Software-defined Modeling Method Of Cyber-physical System Driven By Big Data

Cyber-physical system(CPS)deeply integrates the information world with the physical world through a variety of sensors,changing the way people interact with the physical world.With the development and progress of human society,the generation of data increases exponentially,which makes CPS face great challenges.First,as an embedded system,CPS has limited storage and computing capacity.How to store and calculate massive CPS data is a challenge.Secondly,CPS is a 3C system combining computing,communication,and control.With the increase of CPS types and numbers,how to control and manage these CPS nodes is another challenge.To solve these two problems,this paper combines the ideas of cloud computing and software definition and proposes a CPS architecture based on big data and software definition.First,the control layer is separated from the CPS physical equipment by using the concept of software definition,and is centrally controlled and managed by the control center.Then,the cloud computing is used to analyze and process these data,and the processed results are fed back to the control center to control and manage physical equipment.This architecture combines the big data technology and software definition into CPS,changing the traditional CPS data storage and computing methods,as well as the interaction between CPS.Finally,the CPS architecture is applied to the Internet of Vehicles system,and a five-layer Internet of Vehicles CPS system is proposed,including application layer,cloud computing layer,control layer,roadside unit layer and vehicle unit layer.In this paper,we analyze this structure in terms of architecture and data modeling.First,we use the Architecture Analysis & Design Language(AADL)to model the Internet of Vehicles system.By adding annex,we conducted end-to-end flow analysis and the safety analysis of the overall structural design of the model and verified the effectiveness and efficiency of the CPS architecture based on big data and software definition.Then,we take the vehicle sensor data as an example and propose a neural network data processing model based on the residual structure.By comparing with K nearest neighbor,random forest,decision tree and traditional neural network model,the neural network model based on residual structure proposed in this paper is superior to these methods,which validates the effectiveness and accuracy of the data analysis and processing method proposed in this paper.