随机奇异系统多传感器信息融合Kalman多步预报器

Multi-sensor Information fusion Kalman Multi-step predictors for Stochastic Singular Systems

应用Kalman滤波方法和奇异系统典范型分解,对单传感器随机奇异系统,给出了Kalman多步预报器新算法。对带多传感器随机奇异系统,基于线性最小方差标量加权融合算法,给出了具有两层融合结构的多传感器分布式最优信息融合Kalman 多步预报器。同时给出了任两个传感器之间的预报误差协方差阵的计算公式。当各传感器子系统存在稳态Kalman滤波时,又给出了稳态信息融合Kalman多步预报器。稳态权重可在各子系统达到稳态时通过一次融合计算获得,避免了每时刻计算协方差阵和融合权重,便于实时应用。仿真例子验证了其有效性。

Using Kalman filtering method and a decomposition in canonical form for singular systems, new algorithms for Kalman multi-step predictor are presented for stochastic singular systems measured by single sensor. A decentralized optimal information fusion Kalman multi-step predictor with a two-layer fusion structure is proposed based on the fusion algorithm weighted by scalars in the linear minimum variance sense for stochastic singular systems measured by multiple sensors. The computation formulas for the prediction error covariance matrix between any two subsystems are given. Furthermore, steady-state information fusion Kalman multi-step predictor is also given when steady-state Kalman filter exists for each sensor systems. The steady-state weights can be obtained only by one time fusing after all subsystems enter steady states. It avoids computing covariance matrices and fusion weights at each time step. It is convenient to apply in real time. Simulation example proves its effectiveness.