Research On Task Unloading Method In Mobile Device Cloud

With the continuous innovation of mobile Internet and wireless communication technology,the number of mobile applications and services has maintained a rapid growth,making more and more computing intensive tasks running on mobile devices,such as face recognition,interactive games and augmented reality.Usually performing these computing intensive tasks requires a lot of computing resources and energy consumption of the device.At the same time,the number of intelligent terminals,such as smart phones,tablet computers and vehicle terminals,has also increased exponentially.These heterogeneous terminals have great differences in battery capacity,memory space,CPU computing power and other performance,and a large number of terminal devices can not meet the requirements of new mobile applications for low latency and high reliability.Computing offload has become a possible solution to coordinate the contradiction between the high requirements of new mobile applications and the limited resources of terminal devices.Different from the traditional centralized remote cloud service,mobile device cloud is often used to refer to a small cloud composed of several mobile devices with the functions of computing and storage,which is close to the user in physical distance.Devices in the cloud cooperate with each other by sharing communication and computing resources,so as to meet the needs of users.Terminal devices can delegate tasks to neighboring available devices in the same network,and the entrusted devices perform computing,storage and other operations,and then expand their own capabilities to neighboring nodes to complete tasks that are difficult to complete under their own capacity constraints.Because of the close physical distance between devices,they have the characteristics of high flexibility,which can save energy consumption or reduce time delay Therefore,the proposal and utilization of mobile device cloud is very valuable.The research content of this thesis is the task unloading in the cloud environment of mobile devices and proposes a feasible offloading plan for the system's time delay,energy consumption and other indicators,and gives proof in combination with the simulation results.The contributions of this thesis are as follows:(1)In view of the free task unloading in the mobile device cloud,a task allocation method based on Kuhn-Munkres(KM)algorithm is proposed.By using the idle resources of the surrounding nodes,the total delay or total energy consumption of all nodes in processing tasks is reduced.Firstly,according to the multi-dimensional attributes of the task,such as the calculation load and the latest completion time,the priority of task execution is determined by Analytic Hierarchy Process(AHP).Then,the optimization model of time delay and energy consumption is established,which is transformed into the maximum weight matching problem of bipartite graph.The KM algorithm is used to solve the optimal solution of task allocation,and the efficient cooperative execution of terminal nodes at the edge of the network is realized.Through simulation verification,the algorithm proposed in this thesis can effectively decrease the delay or energy consumption of task unloading compared to the task being executed locally.(2)Considering the selfishness and personal rationality of users,a task unloading method based on Gale-Shapley(GS)matching algorithm in mobile device cloud is proposed.First of all,we design a reasonable price for each task initiating device based on the task attributes and the characteristics of the local device.The price is used to measure the cost that the task initiator needs to pay per unit time of the task executor.Then,based on the price,we can calculate the benefits of the task initiator unloading the task and the task executor executing the unloaded task.Finally,we take the social utility maximization as the goal,solve the problem by using GS bidirectional matching algorithm,and compare with the central solution.Through simulation analysis,when the social utility is close to the central method,the computational complexity of the proposed algorithm is much lower than that of the central method.