| |
| Identification of switching autoregressive exogenous (SARX) systemsis discussed in this paper. On the basis of a relation between theKalman filter and recursive least squares (RLS) estimation, it isshown that when external noises are white and Gaussian, then,some stochastic processes can be constructed whichare white if and only if experimental data are generated by the samesub-ARX model. Based on this observation, a method is developed toidentify switching times of a SARX system, and a procedure formerging experimental data generated by the same sub-model isobtained. Using these merged data, estimation of sub-models can becarried out by resorting to standard linear identificationtechniques. As a main feature, the obtained SARX systemidentification algorithm depends neither on the number of sub-models nor onthe order of each sub-model. Some numericalexperimental results are also included to illustrate theeffectiveness of the proposed algorithm. |
|
| Keywords:Hybrid System Identification; Kalman Filtering; Recursive LeastSquares Estimation; Switching Time |
|
