Receding Horizon Estimation Of Unreliable Networked Control Systems

The networked control systems is a real-time distributed feedback control system which realizes spatial distribution of sensors,controllers,actuators and controlled objects through shared communication network.It has been widely concerned on and studied in aerospace,process control,environmental control,traffic control and other fields.Sensor network can carry out data transmission and information processing between sensors and observers.Due to inherent limitations of the shared communication network,the data in transmission process inevitably presents time-delay and packet losses,which lead to the unreliability of data transmission in networked control systems and degrade the performance of the systems.In this paper,we mainly investigate the receding horizon estimation for unreliable networked control systems.The new design methods of centralized state estimator and distributed information fusion estimator are proposed for systems with time-delay or packet losses based on receding horizon estimation theory.The research work is completed as follows:The receding horizon estimation for time-delay system.For the multi-channel observation delay system or the random observation delay system,the delayed system is firstly transformed into the delay free system based on the observation reorganization technique.The batch and iteration formulas of the receding horizon estimator are derived based on the linear unbiased estimation theory.Stability analysis and simulation example compared with the existing Kalman filter by MATLAB are given.The receding horizon estimator has good tracking performance.It is worth pointing out that the proposed receding horizon estimator is obtained by solving two type of difference Riccati equations which are with the same dimension as that of the original system.It is shown that the proposed methods effectively avoid the computational complexity caused by the state augmentation method and reduces the amount of calculation.The receding horizon estimation for diagonal matrix packet losses system is studied.The diagonal matrix multiplicative noise is introduced to represent packet losses.The model of packet losses system is established.Based on the theory of linear unbiased estimation and the operator of Hadamard product,the receding horizon estimator in batch form and iteration form are proposed,and the stability analysis and simulation example are given.It is worth pointing out that,distinct from the scalar packet losses system model,the diagonal array losses system model can effectively describe the packet losses situation corresponding to each state component,which is more general.In addition,the gain of proposed receding horizon estimator is obtained by solving a set of difference Riccati equations including Hadamard product.The receding horizon estimation for dual-channel time-delay system with diagonal matrix packet losses is studied.Based on the research results of the time-delay system and the diagonal array packet losses system,the two-channel time-delay system model with diagonal array packet losses is established.And the observation reorganization technique is used to transform the system into the delay free system which is just with multiplicative noises.The new receding horizon estimator is derived.And the stability analysis and simulation example are given.The receding horizon estimation of multi-sensors distributed information fusion is studied.Aiming at the model of multi-sensors scalar packet losses system or multi-sensors time-varying delay system,based on the linear unbiased estimation,the scalar weighted fusion algorithm,and the observation reorganization technology,the local estimator and distributed receding horizon estimator are derived.It is proved theoretically that the fusion estimation error variance matrix is smaller than the local estimation error variance matrix.At the same time,the simulation results show that the distributed receding horizon estimation based on scalar weighting algorithm has better tracking performance than the local receding horizon estimation.