Kalman Filtering and Quasilinearization, A Comparative Discussion of Two Procedures for Parameter Estimation,
Abstract
Two techniques which have had wide application in the identification of system parameters are Kalman filtering and quasilinearization. This paper discusses the quantitative differences between these two procedures and shows that corresponding estimates will be identical provided that the Kalman estimation is performed with the following restrictions: The state vector about which the process equations are linearized is the solution to the state equations containing the smoothed parameter and initial condition estimates computed at the end of the last data cycle. No updating of this trajectory should be performed as the data is serially processed. At some time after the start of the data record the recursive equations should be initialized with the parameter and state estimates which correspond to a least squares curve fit of the data up to that time. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1972
- Accession Number
- AD0739606
Entities
People
- Howard Kaufman
Organizations
- Rensselaer Polytechnic Institute