Performance bounds of forgetting factor least-squares algorithms for time-varying systems with finite measurement data
Authors:
- 14 March 2005
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 52 (3) , 555-566
Abstract
This paper on performance analysis of parameter estimation is motivated by a practical consideration that the data length is finite. In particular, for time-varying systems, we study the properties of the well-known forgetting factor least-squares (FFLS) algorithm in detail in the stochastic framework, and derive upperbounds and lowerbounds of the parameter estimation errors (PEE), using directly the finite input-output data. The analysis indicates that the mean square PEE upperbounds and lowerbounds of the FFLS algorithm approach two finite positive constants, respectively, as the data length increases, and that these PEE upperbounds can be minimized by choosing appropriate forgetting factors. We further show that for time-invariant systems, the PEE upperbounds and lowerbounds of the ordinary least-squares algorithm both tend to zero as the data length increases. Finally, we illustrate and verify the theoretical findings with several example systems, including an experimental water-level system.Keywords
This publication has 31 references indexed in Scilit:
- Application of dual-rate modeling to CCR octane quality inferential controlIEEE Transactions on Control Systems Technology, 2003
- Finite sample properties of system identification methods2003
- Generalized predictive control for non-uniformly sampled systemsJournal of Process Control, 2002
- Analysis of dual-rate inferential control systemsAutomatica, 2002
- Ripple-free conditions for lifted multirate control systemsAutomatica, 2001
- Identification of fast-rate models from multirate dataInternational Journal of Control, 2001
- Finite sample properties of system identification of ARX models under mixing conditionsAutomatica, 2000
- Finite sample properties of linear model identificationIEEE Transactions on Automatic Control, 1999
- Multirate sampled-data systems: all H/sub ∞/ suboptimal controllers and the minimum entropy controllerIEEE Transactions on Automatic Control, 1999
- ℋ/sub 2/-optimal design of multirate sampled-data systemsIEEE Transactions on Automatic Control, 1994