Research On State Estimation Fusion Method For Observed Delay System

The fusion theory of state estimation for multi-sensor systems has been widely used in military and civilian fields.The fusion of estimation information can make full use of the observation information from different sensors,so as to obtain the best description of the state of the system,improving the reliability of the system.However,for the delayed of observation system working in a complex environment,owing to the limitations of the communication network bandwidth,the complex noise correlation,the mismatch of the state space model,and the asynchronous multi-sensor observations,etc.,it has brought a series of new problems and difficulties.This paper is based on the observation of randomly delayed processing,noise correlation decoupling,system model uncertainty processing and multiobservation delayed characteristic inconsistent processing,combined with Gaussian filtering,strong tracking filtering,sequential filtering and distributed federal filtering technology,Respectively,a series of local filtering algorithms and state estimation fusion algorithms are proposed to reduce the impact of observation information with delayed lag on the estimated performance of the fusion system.The specific research contents of this paper are summarized into the following points:Firstly,in view of the fact that the traditional filtering algorithm in the fusion system cannot simultaneously solve the state estimation problem of observations with randomly delayed and process noise and observation noise,the filter with two-step randomly delayed measurement and correlated noises is proposed.At first,under the framework of a Gaussian filter,by extending the state and introducing an orthogonal transformation matrix,a Gaussian filter formula with two-step randomly delayed measurement and correlated noises is derived.Then,the proposed algorithm is realized by first-order linearization.At last,simulation experiments prove the effectiveness and superiority of the proposed algorithm.Secondly,aiming at the problem of inaccurate system model in the design process of the local filter algorithm in the fusion system,the accuracy of state estimation is reduced,a strong tracking filter with two-step randomly delayed measurement and correlated noises is proposed.At first,the operating mechanism of the standard strong tracking filter algorithm is analyzed.Then,the principle of extended orthogonality is given.The fading factor is calculated and introduced into the extended state prediction covariance matrix of the system to realize real-time online adjustment of the filter,forcing the residual of the filter output.The difference sequences remain orthogonal to each other at different times.At last,simulation experiments verify that the proposed algorithm can further improve the estimation accuracy and has more robust.Thirdly,research on the fusion method of state estimation for fusion systems,and a strong tracking sequential federated filter algorithm for adaptive information distribution is proposed.At first,the standard federal filtering algorithm is derived and analyzed.Then,for the inconsistency of the delayed characteristics of the observation values of the multi-sensors,the measurement update part of the proposed algorithm is completed sequentially using the observation values,and the attenuation information is used to complete adaptive distribution in the local filter to get the final estimated fusion value.At last,simulations verify that the proposed fusion algorithm can accurately estimate the state of the system.