Distributed Nonlinear Fusion UKF
Zhongyao Hu, Bo Chen, Rusheng Wang, Zhen Hong, Qiuxia Chen · 2020
This paper is concerned with the distributed fusion estimation problem for nonlinear multi-sensor systems. As is well known, the weighted fusion methods minimize the trace of fusion estimate error covariance, and it can be divided into three categories: scalar weighting, diagonal matrix weighting and matrix weighting, in which the matrix weighting has the highest accuracy. However, the matrix weighting needs to know the cross covariances between the local estimates, which cannot be easily and precisely calculated in the nonlinear systems. To overcome this shortage, this paper proposes a novel method to calculate the cross covariances between the local unscented Kalman filters based on statistical linear regression, and the corresponding distributed nonlinear matrix weighting fusion criterion is developed. Finally, a target tracking system is employed to show the advantages and the effectiveness of the proposed method.