一种计及观测噪音的分系统递推辨识方法

本文提出一种可以处理带测量噪音的多输入多输出线性系统递推参数辨识方法。目前多输入多输出系统的在线参数估计方法或者需要将模型正规化,或者要做大量高维矩阵求逆运算,不仅运算时间长,而且很可能由于矩阵逆的不存在而导致估计失败。本文方法将多输入多输出系统参数估计问题分解为若干步多输入单输出系统参数估计问题,把高维矩阵求逆化为若干步求倒数运算,从而避免了可能由此而导致的估计失败。本方法直接估计线性系统参数,不需将模型正规化,从而节省了实际运算时间。最后,根据有观测噪音的试飞数据对本文方法进行之检验。

A PARAMETERS IDENTIFICATION METHOD IN MEASUREMENT NOISE WITH DECOMPOSITION SYSTEM

This paper presents an recursive algorithm for linearized Multi-Input-Multi -Output system parameters identification in measurement noise. Generally, recursive parameter identification methods either have to standarize the model of the objects or have to calculate many times for inversing the matrics, not only taking a lot of operational time but also tending to diverge when the inverse of the matrix does not exist. This method decomposesthe Multi-Input-Multi-Output linearized system parameters identification into a number of Multi-Input-Single-Output system parameter identifications, anP then to inverse the matrix becomes a number of times of invering the scalar quantity. That will be able to avoid divergency. And the method directly estimates parameters without standarizing the model, so that saves the operational time in practice. In the end, the method is checked with flight test data in measurement noise.