MBDLM中卡尔曼滤波算法的改进研究

Improvement of Kalman Filter Algorithm in MBDLM

根据多变量Ｂａｙｅｓｉａｎ动态线性模型（简记为ＭＢＤＬＭ）在递推过程中出现的方差阵不正定、精底低以及滤波发散等情况进行分析，对原有卡尔曼滤波算法加以改进，给出了比较适用的递推算法。

This paper mainly studies the updating recurrence algorithms of multivariate Bayesian dynamic linear model (MBDLM).In practice we may fale some ofstacles, such as computation fails to assure symmetry and positive definiteness, low precision and filter divergence. This thesis provides more suitfal updating algorithms according to different proflems on the fasis of original formulas.