Research On Task Offloading Strategy Based On Mobile Edge Computing

In recent years,in the face of the massive amount of data generated by many emerging applications,mobile devices themselves are no longer able to meet the demand due to limitations such as battery capacity and computing power.While cloud platforms contain rich computing resources,data transmission delays can put tremendous pressure on the entire network,so considering mobile edge computing with intelligent processing capabilities at the edge of the network has become a potential research direction in the 5G scenario.The mobile edge computing paradigm accelerates the response time of user services by offloading task data to the edge for execution.To take fuller advantage of computing at the edge,task offloading techniques have been widely discussed in academia as a key research direction in mobile edge computing.In this paper,task offloading strategies for different mobile edge computing scenarios are investigated to address the problems in existing work.(1)A heterogeneous multi-edge server system in a dense cell is investigated by combining dense cell network technology with mobile edge computing technology.First,a Directed Acyclic Graph(DAG)model is constructed based on the dependency constraints that exist between tasks in mobile applications.Then,an optimization problem is proposed with the goal of reducing the average completion delay of system applications based on satisfying the application deadlines.Finally,a three-stage offloading algorithm is designed by analyzing the urgency of the tasks to be processed.The first stage assigns application priorities,the second stage assigns task priorities according to the inverse-order dynamic programming method,and the third stage task offloading,where tasks are scheduled separately on heterogeneous computing resources based on prescribed principles.Simulation experiments show that this paper effectively improves the completion rate of applications and reduces the average completion delay of system applications by jointly scheduling multiple applications and multiple tasks.(2)By combining computational offloading techniques with load balancing techniques,an edge network collaboration system with limited resources is constructed.First,the Breadth First Search(BFS)algorithm is used to design the scheduling list model to satisfy the dependency condition constraints between tasks and to maximize the parallelism between tasks in the horizontal direction.Then,cross-server collaboration is introduced into the model in order to coordinate the load situation of different edge servers.Then,the problem of reducing system completion delay is decomposed into a joint offloading and migration problem for each scheduling layer,and the joint problem can be modeled as a multi-leader-multi-follower Stackelberg game considering the interaction effects between tasks and edge servers.Finally,the existence of Stackelberg equilibrium is proved,and a Q-value-based offloading algorithm and a distributed iterative migration algorithm are proposed to achieve Stackelberg equilibrium.The simulation results show that both the model and algorithm in this paper can effectively reduce the system completion delay in multi-user and edge server environments.(3)Based on the limited resources of mobile terminal and edge-end,a collaborative network offloading system is designed for cloud-edge-end to maintain the stability of service quality when the number of terminals increases.First,different offloading scenarios between tasks are discussed considering the communication between tasks.Then,a multi-objective offloading model that minimizes application completion delay and mobile terminal energy consumption is proposed for different types of terminal requirements.Then,priority is assigned to each task based on the average computational value of the task,through which the execution order of tasks can be reasonably arranged,thus avoiding malicious competition for computational resources by multiple tasks and preparing for subsequent task offloading.Finally,a D-NSGA algorithm is proposed to find the optimal offloading strategy.a two-population strategy is introduced in the D-NSGA algorithm considering the characteristics of different populations,and the diversity of individuals is explored by using different ranking selection methods for the two populations to avoid the algorithm from falling into local optimum.Simulation results demonstrate the superiority of the proposed strategy compared with other baseline strategies in reducing the application completion delay and the energy consumption of mobile terminals.