Distributed State Estimation Of Uncertain Observing System

In the practice, there exist many stochastic systems with uncertain observations. For example, due to the effects of environment and unreliable communication networks, data loss may occure in the networked control systems and sensor network. Thus the uncertainty appears in measurement equation. In this paper, we study the systems with data loss which can be described by a Bernoulli distributed random variable. For multi-sensor discrete-time stochastic linear systems with uncertain observations, we study the distributed information fusion estimation algorithms. The main contents include the distributed information fusion white noise estimators, state estimators and state estimators with unknown disturbances in sensor observations.Based on the innovation analysis method, the linear minimum variance optimal white noise estimators, including input white noise estimators and observation white noise estimators, are presented for single-sensor discrete-time stochastic systems with uncertain observations and correlated noises. And the steady-state white noise estimators are given for the stable systems. Furthermore, the unified asymptotically stable Wiener white noise estimators are obtained based on the steady-state Kalman estimators. Finally, based on the distributed optimal weighted fusion estimation algorithms in the linear minimum variance sense, the distributed weighted fusion input white noise estimators are proposed for multi-sensor systems with uncertain observations. The estimation error cross-covariance matrix between any two sensor subsystems is derived. To compare with the centralized fusion estimators on accuracy and computational cost, the centralized fusion estimators are also given.Based on the innovation analysis method, the linear minimum variance optimal and steady-state state estimators are presented for single-sensor discrete-time stochastic linear systems with uncertain observations and correlated noises. Furthermore, the unified asymptotically stable Wiener state estimators are obtained based on the steady- state Kalman filter. In addition, the state estimators are developed for a single-sensor stochastic system with uncertain observations and correlated noises both at the same time and the neighborhood time. Furthermore, the cross-covariance matrix between any two sensor subsystems is derived for multi-sensor systems. Then, the distributed weighted fusion state estimators are obtained by using the distributed weighted information fusion estimation algorithms. And the comparison of accuracy between the distributed fusion and the centralized fusion is done.Firstly, for stochastic systems with uncertain observations and sensor unknown disturbance inputs, the linear unbiased minimum variance state estimators, including a priori filter and a posterior fitler, are presented based on single sensor, which are independent of unknown inputs. Then, the filtering error cross-covariance matrix between any two sensor subsystems is derived for multi-sensor systems. Finally, the distributed weighted fusion state estimators are obtained for multi-sensor systems based on the distributed weighted fusion estimation algorithms. And the comparison research is made between the distributed fusion estimation and augmented centralized fusion estimation.