2000 Fiscal Year Final Research Report Summary
Research on System Identification Based on Modeling Accuracy
| Project/Area Number |
10650432
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Control engineering
|
|---|
| Research Institution | TOTTORI UNIVERSITY |
|---|
Principal Investigator |
UOSAKI Katsuji Tottori University, Dept.Inf. and Knowl.Eng., Professor, 工学部, 教授 (20029151)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HATANAKA Toshiharu Tottori University, Dept. Inf. and Knowl.Eng., Research Associate, 工学部, 助手 (10252884)
|
|---|
| Project Period (FY) |
1998 – 2000
|
| Keywords | System Identification / Recursive Estimation / Stochastic Approximation / Nonlinear Systems / Change Detection / Random Varying Truncations |
|---|
| Research Abstract |
Generally speaking, identification error should be minimized in system identification. The error comes from several sources, such as observation error and computation error. In this research, such identification errors have been discussed and a novel identification approach has been considered and some applications of identification have been discussed..
1. Analysis of identification errors
Computation error in system identification has been discussed using the idea of interval arithmetic. Based on the analysis, an identification approach has been developed, of which applicability will be considered in the future.
2. Recursive estimation based on stochastic approximation Stochastic approximation method to estimate the time-varying root of the regression function has been discussed. To improve its convergence, a transformation function for residuals is constructed using the empirical distribution of the residuals. The idea of random varying truncations is applied to relax the convergence conditions. These ideas are applied to design nonlinear filters.
3. Change detection in systems with uncertainty
Change detection has been discussed for the systems with model uncertainty, which is caused by identification error.
|
