引言线性滤波的典型问题可描述为:对t∈[0,T]观测到 y(t)=s(t)+n(t) 其中n(t)是功率谱密度为N_o的白色高斯噪声,s(t)是有待恢复和估计的未知信号。在实际中s(t)有不止一种的描述方式,例如,s(t)可以是一个多项式 s(t)=a+bt+ct~2 其中a,b,c是未知参数。自卡尔曼滤波问世以来,应用最广泛的方式是把s(t)描述为某个动态系统在白噪声激励下的输出,例 Introduction The typical problem of linear filtering can be described as: y (t) = s (t) + n (t) where t (t) is the white Gaussian noise with power spectral density N_o, s (t) is an unknown signal to be recovered and estimated. For example, s (t) can be a polynomial s (t) = a + bt + ct ~ 2 where a, b and c are unknown parameters. Since the advent of Kalman filtering, the most widely used method is to describe s (t) as the output of a dynamic system under white noise excitation, for example