Research On Intelligent Matching Of Terminal Tasks And Resources In Cloud-Edge Collaboration

With the development of Io T technology and the popularity of wireless networks,the number of mobile user devices is increasing and there is a higher demand for performance indicators such as latency,energy consumption and throughput.The traditional centralized processing model of cloud computing can hardly meet the high performance requirements of users,and edge computing technology has emerged.Edge computing sinks cloud services to provide powerful computing,storage and communication capabilities at the edge of the network.How to fully combine the advantages of cloud-centric and edge computing models and rationalize task offloading and resource allocation is an important issue that needs to be addressed urgently.We take edge computing and cloud-centric collaboration as the background,and focus on how to make task offloading decisions adaptively and solve task correlation problems.Based on the existing theories,the problem of minimizing the delay and energy cost in the task offloading mechanism of cloud-edge collaboration network terminals is solved by considering computational resources,channel models,and adaptive offloading decisions.The main work of the thesis is as follows:(1)Game-theory-based task offloading and resource scheduling in cloud-edge collaborative systems.To analyze the multi-terminal user coarse-grained task offloading and resource scheduling problem in the cloud edge collaborative network,the thesis considers the heterogeneity of mobile devices and inter-channel interference,takes minimizing the weighted sum of system delay and energy consumption as the optimization objective,constructs a multi-terminal user,multi-access edge server system architecture,establishes the task offloading decision of multiple end-users as a task offloading model based on game theory The task offloading strategy is obtained by an improved particle swarm optimization algorithm based on game theory,which achieves the Nash equilibrium of the multi-user computational offloading game.Finally,it is demonstrated through simulation experiments that the algorithm can effectively reduce the system overhead in the task offloading process.(2)Deep reinforcement learning-based fine-grained task offloading with cloud-side collaboration.To further explore the task offloading mechanism in the cloud edge collaborative network,a fine-grained task partitioning and offloading strategy is studied,a fine-grained task partitioning model is constructed,and a fine-grained computing task offloading method based on deep reinforcement learning is proposed.The model and method consider the adaptation of computing,storage,bandwidth and other resources to task offloading,and solve the optimization problem of minimizing the sum of task offloading delay and energy consumption weight.First,the end tasks are constructed into a directed acyclic graph structure that can be executed in parallel based on the dependency-aware relationship between tasks,and then the offloading decision is made by the actor network at the edge nodes,and the optimized parameters are uploaded to the global network located at the cloud center,and the network parameters are updated in collaboration with the cloud center.Multiple edge nodes execute asynchronously to update the offloading model in time through feedback rewards,thus tuning the actor-critic network to obtain the optimal solution for the offloading decision.Finally,the proposed algorithm is experimentally demonstrated to be robust and effective in reducing task execution delay and energy consumption.(3)Research on the application of cloud-side collaborative task scheduling method in "Intelligent Gatekeeper".The epidemic has swept across the world,bringing a huge impact on our production life and economic development.Although the joint prevention and control mechanism and grid-based management have effectively suppressed the spread of the epidemic,they have consumed huge human and material resources and increased the security risks of prevention and control personnel.In the framework of cloud-side-end synergy theory and technology,an automated epidemic prevention and control "intelligent gatekeeper" system based on big data collection and artificial intelligence prediction can analyze and process large-scale epidemic data in a timely manner,determine the risk level of users to make prevention and control measures,and liberate manpower.The task offloading strategy studied in this thesis can provide a design idea and technical support for the "Intelligent Gatekeeper",build a set of regional grid-based management system with more timely response,better function and more intelligent operation,and play a role in promoting cost reduction for accurate community epidemic control.