Weighted fusion robust steady‐state estimators for multisensor networked systems with one‐step random delay and inconsecutive packet dropouts

Abstract

Summary In this paper, the weighted fusion robust steady‐state Kalman filtering problem is studied for a class of multisensor networked systems with mixed uncertainties. The uncertainties include same multiplicative noises in system parameter matrices, uncertain noise variances, as well as the one‐step random delay and inconsecutive packet dropouts, which modeled by sequences of Bernoulli variables with different probabilities. By defining a new observation vector and applying the augmented method, the system under study is converted into one with only uncertain noise variances. The sufficient conditions for the existence of steady‐state estimators are given. According to the minimax robust estimation principle, based on the worst‐case subsystems with conservative upper bounds of uncertain noise variances, the robust local steady‐state Kalman estimators (predictor, filter, and smoother) are proposed. Applying the optimal fusion algorithm weighted by matrices and the covariance intersection fusion algorithm, the two kinds of robust fusion steady‐state Kalman estimators are derived in a unified framework. The robustness of the proposed fusion estimators is proved by applying the permutation matrices and the global Lyapunov equations method, such that, for all admissible uncertainties, the actual steady‐state estimation error variances of the estimators are guaranteed to have the corresponding minimal upper bounds. The accuracy relations among the robust local and fusion steady‐state Kalman estimators are proved. An example with application to autoregressive moving average signal processing is proposed, which shows that the robust local and fusion signal estimation problems can be solved by the state estimation problems. Simulation example verifies the effectiveness and correctness of the proposed results.

Fusion steady‐state robust filtering for uncertain multisensor networked systems with application to autoregressive moving average signal estimates

“…Furthermore, from Lemma 2 we have that X ai = X ai and X aij = X aij . Comparing (31) and (32) yields Q w fi = Q w fi , from (33) we have that Q w fij = Q w fij , therefore, from (50) we obtain that Q w g = Q w g . From (7) and ( 8) we have that (37) and (38) (39) we have that…”

“…It is different from the proof method given by References 30, where the permutation matrices are not used. It is also different from References 9,31, where the permutation matrices are only used to prove the robustness of distributed weighted fusion robust estimators. An application example for ARMA signal estimation is put forward, and the robust CF steady‐state signal estimators are derived based on the robust fusion steady‐state state estimators.…”

Robust steady‐state matrix‐weighted fusion filtering for multisensor multichannel autoregressive signal with multiple uncertainties

“…7 However, this hypothesis may not always true in real situations since uncertain perturbations, stochastic parameters, unmodeled dynamics, etc. [8][9][10][11][12][13][14][15][16] If the uncertainties exist in the dynamic system model, then the result of standard Kalman filter will not achieve the expectation. This issue can be solved by designing a robust Kalman filter (RKF).…”

“…As is generally known, an important hypothesis in Kalman filtering theory is that the state‐space model under study is accurate 7 . However, this hypothesis may not always true in real situations since uncertain perturbations, stochastic parameters, unmodeled dynamics, etc 8–16 . If the uncertainties exist in the dynamic system model, then the result of standard Kalman filter will not achieve the expectation.…”

Event‐triggered Kalman consensus filter for sensor networks with intermittent observations

“…On the other hand, it is common in practice that observations of sensors are not consecutive, due to network‐induced packet loss, 28 random sensor failure, 29,30 and so forth. And the filtering problems for sensor networks with intermittent observations have attracted many attentions 31‐35 .…”