Various Predictive Control Researches Of Networked Systems

With the development of industry and the progress of science and technology,the network is gradually embedded in the control systems,where distributed components communicate and transmit data with each other via the network.Compared with the traditional point-to-point systems,the networked control systems have brought many advantages,such as no limitation of space,lower consumption and cost of energy,easier to maintain and expand,sharing the same resources.Thus,networked control systems are widely applied in various fields,such as vehicle engineering,aerospace and maritime engineering.But at the same time they also bring some problems,such as packet dropouts,packet disorder,delays,which are induced by the network.These problems may damage the performance of the system,and even make the system unstable.To solve these problems,a new kind of methods named predictive control mechanism is proposed to compensate for delays or packet dropouts.Predictive control is a good way to track signals,resistant to interference and overcome the uncertainties of the models,without requiring the accurate mathematical models of the system.It also can deal with the pure lag processes directly,so it has an excellent control performance in practical applications.To compensate for packet dropouts and networked-induced delays,this paper mainly concerns about three predictive control methods on the complex networked systems,i.e.predictive control method based on observer,T-S fuzzy predictive control method and data-driven predictive control method.First of all,a method of predictive control based on observer for multi-agent system to compensate for the networked-induced delays is presented in this paper.In the networked control systems,because the delays are usually influenced by network protocols and time synchronization between different components,there are some errors between the actual values and the measured values of the delays.That is to say,the actual values of delays are not always equal to the measured values of delays,i.e.the delays don't match.Taking into account that the delays don't match in the discrete multi-agent system,this paper provides a predictive control method based on observer to compensate for communication delays.What's more,by using the output feedback control based on observer,the sufficient conditions which make the closed-loop multi-agent system achieve stable and consistent are obtained.Secondly,this paper presents a T-S fuzzy predictive control method to compensate for the data packet dropouts.According to the T-S fuzzy system with different memberships,a new event-triggered condition is designed.The new event-triggered condition can not only avoid the disadvantages that relative event-triggered mechanism send more data when it is close to the equilibrium point of the system,but also avoid the detect that the absolute event-triggered mechanism send more data when it is away from the equilibrium point.In this way,it can reduce the number of data to be transmitted.Under the new event-triggered condition,the system can save more resources of calculation and bandwidth-limited network.Then based on the T-S fuzzy predictive control,packet dropouts can be compensated.By using the public Lyapunov function and imperfect premise matching method,the sufficient conditions of the bounded stability of the system are obtained.Finally,this paper provides a data-driven predictive control method.Applying this method,we can only consider the input and output data of the system,without knowing the accurate model of the system.In this paper,for the nonlinear networked control systems with packet dropouts and delays in both the sensor to controller channel and the controller to actuator channel at the same time,we transform it into an equivalent linear model at first.Then,a delay compensator,which is based on the sequence of predictive control,is designed to compensate for the packet dropouts and delays.And we can obtain the sufficient conditions which make the system stable and the output error convergence.