非均匀采样系统的一种递推辨识方法

在非均匀采样系统辨识方法中, 通常利用重采样、数值积分等方法来处理非均匀采样数据, 所用模型多为连续有理分式传递函数, 在递推形式下非均匀采样对象又常局限于``数据缺失'的情况. 本文研究更为一般的异步非均匀采样的多变量系统, 采用连续时间状态空间模型描述, 推导了模型参数、参数梯度和系统状态之间的相互递推关系, 构成一种可变迭代间隔的递推辨识算法, 在每次输出采样点上仅更新模型中受当前采样数据影响的参数. 这种辨识方法可以适用于任意非均匀采样系统, 多采样率系统也可作为一种特例适用于本算法. 仿真结果表明, 所提的方法是可行有效的.

A Recursive Identification Method for Non-uniformly Sampled Systems

In the non-uniformly sampled system identification, non-uniform data are usually preprocessed first, by e.g. resampling or numeric integration. Rational models of continuous fractional transfer function are adopted in most cases. Moreover, in the existing iterative algorithms the ``data missing' cases are mostly considered. In this paper, a general recursive identification method is developed for multivariate systems with asynchronous non-uniform samples. Continuous state space model is adopted in the identification method, where three iterative forms, namely the model parameters, their gradient, and the system state, are derived through the afore-defined mathematical model. Hence, a recursive identification with varying iterative interval is proposed: part of the whole parameter set takes the recursion with only those involved parameters being updated. This method can handle arbitrarily sampled non-uniform data, so can immediately be applied to multirate systems, which are special cases of non-uniformly sampled systems. The simulation results illustrate the feasibility and effectiveness of the proposed method.