Research On Task Offloading Technology In MEC System With Limited Resources

With the rapid development of the mobile communications and the Internet of Things,more and more mobile terminals are connected to the network,which will generate a large number of emerging tasks such as face recognition,video rendering,and simulation of reality.Due to the limited computing power and battery capacity of mobile terminals,it will be difficult for them to support these emerging tasks.Mobile Edge Computing(MEC)offloading technology is performed by offloading terminals' tasks to the MEC server with strong computing power,which can effectively reduce the calculation delay and calculation energy consumption of the task,and improve the users' experience.However,as MEC server components are constrained by the cost of updating,they will gradually be unable to meet the ever-increasing computing demands.Therefore,this thesis focuses on the research of offloading technology for the resources limited MEC system,the main works are as follows:1.This thesis analyzes the current research status of MEC and MEC offloading technology,and concludes that in order to solve the limited resources of the MEC system,it is necessary to consider whether the MEC has additional resources supplements.When there is no resources supplement,MEC needs to optimize its own limited resources;when it has resources supplement,MEC can be offloaded through resources expansion.2.In the single MEC scenario without additional resources supplement,this thesis adopts the method of price-constrained task offloading to alleviate the problem of insufficient resources faced by MEC.On this basis,this thesis proposes a joint offloading scheme to maximize MEC revenue.The scheme uses Stackelberg game to model the interaction between the mobile terminal and the MEC server,and decomposes the original optimization problem into multiple unrelated sub-problems.Among them,task pricing is accomplished through differentiated pricing strategies;task offloading is accomplished by designing a multi-task offloading algorithm;computing resources for tasks are allocated through convex optimization method;the original value is updated through improving simulated annealing algorithm.Through simulation comparison,the proposed scheme can increase the number of offloading tasks and the total revenue of the MEC server in scenarios with limited computing resources.3.In the multi-MEC scenario with resources supplement,the problem of limited MEC resources in hot-spot cells is solved by means of master-slave MEC cooperatively offloading.On this basis,this thesis proposes a master-slave MEC joint offloading scheme based SDN to minimize the total cost of task execution in hotspot cells.The scheme solves the optimization problem in an iterative manner.In each iteration,the Greedy Based Multi-MEC Selection(GBMS)Algorithm is designed to complete the task allocation among multiple MEC servers;the calculation resources allocation is completed by solving the convex optimization;finally,the scheme try to offload the most costly tasks that were executed locally.The simulation results show that the proposed scheme can alleviate the problem of limited MEC resources in hot-spot cells and effectively reduce the total cost of task execution in hot-spot cells.