Study For State Estimation Problems In Networked Control Systems

State estimation is not only an important branch of the modern decision-making and control theory, but also the important precondition for guaranteeing the safety and economy of business processes. In the 21st century, control system becomes an integration of network, communications and control systems. However, in Networked Control Systems (NCSs), since the introduction of network, it makes control system become larger, more complex and include various uncertainties, which makes classical state estimation theory difficult to be applicable. Hence, state estimation problems in Networked Control Systems become a forefront topic in automation field.In the paper, in the network environment where exists various types of uncertainties including intermediate uncertainty and parameter uncertainty, some state estimation problems of Networked Control System are studied aiming to propose appropriate strategies of state estimation and some state estimation stability conditions.When observations are partitioned into several parts to deliver through network subject to communication limitation, for discrete-time linear systems with partial or entire packet losses described by a two state Markov chain process, based on a two state Markovian packet dropout model and the expanded Kalman filter updates with partial packet, conditions for stabel estimator are given.For discrete-time linear systems simultaneously with Markovian delay and packet loss, through setting a buffer with appropriate length at estimator site, a discrete time-invariant system with multiple-state Markovian delay and packet loss is modeled as a time-variant system whose data reception process is described by a two-state Markov chain. Based on the proposed Finite Reception History Estimator (FRHE), the optimal FRHE design strategy is proposed under a known maximum delay.For linear discrete time-varying systems subject to limited communication capacity which includes measurement quantization, randomly transmission delay and data-packet dropouts, by transforming quantization effects into norm-bounded uncertainties and absorbing time-delay and packet loss variables into the stochastic matrices of the system's representation, robust filtering and variance-constraint filtering, respectively, are proposed.As wired communication network is replaced by wireless communications network rapidly, multi-packet loss problem is particularly prominent. For discrete time-varying uncertainty system with multiple packet dropouts, based on transforming the consecutive packet losses rate into a stochastic parameter in the system's representation, robust filtering and variance-constraint filtering, respectively, are proposed.