Fusion Filtering For Stochastic Uncertain Systems With Time-correlated Nois

Multi-sensor information fusion state estimation has become an important branch of modern decision and control theory.Time-correlated noise widely appears in many engineering applications.Based on the projection theory,this paper studies the fusion filtering problem of stochastic uncertain systems with time-correlated noise by combining three methods of distributed fusion,centralized fusion and sequential fusion.The main contents are as follows:For stochastic uncertain systems with time-correlated observation noise and multiplicative white noise in the observation equation,a centralized fusion filter and a sequential fusion filter in the linear minimum variance sense are both designed by introducing the observation noise estimator.Based on mathematical induction,it is proved that the sequential fusion filter is equivalent to the centralized fusion filter under the same initial values,that is to say the proposed sequential fusion filtering algorithm also has global optimality.Further,the multiplicative noise in the equation of state is also considered for the system and the state multiplicative noise is correlated with the measurement multiplicative noise.By introducing an estimator of the product of state multiplicative noise and system state,centralized fusion filter and sequential fusion filter are designed in the linear minimum variance sense,and their accuracy equivalence property is proved.For the stochastic uncertain system with time-correlated observation noise and one-step autocorrelation process noise,the observation difference method is used to transform the original system into the one with one step state delay,one step autocorrelation and one step cross-correlation noise.The filter,predictor and smoother of local single sensor subsystem are designed.The calculation formula of cross-covariance matrices between any two local estimation errors is derived,and then the distributed fusion estimator in the linear minimum variance sense is given.At the same time,the centralized fusion estimator is derived.