Multisensor Accompanied Shaped Self-tuning Information Fusion Filter And Its Application

Multisensor information fusion (Multisource information fusion or Multisensor data fusion) is a new branch of science, which involves the automatic control theory, computer science, mathematics, military science, artificial intelligence, communication technology, electronics technology, etc. It has been applied to many high-tech fields including defense, guidance, target tracking, communication, signal processing, GPS positioning and robots, etc. Multisensor information fusion detects the same objects by multisensor, avoiding the limitation of single sensor, and it can provide more general and accurate information. As a new technology of the multisource data processing, the multisensor information fusion can fuse multisensor detection data from one object or local estimations of all sensors, and generate more accurate estimation than single source.For the multisensor systems with companion form, and for the parallel and series multisensor systems with companion form, when the model parameters and noise variances are unknown, the online information fusion estimators of unknown AR parameters for the equivalent Autoregressive Moving Average (ARMA) and Autoregressive (AR) models parameters are presented based on the Recursive Instrumental Variable (RIV) method, and the online information fusion estimators of the noise variances are presented based on the correlation method. And the online information fusion estimators of the Moving Average (MA) model parameters of the ARMA model are presented based on the G-W algorithm with dead band. They have consistency.The self-tuning information fusion Kalman filter and Wiener filter are presented by substituting the online information fusion estimators of unknown model parameters and noise variances into the optimal fusion Kalman filter and Wiener filter weighted by scalar for components. Their accuracy are higher than that of each local self-tuning filters.By the dynamic error system analysis (DESA) method, it is proved that the self-tuning fused Kalman filter converges to the optimal fused Kalman filter in a realization. So it has the asymptotic optimality.Applying the proposed results to the multisensor AR and ARMA signals with unknown model parameters and noise variances, based on the transformation of the ARMA model to state space model, the signal estimation problem can be transformed to the state estimation problem. Substituting the online fused estimators of unknown model parameters and noise variances into the optimal fused Wiener filter, the self-tuning fusion Wiener filter is presented. By the DESA method, it is proved that the self-tuning fused filter converges to the optimal fused filter in a realization, so it has the asymptotic optimality.Many simulation examples show proposed results'correctness and effectiveness.