两级交互式多模型算法

A TWO-STAGE INTERACTING MULTIPLE MODEL ALGORITHM

在混合估计中 ,交互式多模型滤波器 ( IMM—— Interacting Multiple Model)以其优越的性能而受到广泛的研究 .由于马尔可夫参数的限定 ,交互式多模型在模型数较多时会出现精度下降 ,从而限制了它在高维参数空间建模的有效性 .利用模型集的概念 ,首次提出了双马氏过程的模型切换假设 ,从而构造出一种两级交互式多模型滤波器 .通过辨识系统噪声的多个统计参数比较了两级交互式多模型滤波器与常规交互式多模型滤波器 .结果表明 :对于大信噪比信源 (即小的量测噪声 ) ,两级交互式多模型滤波器与常规交互式多模型滤波器性能基本相当 ;而对于小信噪比信源 (即大的量测噪声 ) 。

In hybrid estimation, interacting multiple model (IMM) estimator is one of the most cost effective schemes. But IMM will degrade if too many models are chosen. This is due to the limit of Markov parameters. In this paper, a dual Markov switching process is proposed. Then a two stage IMM is given. Simulations show that the new algorithm is better than IMM for large measurement noise and almost the same as IMM for small measurement noise when two abruptly changing statistic parameters of process noise are identified.