Design Of Networked Control System Based On Communication Feature Learning

Networked control system is a closed-loop control system formed by connecting sensors,controllers and actuators distributed in different geographical locations through digital communication network.Compared with traditional control system,networked control system has been widely used in intelligent home,telemedicine,intelligent transportation,industrial automation and other fields because of its advantages of low cost,easy maintenance and remote control.However,the non real-time nature of digital communication network challenges the real-time requirements of control system,and the design and analysis of networked control system with communication constraints such as packet loss and delay have attracted the attention of scholars.In the past research,whether it is based on traditional control algorithm to deal with communication constraints or communication and control joint design method,most of them need to know the characteristics of network communication.It is usually assumed that the characteristics of network communication are known.However,in some complex network environment,communication characteristics are easy to change,the actual communication characteristics are also easy to change when the system is running,and the law of change is unknown.Therefore,the controller designed on the basis of given packet loss rate or delay value can not meet the real-time requirements of the control system,even the system performance deterioration or even instability due to the change of network state.In this thesis,the characteristics of dynamic network communication in networked control system are studied,and the corresponding optimal control strategy and the design and analysis of dynamic data package are proposed.The specific work is summarized as follows:(1)In view of the dynamic change of packet loss characteristics in networked control system,a learning algorithm of packet loss characteristics is proposed,and the optimal controller is designed according to the learning results,which solves the problem that the packet loss rate should be assumed in advance when the traditional optimal controller is designed.Firstly,the initial Bernoulli model is established to solve the packet loss problem in the communication network of networked control system.Based on Bayesian inference,a method of learning the current network packet loss model is proposed.Then,the packet loss rate of the network is estimated in real time by the latest packet loss model.Then,the optimal controller is designed according to the real-time estimation of packet loss rate.Finally,the effectiveness of the optimal controller based on packet loss rate learning in improving the performance of traditional optimal controller system is proved by MATLAB numerical simulation.(2)Aiming at the problem of dynamic change of network delay characteristics in networked control system under shared network,a learning algorithm of delay characteristics is proposed,and a dynamic packing scheme is designed according to the learning results,which solves the problem of control system consuming too much network resources under shared network environment.Firstly,communication constraint modeling is carried out to solve the problem of packet loss delay in networked control system.Based on the Bayesian inference method introduced in Chapter 2,a model learning and updating scheme using the recent delay data is proposed.Then,a delay estimation method based on real-time communication characteristic model is designed.Then,the problem of effective utilization of network resources in the packet based control framework is considered,which is conducive to the design of a dynamic packet scheme for real-time estimation of network delay.Finally,the effectiveness of the scheme in saving network resources is verified by MATLAB numerical simulation.