DocumentCode
3169399
Title
Experiment Design for Structured Nonlinear System Identification
Author
Hsu, Kenneth ; Vincent, Tyrone ; Poolla, Kameshwar
Author_Institution
Univ. of California, Berkeley
fYear
2007
fDate
9-13 July 2007
Firstpage
1554
Lastpage
1559
Abstract
When identifying nonlinear functions in interconnected systems, it is often desired that the exciting signal be sufficiently dense both spatially and temporally. However, density requirements are only realizable in infinite-length experiments. In this paper, we consider metrics for determining the quality of finite-length signals for the purpose of identification. Although the resulting experiment design problem is non-convex, we provide some insight that warrants further exploration of possible relaxations for the problem.
Keywords
identification; nonlinear systems; signal processing; density requirement; finite-length signal; interconnected system; nonconvex experiment design; nonlinear function identification; structured nonlinear system identification; Cities and towns; Control systems; Heart; Interconnected systems; Mechanical engineering; Nonlinear control systems; Nonlinear systems; Sensor systems; Signal design; Signal processing;
fLanguage
English
Publisher
ieee
Conference_Titel
American Control Conference, 2007. ACC '07
Conference_Location
New York, NY
ISSN
0743-1619
Print_ISBN
1-4244-0988-8
Electronic_ISBN
0743-1619
Type
conf
DOI
10.1109/ACC.2007.4282750
Filename
4282750
Link To Document