<p>The problem of on-line identification of linear multivariable discrete-time systems from input-output data is considered. A study has been made of the relative effectiveness of the four different models used in the area of identification of linear multivariable systems (transfer-function matrix, impulse response matrix, input-output difference equation and state space). The features of each model and its effect on the complexity of the identification algorithm as well as the bias of the paramter estimates while using the ordinary least-squares method have been studied. Different on-line algorithms have been proposed for the identification of the given system directly in each of the four different model representations. These algorithms estimate the parameters of the system from noisy measurements and no knowledge of the noise characteristics is required. The identification of a given multivariable system has been deccomposed into the identification of m subsystems (where m is the number of outputs) and the parameters of each subsystem are estmated independently from each other. The problem of structure determination has been considered, and algorithms have been proposed for the estimation of the structural parameters of the transfer-function matrix and the state space representations from noise-free as well as noisy measurements. Also, a two-stage bootstrap algorithm has been derived for combined parameter and state estimation of linear multivariable systems.</p> / Doctor of Philosophy (PhD)
| Identifer | oai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/11656 |
| Date | 05 1900 |
| Creators | El-Sherief, Hossny E. |
| Contributors | Sinha, Naresh K., Electrical Engineering |
| Source Sets | McMaster University |
| Detected Language | English |
| Type | thesis |
Page generated in 0.0021 seconds