Research On Intelligent Task Scheduling Strategy In MEC And Its Application In ITS

With the development of intelligent mobile terminals and the popularity of emerging artificial intelligence applications(such as driverless,virtual reality,and augmented reality),mobile devices are facing huge challenges due to their limited computing power and battery capacity.Traditional cloud computing has powerful computing capabilities,but facing the limitation of response time and transmission bandwidth brought by long distance,those computation-intensive and delay-sensitive applications cannot get a good feedback result.To overcome these difficulties,mobile edge computing came into being.MEC has quickly become the mainstream direction of future network development due to its powerful computing and storage resources,as well as short distances,low latency,high bandwidth,and low energy consumption.The main problem in MEC is to solve the task scheduling problem,and the key of task scheduling is how to perform computing offloading,which is also the focus of this paper's research work.At present,many researches on users' computing offloading fail to consider the problem of users' choice of server.In the actual situation,the server's limited computing resources and complex network environment will lead to many problems,and its unbalanced resources and load become the key factors that limit the server's scalability and improve network performance.Therefore,considering the finiteness of server resources and thecomprehensive consideration of time delay and energy consumption during task scheduling,this paper designs an intelligent task scheduling strategy in MEC,and applies the strategy to intelligent transportation system.In this paper,the task scheduling problem in the MEC environment is studied as follows: at first,this paper expounds the research status of MEC,and analyzes the current research focus.Secondly,the relevant theories and techniques used in this paper are introduced.Then the intelligent task scheduling strategy in MEC is modeled and designed in this paper.In the multi-user multi-server scenario,computing offloading is an important research direction of task scheduling.At first,this paper designed a kind of task scheduling strategy based on Priority,which provides more options for computing offloading of mobile users and makes the computing resources allocation is difficult to ascend.The server provides a priority level for each user who applies for offloading request.In the case that the user is faced with multiple servers which can choose to computing offloading,the server with the highest priority level shall be selected first for unloading.This strategy set up under the constraints of local maximum delay calculated computing offloading total success rate maximization problem of the optimization model.Based on the proposed Priority-Based task scheduling strategy,Q-Learning algorithm in Reinforcement Learning and convolutional neural network in Deep Learning are used to further design a task scheduling strategy based on Deep Q-Learning,which improves the efficiency of the Priority algorithm.In this paper,a street scene is simulated and an experiment is carried out for the intelligent task scheduling strategy under the vehicle network scenario in MEC.The simulation results show that the algorithm designed in this paper can largely improve the success percentage of the computing offloading.At the same time,also can effectively reduce the user in theprocess of task scheduling time delay and energy consumption.Especially in the case of edge computing server resource shortage,the intelligent task scheduling strategy designed in this paper can make full use of the computing resources and storage resources of the server,improve the performance of the whole system.