Smoothing Algorithm for Multi-sensor Fusion with General Correlated Noises
一般相关噪声下多传感器平滑融合算法

In view of multi-sensor fusion estimation performance for maneuvering target tracking, a new smoothing-fusion algorithm is proposed for discrete-time linear system with general correlated measurement and process noises. The correlations between the errors are calculated precisely by analysis of the error transmission property. Based on the linear unbiased minimum variance estimation theory, the new algorithm estimates the system states recursively by using centralized expanding-dimension method with all measurements in the given interval. Compared with the uncorrelated or partially correlated Kalman smoothing-fusion algorithm, the new fixed-interval smoothing-fusion algorithm is superior under the hypothesis of Gauss distribution of noise, and the fixed-lag algorithm is suboptimal. Simulation results verified the superiority of the new proposed algorithm in the general correlated noises. It was also shown that its improvement of the tracking performance increased with the increasing of the correlation coefficient.