Research On Computing Offloading Of Mobile Edge Computing System Based On Computation Rate Maximization

Due to the rapid development of the Internet and artificial intelligence,different low-latency and high-computation applications(such as virtual reality and remote surgery)appear in wireless devices(such as smart phones,wearable devices and iot sensors/massagers).However,the limitation of computing capacity of mobile terminal devices cannot meet the high delay requirements of such applications.In order to meet the delay requirements of such devices and the large amount of computing data,mobile edge computing(MEC)[^1]-[4]has become an effective solution.MEC offloads computing tasks to nearby network edge devices or base stations through wireless channels,providing end users with cloud-like computing capabilities at network edge(for example,access point(APs)and base station(BSs)),providing a better solution for some delay-sensitive and computation-intensive mobile applications.Aiming at the problem of computing offloading optimization in MEC system,this thesis studies the maximum number of users'total computing bits under the condition of time delay and energy consumption.The main work and achievements of this thesis are as shown below:(1)Two models of offline and online multi-user computing bit maximization in mobile edge computing system are proposed,the problem is modeled as a problem of the maximum total number of computing bits of optimized computing offloading and local computing,and the closed solution of the optimal resource allocation scheme of the problem is derived by using the convex optimization method.(2)Under the offline scheme,an iterative algorithm based on ellipsoidal method is proposed to solve the problem of total computing bit maximization of users under the established system model,and the secondary gradient of the ellipsoidal algorithm is updated so that the optimization variables in the problem tend to be optimal in the process of iterative update step by step,and finally the total computation bits of the user are maximized under the constraints of delay and energy.Finally,the data results are simulated and verified in Matlab,and the advantages of the proposed schemes and algorithms are highlighted compared with three kind of other schemes.(3)Under the online prediction scheme,the condition that the energy has a causal relationship with the task and the energy collected in the future of the prediction and the channel state information are accompanied by the prediction error,an online resource allocation scheme based on the sliding slice window is further proposed,the original problem is divided into the maximum local calculation and offloading calculation bit problem under a single time slot window,and a new joint optimization algorithm is proposed to achieve the optimal solution of the local calculation and task offloading calculation at the user end.Last,it is verified by data simulation,which proves that the proposed joint optimization algorithm has better performance than the benchmark scheme.Therefore,the results of this thesis will provide a new method for optimizing multi-user resource allocation in wireless dynamic MEC systems.