEC systems to meet the needs of users is a huge challenge.Firstly,the thesis analyzes the research status of MEC,and summarizes the problem of solving the limited resources of MEC system.It can be considered fr-om two perspectives:(1)introducing auxiliary nodes to expand the computation resources of MEC.(2)Jointly optimizing the computing and wireless resources in network.Secondly,for a single cell,a joint optimization strategy based on task caching is proposed for computing offloading and resource allocation.The innovation lies in the idea of combining cloud and MEC to cache tasks.Because of the geographical popularity of computing tasks in some scenarios,caching the execution results,which are repeated requests firequently can effectively reduce the execution delay of tasks and greatly relieve the computation pressure of MEC servers.The computation offloading,resource allocation and task proactive caching are jointly optimized to minimize the execution latency subject to the constraints of the radio,computation and storage resources.To solve this problem,this thesis divides it into two parts:1.By designing a task proactive caching algorithm combined with MEC and remote cloud to determine the cache status of tasks irn each time slot;2.A heuristic algorithm based on greedy strategy is designed to solve the remaining resource allocation and task execution mode selection problems.The simulation results show that compared with other benchmark schemes,such as local execution scheme and no caching scheme,this scheme can achieve the lowest total delay of task execution.Thirdly,aiming at multi-cell,this thesis proposes a master-slave MEC server cooperative computing offloading strategy based on SDN.The innovation lies in considering the idea of multi-cell computing load balancing based on SDN centralized control airchitecture.The task execution mode selection and computing resource of multiple MEC servers are jointly considered in this thesis,and an optimization problem with computation resource constraints is modeled to minimize the system cost.To solve the Mixed Integer Nonlinear Programming(MINLP)problem with two-dimensional 0-1 task execution decision vectors and continuous computing resource allocation vectors,a heuristic algorithm based on simulated annealing is designed,which achieves the convergence state through continuous iteration and cooling to obtain the sub-optimal solution.The simulation results show that the proposed scheme achieves the best system performance in most cases,compared with the scheme of randomly selecting task execution nodes,the scheme without considering the cooperative MEC server and the local execution scheme.Finally,this thesis summarizes and prospects.