Research On QoE-oriented Edge Computing Task Offloading Technology

With the rapid development of the Internet of Things technology and the increasing complexity of data,the traditional cloud computing model cannot effectively meet the needs of users for low latency and low bandwidth.In order to solve such problems,this article is based on edge computing,combined with the Internet of Things technology,to build the Intelligent Internet of Things(AIoT).Firstly,the coarse-grained data collected by the sensing layer terminal is preprocessed to solve the impact of redundant data on the accuracy of the final result;secondly,an adaptive task offloading method is used to offload the fine-grained data stream obtained by preprocessing to The optimal edge node processing improves the efficiency of the task offloading method and the computing resource utilization of the edge node;through the design of related monitoring methods,the reliability of the task offloading method is guaranteed,and the purpose of improving QoE is achieved.The main work of this research is as follows:1.Aiming at the problem of a large number of redundant data in the data stream collected by the Internet of Things,a data stream granularity division method based on an improved residual network is designed.In order to solve the problem of gradient explosion and gradient disappearance in the training process,CNN and ResNet-18 models are deeply analyzed,ResNet-18 is selected as the basic model,DLBHC algorithm is introduced,and ResNet-18 is optimized.Add an intermediate layer between the penultimate layer of the model and the final task layer of the model,convert the obtained high-dimensional feature vector into a string of low-dimensional binary hash codes,and compare the binary of the test data with the tested data Hash coding,calculate the Hamming distance between them,and judge whether the Hamming distance is within the threshold range,and realize the granularity division of the data stream.This method can effectively reduce the impact of redundant data on the results,ensure the accuracy of the final results,and improve QoE.Experimental simulation shows that the accuracy of DLBHC + ResNet-18 is 3.69% higher than that of ResNet-18.The accuracy rate of DLBHC + ResNet-18 is 6.94% higher than that of DLBHC + Alex Net.2.Aiming at the load balancing problem in the task offloading process,an adaptive task offloading method based on ACS is proposed.Load balancing factor is added to the setting of total pheromone,and the optimal edge is selected according to the change of load balancing factor Nodes complete computing tasks,and this method can effectively reduce the impact of high latency in the execution of the task offloading method,and improve QoE.In order to ensure the normal operation of the task offloading method,an offloading monitoring method based on MQTT is designed.Through the callback message received on the client,the data uplink and data downlink status during the task offloading process are monitored in real time.The relevant status code provided by the callback message can solve possible problems in a timely manner,ensure the reliability of the task offloading method,and meet the QoE to the greatest extent.Experimental simulations show that when the number of iterations is the same,the convergence rate of this method is 12.59% higher than that of the traditional ACO algorithm,and the convergence rate of this method is 3.28% higher than that of the MMAS algorithm.