Analysis And Design Of Networked Predictive Control Systems With Time-varying Delays

In recent years,networks are increasingly used in control systems with the widespread use of computer networks and the rapid development of communication technologies,which results in networked control systems.Considering that the control objects are increasingly complex and the distribution areas are continuous expanding,the limitations of traditional point-to-point control have become prominent,such as complicated wiring,high cost,maintenance difficulties,poor scalability,and hard to share information.Networked control systems have much advantage in these aspects,but the factors of networks bring new problems to control system,such as network-induced delays,packet dropout,packet disorder,resource limitation,quantization and so on.Much attention has been paid on networked predictive control since it is an effective method to compensate network-induced delays and packet dropout.However,simple predictive control has problems such as large on-line calculations,high bandwidth occupancy and may worsen network environment,especially in systems with limited network resources.Therefore,this paper studied how to design,analyze and improve networked predictive control methods to solve problems in networked control systems better.The main research results of this paper are as follows:Firstly,an event-triggered networked predictive control scheme is proposed for networked control systems with network-induced delays and packet dropout.Assuming that not all system states are measurable,an observer is designed to obtain the state estimation.At the same time,an observer-based event generator is designed to reduce unnecessary packet transmission in feedback channel networks to save network bandwidth resources.In addition,a predictive controller is designed to obtain the future control inputs based on state information and control inputs available.The selector chose control input suitable for current moment and acts on the controlled object through actuator.Finally,sufficient conditions of asymptotic stability and the parameters of the observer and the controller are obtained.Secondly,an H_?event-triggered predictive control scheme is proposed for networ-ked control systems with sensor saturation and actuator saturation.A state estimator is designed to deal with state estimation problem with measurement noise and sensor saturation.To reduce unnecessary packet transmission in feedback channel networks and save network resources,an event generator is designed according to the error of the state estimation.In addition,a buffer-based event-triggered predictive controller is designed to compensate network delays,by which the bandwidth utilization of forward channel networks is also improved.By solving a set of linear matrix inequalities,sufficient conditions for asymptotic stability of the closed-loop networked control system are obtained.At the same time,the desired H_?performance index is also achieved.Thirdly,for networked control system with time delay and packet dropout,the problem of variable-gain networked predictive control under the non-synchronization of sensor network node and controller node clock is studied.According to the switching system theory,the condition of asymptotic stability of closed-loop system is obtained and given in the form of matrix inequality.Considering that the nonlinear matrix inequalities are difficult to be solved directly,the method of cone complementarity linearization is used to convert the nonlinear matrix inequality into a linear matrix inequality,and then the controller parameters are obtained by solving the optimization problem with constraints.By comparing control effect of the variable gain predictive control scheme with that of local control scheme and traditional fixed gain predictive scheme,it is proved that the designed scheme can get closer to the local control effect.Finally,a networked control experiment platform is set up in order to verify the effectiveness of theoretical results in practical applications.The experimental platform consists the linear inverted pendulum of Googol company and the NetCon system of the Institute of Automation,Chinese Academy of Sciences.The inverted pendulum is used as the plant of the networked predictive control system.One NetController is connected to the inverted pendulum as sensor and actuator,the other NetController is used as a remote controller to implement the control and calculation function.The effect of different algorithms on control effect is compared by the experiment results and the effectiveness of the networked predictive control method in practical applications is verified.