Fusion Filtering With Unknown Interference And Packet Loss Compensation Networked Systems

With the development of network technology and the further research of control theory,control and estimation problems of networked systems have been the current research hotspots.They are widely used in the national economic construction due to low cost,high efficiency and easy remote operation of networked systems.On account of increasing scale of networks and raising complexity of systems,control and estimation for networked systems are becoming difficult and challenging.Because of the limited carrying capacity of the server and network bandwidth,the state information or control command data are inevitably lost during network transmission.In addition,the network channel is also easily affected by various external interferences.Except for network channel factors,the uncertainties of multiplicative noises often make the influences to states and observations of systems.Therefore,considering the above problems,this paper studies state fusion estimation problems for multi-sensor networked systems with unknown communication interferences,compensations of packet losses and multiplicative noise uncertainties.The main contents are as follows:For the multi-sensor networked systems with unknown communication interferences and packet losses,when observations are missing,a new observation model is established by taking the predictions of the current missing observations as compensations.Based on the linear unbiased minimum variance estimation criterion,the distributed and centralized fusion state predictors and filters are designed.The cross-covariance matrices of any two local prediction or filtering errors of subsystems are derived.The stability of the filters is analyzed.Finally,on the basis of state estimation,estimators for the unknown interferences are proposed.For multi-sensor networked random systems with multiplicative noise uncertainties consisting in states and observations simultaneously,missing observations and unknown communication interferences,based on new models of observations established by using the compensations of the predictors for packet losses,the distributed and centralized fusion state predictors and filters independent of unknown communication interferences are designed.The cross-covariance matrices between any two local predictor and filter errors for subsystems are derived.The accuracy of estimation algorithms with and without compensations of the predictors for packet losses is compared.At last,the estimators for the unknown interferences are given.