Unified optimal linear estimation fusion. I. Unified models and fusion rules

Abstract

This paper deals with data fusion for the purpose of estimation. Three fusion architectures are considered: centralized, distributed, and hybrid. A unified linear model and general framework for these three architectures are established. Optimal fusion rules in the sense of best linear unbiased estimation (BLUE), weighted least squares (WLS), and their generalized versions are presented for cases with either complete, incomplete, or no prior information. These rules are much more general and flexible than previous results. For example, they are in a unified form that are optimal for all the three fusion architectures with arbitrary correlation of local estimates or observation noises across sensors or across time.. They are also in explicit forms convenient for implementation. The relationships among these rules are also presented.

Optimal linear estimation fusion-part VII: dynamic systems

“…It will never be worse than and may outperform the best recursive BLUE without prior It follows from Theorem 1 of [12,11] that the optimal recursive BLUE with prior (LMMSE) fuser is given bŷ…”

Distributed Kalman Filtering

“…Other related publications cited in the Table 2.7 are [338]- [339], [337], [332], [333,334], [37], [46], [47], [48], [48], [60], [59,58], [57], [56], [55], [54], [53], [52]. Other related publications cited in the Table 2.8 are [67], [68], [91], [93,94], [95], [115], [116], [117], [124], [125], [127] and [128].…”

Distributed Weighting Fusion and Covariance Intersection Fusion Kalman Smoother for Systems with Colored Measurement Noises

“…Unified fusion rules for the optimal linear estimation fusion and several distributed weighting state fusers were presented in [2][3][4][5], where the three distributed weighting fusers have the accuracy relations: the accuracy of the fuser weighted by matrices is higher than that of the fuser weighted by scalars, and the accuracy of the fuser weighted by diagonal matrices is between of them. However, all of the above weighting fusers have the limitation that in order to compute the optimal weights, the computation of the cross-covariances between the local estimation errors is required, while the cross-covariances are usually unknown [6] or their computation is very complex [7] in many applications.…”