Research On Computation Offloading Mechanism In Tactical Internet Of Things

With the rapid development of new applications such as augmented reality,face recognition and crowdsensing,huge computing needs have been generated and higher requirements have been put forward in response delay and energy consumption.However,limited in computing resource,storage capacity and battery capacity,Io T devices cannot meet the above requirements.Edge computing is an ideal solution to this problem.By providing a low-latency and high-bandwidth computing environment at the edge of the network,tasks are migrated to edge servers for processing,thereby effectively reducing response delay and energy consumption.However,in the tactical Internet of Things,the communication environment is relatively harsh,and the network condition shows the characteristics of easy interruption,intermittent connection and low bandwidth.In this environment the use of computation offloading technology needs to consider the real-time network conditions between edge and cloud and between devices and edge,and adjust the offloading strategy according to context changes,so as to achieve highly available edge service.This thesis focuses on this issue.First,in the tactical Internet of Things,the network connection between edge and cloud is unstable and interrupted.When Io T devices migrate tasks to the cloud or edge servers,the communication cost is seriously affected by the network,may not be suitable for offloading.Therefore,the Io T distributed application was abstracted as a task dependency model,and the dependencies between tasks were analyzed considering the task calculation cost and communication cost.The multi-site collaborative computing problem was modeled as a cost model.To solve this problem and achieve best offloading strategy,a genetic-based computation offloading algorithm was proposed,which considered network conditions during the application execution.The simulation results show that the algorithm could effectively reduce the application execution cost in the tactical Internet of Things and has better performance than other solutions.Then,this thesis further carried out research on computation offloading when the network between iot device and edge server is inturrupted.At this time edge servers cannot provide effective service.However,in the edge environment,the network condition between iot devices are relatively good and some idle deivces are sufficient in computing resource,which have the ability to provide computing resource to cooperate for task execution.Therefore,this thesis conducted research on edge-end collaborative computing,using idle deivces to provide computation offloading service and combining with Stackelberg game to analyze the decision-making process between task devices and idle devices.Then a distributed two-layer iterative algorithm was proposed to solve the problem and achieve the balance between cost and utility.The simulation results show that the designed algorithm can effectively reduce the cost of task devices and improve the system utility.Finally,an edge computing system was designed and implemented,whose functional modules of the device,edge gateway and cloud platform are developed.The feasibility of the proposed solution was tested through image stitching application and object recognition and tracking application.The test results show that the system can achieve the expected results.