Design And Implementation Of Joint Optimization Strategy For Computation Offloading And Resource Allocation In Mobile-Edge Computing

In recent years,with the development of the Mobile Internet,more and more types of computation-intensive services have emerged in smart terminals,such as face recognition,augmented reality,and virtual reality.These services have the characteristics of high computational complexity,high energy consumption,and delay sensitivity,which will not only increase the network load but also bring huge challenges to the computation capacity and endurance capacity of mobile devices.As one of the core technologies of 5G,mobile edge computing can provide computation,communication,and other resources at the edge of the network close to users,which plays an important role in alleviating the pressure on the central network and reducing the time delay and energy consumption of computation-intensive services on mobile devices.The specific work of this paper is as follows:Firstly,this paper summarizes the domestic and international research status related to mobile edge computing,describes the related technologies of mobile edge computing such as computation offloading and resource allocation,and analyzes the characteristics of ultra-dense networks and the principles of deep reinforcement learning.Secondly,combined with the ultra-dense network,this paper proposes a resource allocation algorithm and a computation offloading algorithm in the mobile edge computing scenario.The resource allocation algorithm analyzes the status of each edge server for computation-intensive services and makes the best resource allocation strategy.Based on deep reinforcement learning,the computation offloading algorithm considers the influence of external communication environment and factors of mobile device and cooperates with the resource allocation algorithm to give the best computation offloading decision.In this paper,the greedy algorithm and the binary particle swarm optimization algorithm are used as the comparison algorithm.The results show that the proposed algorithm has better performance.Thirdly,based on the algorithm proposed above,this paper designs and implements an edge computation offloading and resource allocation system for face recognition services.This paper describes the design and implementation of the system from three aspects: cloud,mobile device,and edge server.In the cloud,this paper uses Docker to create and distribute face recognition service images.On the mobile device side,this paper mainly uses Tensor Flow Lite to deploy the algorithm model on Android devices.At the same time,this paper uses MTCNN and Mobile Face Net to realize the face recognition service on Android devices.On the edge server side,this paper obtains the resource allocation table based on the resource allocation algorithm and uses Docker to realize the automation of the configuration and processing of the face recognition service and the automation of returning the recognition result to the mobile device.Finally,this paper builds the corresponding test topology to test the function and performance of the system,including the verification of the effectiveness of each functional module on the mobile device side and the edge server side,as well as the delay and energy consumption performance of the system under different environmental parameters and different algorithm strategies.The results show that the proposed joint optimization strategy of computation offloading and resource allocation has better performance in the comprehensive cost of delay and energy consumption.