Event-Triggered Distributed Filtering For Several Classes Of Time-Varying Stochastic Nonlinear Systems

The distributed filtering problem based on sensor network transmission is one of the main research issues in the control field.The rapid development of science and technology not only improves the sensor performance and expands the application scope of sensor network,but also brings new problems and challenges to the networked system.For example,the service life of the network will be reduced by frequent data transferring around network nodes,packet losses,transmission delay and other network induced phenomena will not only degrade the system performance,but also diverge the filtering error.Therefore,the event-triggered mechanism which can increase network resource use efficiency and lighten network burden is significance to ensure the performance of networked filtering algorithm.For the nonlinear stochastic system based on sensor network transmission,exploring the distributed filtering problem on the basis of event-triggered mechanism is an urgent and widely applied research topic.The aim of this thesis is to study the event-triggered distributed filtering problem for several kinds of nonlinear stochastic systems based on sensor network transmission.According to the Kalman filter theory,new distributed filters are designed,and the boundedness of filtering error is discussed to explore the influence of different event-triggered mechanism on filter design and filtering effect.Firstly,the distributed filter design problem is investigated for a class of time-varying nonlinear systems with randomly switching nonlinearities and redundant channels.The randomly switching nonlinearities and randomly missing measurement in channels are modeled by two series of mutually independent random variables which obey Bernoulli distributions.The distributed filter is designed for discrete stochastic time-varying nonlinear systems with dynamic event-triggered mechanism.Based on the Riccati-like difference equation approach,the recursive expressions of upper bound for error covariance are obtained.An appropriate filter gain is constructed to minimize the trace of the upper bound.Under an assumption that the system parameters are norm-bounded and system state is energy bounded,the sufficient condition is established to ensure the mean-square boundedness of the filtering error.Secondly,the resilient distributed filter is designed for discrete time-varying nonlinear systems with correlated noise.The statistical properties of noise are given and process noise and measurement noise are one-step autocorrelation,the process noise is cross-correlated with the measurement noise in two steps.Considering the effect of event-triggered mechanism and filter gain perturbation,a resilient distributed filter is designed.Based on the solutions to the matrix difference equations,the minimum upper bound of the filtering error covariance is obtained.A sufficient criterion is proposed to verify the exponentially mean-square boundedness of the filtering error.Thirdly,the distributed filter design problem is discussed for time-delay nonlinear systems with randomly occurring sensor saturation.The randomly occurring sensor saturation is described by a series of mutually independent random variables.A mathematical model is established to describe the dynamic event-triggered mechanism.Considering the saturation error and event-triggering error at the same time,an upper bound of the filtering error covariance is obtained by designing a recursively distributed filter.Under the assumptions that similar to those in part one,the sufficient condition of mean-square bounded filtering error is given by using stochastic analysis technique.Fourthly,the problem of distributed filter design is studied for multi-rate systems with randomly occurring transmission delay.By introducing Kronecker-?function,a concise model with randomly occurring transmission delay is established.The multi-rate system is converted to a single rate system by lifting technique.A new time delay compensation method is proposed for the case that the sampling interval of the lifted discrete systems may be different,and then an appropriate innovation representation is given.A recursive distributed filter is designed by considering the event-triggered mechanism's scheduling effect on innovation.Based on Riccati-like difference equation method and matrix theoretical knowledge,the upper bound of filtering error covariance is proposed and optimized,and an appropriate filter gain expression is designed.