基于模型空间分解的交互式多模型算法

A Two-Level Interacting Multiple Model Algorithm

由于马尔可夫参数的限定 ,交互式多模型的估计精度会在模型数过多时下降。这限制了它在高维参数空间的应用。通过将交互式多模型建模空间分解 ,构造出一种两级交互式多模型算法。并通过辨识系统噪声的多个统计参数 ,比较了新算法与常规交互式多模型滤波器 。

Due to the limitation of Markov parameters,standard interacting multiple model (IMM) algorithm's handing capability deteriorates if too many models are chosen. In such a case, we propose Two Level IMM. In Two Level IMM, we divide model set including many models into several model subsets. We assume that the transition of one model subset to another belongs to one Markov chain. We also assume that the transition of models is another Markov chain on the condition that the transition of the corresponding model subsets happens. In the simulation we identify process noise with two abruptly changing statistical parameters. Simulation results, shown in Figs.1 through 4, show that Two Level IMM is better than standard IMM. The computational burden of Two Level IMM is about the same as that of standard IMM.