Research On Multi-user Task Offloading Scheme In Cloud-edge-end Environment

With the popularity of mobile computing and Internet of things technology,task offloading has been widely used in image/speech recognition,natural language processing and Internet of vehicles.Effective application of task offloading can deal with large-scale computing tasks and improve the computing efficiency and resource utilization of the system.At present,there are two major problems in the research of task offloading: first,the research on optimization algorithm and offloading strategy is still relatively ideal;second,many task offloading schemes have data security risks,and the task calculation does not fully consider collaborative operation.At the same time,with the popularization of 5G and the emergence of artificial intelligence,the "Cloud-Edge-End" environment has become a hot topic in the current digital field,which solves the limitations of traditional task offloading schemes to a certain extent.Therefore,this thesis focuses on the task offloading methods in the "Cloud-Edge-End" environment,and proposes different task offloading optimization schemes for unilateral and cluster systems.The main tasks are as follows:(1)Aiming at the problems of task offloading under the simple system model,such as ignoring edge-cloud collaboration,too many parameter Settings,and excessive running cost of the optimization algorithm,this paper proposed a single edge node task offloading optimization scheme.In the scheme,a three-layer framework of multi-user task offloading with single computing access point participation is designed,and end users in the simple system architecture can choose three modes: local computing,mobile edge computing and mobile cloud computing.The data receiving and computing units of the edge layer are divided into the base station and the computing access point CAP.The base station is responsible for the computing tasks of the user layer at the receiving end and transfers them to the nearby CAP for computing processing.It avoids the tedious processing of multiple transactions in one unit at the same time,and the forwarding process does not consume energy.The offloading strategy is represented by 0-1 matrix,and the particle swarm optimization algorithm is used to train the optimization goal of the system.In the optimization of energy consumption and delay,the performance advantage is reflected compared with the single calculation method.(2)Aiming at the problems of target conflict,task allocation and security in multi-objective task offloading scheme,a task offloading optimization scheme for MEC cluster is proposed.The scheme facilitates task allocation and forwarding through the MEC cluster to relieve pressure on several computing access points in the edge layer.The scheme uses the relay mechanism to realize the "two-hop" task to reach the cloud through the base station,ensuring the communication quality and data security in the process of task forwarding.The offloading strategy is obtained by combining multiple matrices,and the particle swarm optimization algorithm is used to optimize the system optimization goal.Compared with the single calculation method in three levels of energy consumption,time delay and system loss standard function,the performance improvement of this scheme is reflected.