Variable step size predictor design for a class of linear discrete-time censored system
Variable step size predictor design for a class of linear discrete-time censored system
复制标题
DOI：
10.3934/math.2021614
复制
发表时间：
2021
期刊：
影响因子：
2.2
通讯作者：
Zhifang Li;Huihong Zhao;Hailong Meng;Yong Chen
中科院分区：
文献类型：
--
作者：
Zhifang Li;Huihong Zhao;Hailong Meng;Yong Chen
针对一类线性离散时间删失系统，提出了一种新的变步长预测器设计方法。我们将删失系统分为两部分。其中一部分的系统量测方程不包含截尾数据，另一部分的系统量测方程是截尾信号。对于正常情况，我们采用卡尔曼滤波技术设计一步预测器。对于观测方程被删失的情形，根据删失数据长度确定预测器步长，并分别应用最小误差方差迹、投影公式和经验分析方法给出了具有明显误差情形预测器的增益补偿参数矩阵β（\mathfrak{s}）。最后通过仿真实例表明，基于经验分析的变步长预测器具有较好的估计性能。
We propose a novel variable step size predictor design method for a class of linear discrete-time censored system. We divide the censored system into two parts. The system measurement equation in one part doesn't contain the censored data, and the system measurement equation in the other part is the censored signal. For the normal one, we use the Kalman filtering technology to design one-step predictor. For the one that the measurement equation is censored, we determine the predictor step size according to the censored data length and give the gain compensation parameter matrix $β(\mathfrak{s})$ for the case predictor with obvious errors applying the minimum error variance trace, projection formula, and empirical analysis, respectively. Finally, a simulation example shows that the variable step size predictor based on empirical analysis has better estimation performance.