考察实际中常见的三类典型随机非线性系统(即Wiener、Hammerstein和NARX系统)的辨识,首先概述了现有的递推和非递推辨识算法,然后介绍这三类系统的一个统一辨识框架: 利用系统所确定的过程的马氏性及混合型,将辨识转化为求函数零点的问题,基于扩张截尾的随机逼近算法,得到了递推、强一致的辨识结果,并给出了数值模拟验证辨识算法收敛到真值．

Identification of several classes of stochastic nonlinear systems, i.e., the Wiener system, the Hammerstein system and the nonlinear ARX system, is considered. First, existing recursive and nonrecursive algorithms for identifying these systems are briefly summarized. Then, a unified framework to recursively identify these systems is introduced. Based on the Markov chains and mixing properties connected with these systems, the identification is transformed into root searching problems. Finally, identification algorithms based on stochastic approximation with expanding truncations are introduced
and strong consistency of estimates is established. The theoretical results are verified by simulation examples.