广义系统信息融合稳态与自校正满阶Kalman滤波器

Information fusion steady-state and self-tuning full-order Kalman filters for descriptor systems

中文关键词  广义系统  两段融合  稳态满阶滤波器  互协方差  自校正滤波器

英文关键词  descriptor system  two-stage fusion  steady-state full-order filter  cross-covariance  self-tuning filter

中文摘要

      基于线性最小方差标量加权融合算法和射影理论, 对带多个传感器和带相关噪声的广义系统, 提出了分布式标量加权融合稳态满阶Kalman滤波器. 推得了任两个传感器子系统之间的稳态满阶滤波误差互协方差阵, 其解可任选初值离线迭代计算. 所提出的稳态融合滤波器避免了每时刻计算协方差阵和融合权重, 减小了在线计算负担. 当系统含有未知模型参数时, 基于递推增广最小二乘算法和标量加权融合算法, 提出了一种两段融合自校正状态滤波器. 其中第1段融合获得未知参数的融合估计; 第2段融合获得分布式自校正融合状态滤波器. 与局部估计和加权平均融合估计相比, 所提出的标量加权融合参数估计和自校正状态估计都具有更高的精度. 仿真研究验证了其有效性.

英文摘要

      Based on the fusion algorithm weighted by scalars in the linear minimum variance sense and the projection theory, a distributed fusion steady-state full-order Kalman filter weighted by scalars is presented for descriptor systems with multiple sensors and correlated noises. The cross-covariance matrix of steady-state full-order filtering errors between any two sensor subsystems is derived. The solution can be computed by iteration with any initial value off-line. The proposed steady-state fusion filter avoids computing covariance matrices and fusion weights at each time step, so the online computational burden can be reduced. When the unknown model parameters are involved in the system, a two-stage fusion self-tuning state filter is presented based on the recursive extended least squares algorithm and fusion algorithm weighted by scalars. The first-stage fusion is to obtain the fusion estimate of the unknown parameters. The second-stage fusion is to obtain the distributed self-tuning fusion state filter. Compared with local estimates and weighted-average fusion estimate, the presented scalar-weighted fusion estimates for parameters and self-tuning estimate for state both have better accuracy. Simulation research validates the effectiveness.