基于贝叶斯估计的平滑算法

基于贝叶斯估计的平滑算法是在事后处理的情况下,依据过去直至现在的观测值去估计过去的历史状态,以有效提高精度。而滤波是依据过去直至现在的观测值去估计现在的状态。从理论上讲,由于平滑用了所求估计时刻之后的观测值,平滑算法应比滤波优异一些。本文设计了2个实验仿真计算,在逼近效果和RMS等方面分别与Kalman滤波和双向滤波加权平均进行了比较。经实验证明,无论是逼近效果还是RMS,平滑算法都要更优一些。

A study on smoothing algorithm based on Bayes estimation

A smoothing algorithm is based on Bayesian theory,which should be derived using the Kalman filtering.In the case of post-process,the precision can be improved by estimation using observations from past to future.The filter algorithm uses the observations from past to now.So,in theory,the smoothing algorithm is better than the filter algorithm in precision.Two simulation experiments were taken in the paper,and the results showed that the smoothing algorithm is more effective than Kalman filtering and a weighted combination of forward and backward filtering.