A Novel Edge-Cloud Collaboration Architecture Design And Implementation Base On The Industrial Internet

The industrial technology plays an important role in the national comprehensive production capacity system,and the development of industrial manufacturing is vital for the national economic development.In recent years,with the development of industrial technology and the promotion of "intelligent manufacturing 2025" in China,many computing intensive applications have emerged,like industrial object detection,face recognition,multimedia,etc.Meanwhile,with the development of mobile Internet and wireless communication,the Industrial Internet of Things faces opportunities and challenges.Considering numerous devices,the network infrastructure faces much pressure.As for the Cloud Computing,in the scenario of Industrial Internet of Things,many sensors are physically far away from the cloud platform,which cannot meet the low latency and high quality requirements of delay sensitive applications,the quality of service is intolerable.Edge Computing pushes the data and services from the centralized node to the edge of the network.It integrates the open platform with core capabilities of network,storage,computing and application in the place where close to mobile devices and data sources.Edge computing provides intelligent services which close to the user.Besides,it promotes the development of Internet of Things.However,high computing consumption applications migrate to the embedded system of edge computing.The edge cannot meet the demand because of the heavy workload,the performance is not as good as the cloud.Therefore,combining the advantages of cloud computing and edge computing is important.In the industrial Internet,we need to construct a new edge-cloud collaboration network framework which is Edge-Cloud Collaboration(ECC).In this thesis,firstly,an edge-cloud collaboration framework is proposed.Then,an energy minimization task offloading mechanism with ECC in IoT networks is formulated.In the end,a new edge-cloud collaboration platform which base on the proposed architecture is constructed.The edge-cloud collaboration architecture is deployed to the actual industrial production line in order to improve the efficiency as well as the network performance.The main work is as follows:(1)In the multi task scenarios,a novel edge-cloud collaboration framework is proposed and the demonstration platform is deployed.In order to avoid the heterogeneous problems,the Raspberry PI(RP)computer is used as the working node,and an edge computing oriented edge-cloud collaboration framework is designed.The framework is mainly divided into three layers,the Edge Raspberry PI(ERP)layer,the Monitoring&Scheduling Raspberry PI(MSRP)layer and the Cloud Center(CC)layer.The collaboration communication between different RP and layers are established via wireless connection.Through the time-consuming analysis for each layer of the framework,the overall time delay problem of the framework is modeled to minimize the working time of the edge-cloud collaboration.Through the scheduling of MSRP,ERP can collaborate with CC,the ERP makes full use of the computing resources between the edge and cloud.ERP layer runs two representative edge tasks,face recognition and object detection.(2)In the delay constrained scenario,an edge-cloud matching algorithm which base on an optimal power search algorithm is proposed.Firstly,considering the channel state,delay constraint and utility function,a minimum energy consumption task execution system model with delay constraint penalty mechanism is proposed in the condition of edge-cloud offloading.Then,based on the modeling problem,a binary search algorithm for the optimal transmit power and a task offloading algorithm for fast weight matching are proposed,so as to find the optimal transmit power of edge-cloud collaboration and the appropriate edge-cloud matching.(3)In the actual production line,combing with the IoT node and 5G network infrastructure,an IoT end-edge-cloud collaboration industrial intelligent portrait detection platform is constructed.The deep learning model is integrated under the advantages of high-speed industrial Internet,so that the edge computing and the edge-cloud collaboration technology can be implemented in the actual industrial production line with IoT devices.The proposed framework and platform have practical application value.