| 摘要: |
| 辨识切换线性系统的主要难点在于参数估计问题与子系统划分问题耦合。针对该问题，利用卡尔曼滤波与递推扩展最小二乘的联系，证明当所辨识的带外源输入的自回归滑动平均（ARMAX）切换系统在满足严正实条件下，当且仅当输入输出数据来源于同一个子ARMA系统时所构造的新息序列具有白色性。基于此,提出了一种切换ARMAX系统切换时刻检测算法。仿真计算结果验证了所提算法的有效性。 |
| 关键词: 非线性系统辨识；ARMAX系统 切换时刻检测；新息过程 |
| DOI：10.3969/j.issn.1001-893x.2017.09.006 |
|
| 基金项目: |
|
| Switching time estimation for switching ARMAX systems |
| CHEN Jun,XIONG Jie |
| (Southwest China Institute of Electronic Technology，Chengdu 610036，China) |
| Abstract: |
| The most difficult and essential part in switching linear systems identification is to decouple the date classification and the sub-model parameter identification from experimental data. Against this issue, on the basis of a relation between the Kalman filter and recursive extended least squares (RELS) estimates, it is shown that when external disturbances are modeled as Moving Average (MA) models which satisfy strictly positive realness, then, there exists an innovation process which is white if and only if experimental data is generated by the same Auto-Regressive Moving Average with eXogenous input (ARMAX) model. Based on this observation, a procedure is developed to identify the switching time of a switching ARMAX system. Numerical experiment results are also included to illustrate its effectiveness. |
| Key words: nonlinear system identification ARMAX system switching time estimation innovation process |
