Fusion Estimation Of Stochastic Systems With Time-Correlated Nois

Multi-sensor systems with time-correlated noises are widely used in aerospace engineering,communication systems and tracking systems.Based on projection theory,this paper investigates the globally optimal fusion estimation for systems with time-correlated multiplicative noises and the distributed fusion estimation for systems with time-correlated additive noises and random packet dropouts.The main content is as follows:For multi-sensor systems with time-correlated multiplicative noises,where the time-correlated multiplicative noise is described by a first-order Gaussian-Markov process.First,we introduce a set of virtual state variables,and then the multiplicative noises are transformed into additive noises.Next,we propose the recursive filters for the virtual states,the globally optimal centralized fusion filter based on batch processing and the globally optimal sequential fusion filter based on sequential processing are designed in the sense of linear minimum variance,respectively.The proposed sequential fusion filtering algorithm has the same estimation accuracy as the centralized fusion filtering algorithm.Further,compared to the centralized fusion filtering algorithm,it can reduce the computational burden.Finally,the equivalence of the estimation accuracy of the proposed centralized fusion filter and sequential fusion filter by the mathematical induction method.For multi-sensor uncertain systems with time-correlated additive noises and random packet dropouts,where the time-correlated process and measurement noises are described by the first-order Gaussian-Markov process.The uncertainties of the model parameters are described by correlated white multiplicative noises.When the sensor measurement signal is lost over network transmission,its one-step prediction is used for compensation.Based on the local estimators proposed in the existing literature,the cross-covariance matrices between any two local filtering errors are derived.Further,three distributed weighted fusion filters are proposed based on the distributed weighted fusion algorithm in the linear minimum variance sense.