Covariance Cross-fusion Estimation Of Time-delay Systems

In recent years,with the deepening of the research on multi-sensor information fusion technology,the state estimation problem of time-delay systems has attracted more and more attention.In the multi-sensor system,due to the inherent reasons of the system itself or the existence of unreliable external factors such as the aging of components,and most of the time delay is inevitable.If there is a delay in the control system,it is equivalent to the introduction of additional phase shift,which will make the control system tends to be unstable and have a very negative effect on the control performance of the control system.In particular,because of the existence of time delay,the calculation of the error cross-covariance matrices between the local sensors is especially complicated when the multi-sensor information fusion is carried out.In order to reduce the complexity of the fusion algorithm,it is possible to use SCI fusion estimation algorithm,which only based on the local state estimation and its corresponding error variance matrices.The SCI fusion estimation algorithm is a conservative fusion estimation method because it avoids the calculation of the cross-covariance matrices between the local sensors,but can significantly reduce the computational burden.Based on the projective theorem and the linear minimum variance criterion,this paper uses the SCI fusion estimation algorithm to estimate the system state in the multi-time delay multi-sensor systems.The main research includes the following aspects:First of all,based on the existing local state estimators for a observation delay multi-sensor systems with uncorrelated noise,a state delay multi-sensor system with correlated noise and a state delay multi-sensor system with colored noise,the SCI fusion estimation of the system state are proposed respectively,which can avoid the calculation of the cross-covariance matrices among sensors.Compared with the distributed information fusion estimation algorithm weighted by matrices,the computational burden is reduced and the consistency fusion estimation is obtained.Secondly,by using the modern time series analysis method for the observation delay multi-sensor system with uncorrelated noise,and based on the state space model of the system,the left information decomposition and Gevers-Wouters algorithm are used to construct the ARMA innovation model.The observation delay existing in the system can be naturally embedded into the model.The non-recursive Kalman filter of the local sensor and its corresponding error variance matrices are deduced by using the projective theorem.Based on the unified steady-state white noise estimation theory,fusion Kalman filter of a observation delay multi-sensor system is designed by using the SCI fusion estimation algorithm.Combining modern time series analysis method with SCI fusion estimation algorithm,it is very important for the research of the multi-sensor systems with time delays.Finally,on the basis of the previous research,the two-level fusion estimation structure of the multi-sensor system is designed to realize the SCI fusion estimation of the system state.In the first-level fusion center,the local sensors are grouped according to the similar structure.Each group of sensors sends their observed information to the corresponding local sensor centers,and then uses the weighted observation fusion algorithm to obtain the first-level fusion estimation.In the second-level fusion center,the final fusion estimation of the system state is obtained by using the SCI fusion estimation algorithm.Compared with the single SCI fusion estimation algorithm,the two-level fusion estimation structure has higher estimation accuracy for the system state.Several simulation studies verify the validity and superiority of the SCI fusion estimation algorithm in dealing with the state estimation of multi-sensor systems with time delays.