广义系统最优与自校正信息融合滤波器  

Optimal and Self-tuning Information Fusion Filter for Descriptor Systems

作　　者：马静[1] 曲仲田[1,2] 孙书利[1] 

机构地区：[1]黑龙江大学自动化系 [2]阿城市金京热电厂,阿城150300

出　　处：《科学技术与工程》2006年第12期1591-1595,共5页Science Technology and Engineering

基　　金：国家自然科学基金(60504034);黑龙江省青年基金(QC04A01);黑龙江大学电子工程省重点实验室资助

摘　　要：对带多个传感器广义离散随机线性系统,利用典范型分解,基于线性最小方差各分量按标量加权融合算法,给出了多传感器分布式最优分量融合降阶滤波器,它要求并行计算一系列标量权重。推得了任两个传感器子系统之间的滤波误差互协方差阵的计算公式。同时当系统含有未知噪声统计信息时,基于相关函数又给出了分布式自校正分量融合降阶滤波器。与各局部估计以及状态向量按标量加权融合估计相比,分量融合滤波具有更高的精度。仿真研究验证了其有效性。Using a decomposition in canonical form, a multi-sensor distributed optimal fusion reduced-order filter for each state component is proposed based on the component fusion algorithm weighted by scalars in the linear minimum variance sense for descriptor discrete stochastic linear systems with multiple sensors. It requires in parallel the calculating of a series of scalar weights. The computation formula for the filtering error crosscovariance matrix between any two subsystems is derived. In addition, a decentralized self-tuning fusion reduced-order filter for each state component is also given based on correlation function when the noise statistic information is unknown. Compared with all local filters and the fusion filter weighted by scalars for the state vector, the component fusion filter weighted by scalars has higher precision. The simulation research shows its effectiveness.

关 键 词：广义系统 最优信息融合 自校正信息融合 互协方差阵 降阶滤波器 

分 类 号：O211.64[理学—概率论与数理统计]