Research On Information Fusion Estimation For Several Classes Of The Networked Singular Systems

As we all know,the singular systems are one class of power systems,which have more extensive description forms than the normal systems.The singular systems have complexity and novelty in structure,and have more widespread applications in practical fields,such as the economic systems,navigation systems,electrical circuit network systems and so on.Accordingly,the singular systems have gained considerable attention.Owing to the estimation accuracy of the single sensor would be affected by severe interferences from other external factors and its own influence,so the multi-sensor information fusion algorithm came into being.However,the network-induced phenomenon is prone to occur in the network environment due to imperfect communications,which would degrade the performance of systems.Consequently,it is significant and challenging for the networked singular systems to investigate the information fusion estimation problem with multiple sensors.In this thesis,we aim to propose the reduced-order or the full-order information fusion estimation algorithms for several classes of networked discrete nonlinear singular systems with multiple sensors.The main works are summarized as follows.1.The information fusion filtering problem is investigated for multi-sensor singular systems with fading measurements and stochastic nonlinearity.The phenomenon of the multi-channel fading measurements is molded by a diagonal matrix with diagonal elements over the interval [0,1].The singular systems can be converted into the reduced-order nonsingular systems by utilizing the restricted equivalent transformation method.Then,the local optimal filters for the reduced-order subsystems can be obtained on the basis of the optimal filtering theory.Furthermore,the distributed fusion filter(DFF)is derived,on the basis of the matrix-weighted fusion estimation algorithm(MWFEA),for the original singular systems with multiple sensors in the linear minimum variance sense.Finally,a simulation example is put forward to verify the feasibility of the proposed fusion filtering algorithm.2.The issue of the distributed and centralized fusion estimation is investigated for a class of discrete time-varying nonlinear singular systems by comprehensively taking into account missing measurements and interference of multiplicative noises.In order to cater the practical engineering,both the correlated process noises and the measurement noises are considered.The missing measurements phenomenon for different sensors is characterized by a set of Bernoulli distributed random variables.Firstly,the singular systems can be transformed the reduced-order nonsingular subsystems based on the standard singular value decomposition.Then,the local optimal filters and corresponding DFFs can be derived for reduced-order subsystems be means of the innovation analysis approach and the MWFEA.Moreover,the centralized fusion estimator can be derived on the basis of the augmented measurement model.In the end,a numerical example is used to verify the feasibility and validity of the proposed fusion algorithm.3.The issue of the fusion filtering is investigated for a class of singular systems,where the phenomena of the random transmission delays(RTDs)and packet losses are considered in discrete time-varying singular systems.The Bernoulli distributed random variables are introduced to characterize the phenomena of the RTDs and packet losses.Specifically,when the sensor measurement is lost,the one-step predictor of current sensor measurement is used to compensate for the negative effects induced by packet losses.Firstly,the singular systems are converted into the nonsingular subsystems on the basis of the nonsingular transformation.Then,the new augmented systems with stochastic parameter matrices and correlated noises are introduced based on the measurement compensation model.Subsequently,the local optimal filter can be obtained for the augmented systems by utilizing the projection theory.Then,based on the MWFEA,the DFF is derived in the sense of the linear minimum variance.Finally,a simulation example is carried out to verify the effectiveness and superiority of the proposed distributed fusion filtering algorithm.4.The design issue of the full-order recursive filter is investigated for a class of delayed nonlinear multi-sensor singular systems with packet disorders.The phenomenon of packet disorders is caused by RTDs which take place in the channel from sensor to filter,where a set of random variables obeying known probability distribution are introduced to describe the RTDs.Firstly,the singular systems are converted into the full-order nonsingular systems by utilizing the matrix transformation method.Then,the full-order local recursive filter is designed,which employs the rounding function of the mathematical expectation of the random transmission delays.Subsequently,by using the matrices inequality technique,the upper bound depended on the delay probability of the estimation error covariance is obtained and then locally minimized through appropriately designing the corresponding filter gains.Furthermore,the upper bound of the estimation error cross-covariance can be calculated.Then,the DFF is derived for singular systems by employing the MWFEA.Finally,a numerical example is put forward to verify the feasibility and validity of the proposed fusion filtering algorithm.