Tasks Offloading And Resource Allocation Strategy Of Mobile Users Based On Edge Computing

Mobile Edge Computing(MEC)deploys computation power at the edge of the network close to the data source to meet users’ various business demands of high energy consumption and low delay,thus effectively reducing load pressure of core network.With the rapid growth of edge data,it is imperative to use machine learning methods to fully release the application potential of edge big data.For example,using reinforcement learning to solve tasks offloading problem can effectively improve the overall revenue,while the use of federated learning architecture for coordinating edge computing resources can guarantee the privacy of local data when training data model.Based on above background,the main research work of this thesis is as follows:(1)In the mobility-aware MEC network,a reinforcement learning based user computation task offloading and migration strategy is proposed.The mobility of users may cause them to leave the area covered by the server before offloading tasks are returned,so it is necessary to transfer tasks between base stations to ensure the integrity of offloading process.However,the task migration process will also cause additional overhead and reduce the user benefit.In order to maximize the total benefit of the user system,a task offloading algorithm based on Q-learning algorithm of reinforcement learning is proposed.This algorithm enables the system to calculate the resource state according to the edge nodes,and continuously optimize the calculation offloading strategy through the exploration learning method,so as to reduce the occurrence of the task migration process and ensure the maximum total benefit of the system,which is considering the task processing delay and the user energy consumption.In the simulation experiment,the proposed algorithm is compared with other tasks offloading schemes,and it is verified that the application of this algorithm can bring obvious benefit improvement to the system.(2)In the two-layer heterogeneous MEC network,a user computation task offloading and resource allocation strategy with federated learning is proposed.Considering the characteristics of the two-layer structure of heterogeneous network,the MEC servers are installed on macro base station and small base stations respectively to construct the federated learning distributed structure.Macro base station collects local model parameters uploaded by each micro base station,then summarizes and updates them into a global model that can predict the optimal tasks offloading decision,actively determining task offloading strategy for the users.And then,Lagrange Multiplier Method is used to obtain the optimal edge node computing resources and user power resources allocation scheme.Becasuse of the limited computing power and battery life of the device,local processing tasks or transfer tasks to be offloaded both require power.Compared with other tasks offloading and resource allocation schemes,the simulation experiment proves that the proposed algorithm can effectively save the energy consumption of computation and transmission for users.