Research On The Task Offloading Strategy Based On Mobile Edge Computing

The rapid development of the mobile Internet and the Internet of Things has driven the construction of the future network.Faced with the challenges of high reliability,low latency,large bandwidth,and low power consumption of the future network,a new network architecture called Mobile Edge Computing(MEC)that decentralizes the service platform to the edge of the network.The network provides approximate localized services for terminal devices by wireless network which greatly reducing the delay in service delivery.The computation offloading technology is a key technology to ensure the localization of services.It makes computational intensive tasks offload to the MEC server through the wireless network to complete the computation,so that the computation of the tasks is no longer limited by the computing power and battery capacity of the terminal devices.Under the framework of mobile edge computing,most of the existing computation offloading studies only consider the simple linear topology inside the mobile terminal application,rarely consider the complex topology of the internal module,thus cannot fully utilize the advantages of high transmission rate of mobile edge computing.In addition,due to the limited resources of the MEC system,most of the existing computation offloading studies focus on the load balancing of distributed MEC servers,but there are few studies on computing offloading schemes with specific resource allocation for terminal devices in specific regions.Based on the above two points,this thesis studies the mobile edge computing offloading scheme in single-user terminal and multi-user terminal scenarios.The main research contents are as follows:(1)Under the scenario of single-user computation offloading,a computation offloading strategy based on task partitioning is proposed in this thesis.Since the mobile edge computing architecture with a high transmission rate could withstand relatively frequent computational task offloading,the topology of the application itself and the relationship between the modules will largely influence the performance of the computation offloading.Therefore,the internal structure of the application is considered in this thesis while dividing the computational intensive application into several interdependent components.According to the established application division model,the delay of the execution of the application and the energy consumption of the terminal device are respectively indicated.The energy consumption generated by the terminal device is taken as an optimized target with the constraint of the delay.A heuristic artificial fish swarm algorithm is designed for the optimization model to find a suboptimal solution of this problem.The simulation results show that the task partitioning computing offloading scheme in the single-user scenario proposed in this paper can effectively reduce the terminal energy consumption while meeting the computational delay requirements.(2)Under the scenario of multi-user computation offloading,a computation offloading scheme with resource allocation is proposed in this thesis.Due to the limited computing power and communication capability of the MEC system,there are cases where multiple terminal devices compete for limited resources so that the computation offloading requirements of too many terminals cannot be satisfied at the same time.Therefore,the data communication model,local and edge computation models of tasks in multi-user scenarios are established in this thesis so that we can form the expressions of delay and energy consumption of each model.With the decisions of computation offloading,the communication and computing resource allocation taken as the solution to be optimized,the delay and energy consumption of all terminals are optimized with a certain weight indicator at the same time.According to the optimization model,a heuristic algorithm combining Genetic algorithm and Whale optimization algorithm is designed to find a suboptimal solution for this problem in this thesis.The simulation results show that the computation offloading algorithm in the multi-user scenario proposed in this thesis can greatly reduce the cost caused by task computing under various dynamic conditions.