An Adaptive Kalman Identifier and Its Application to Linear and Non-Linear ARMA Modeling.
Abstract
The problem of accurately replicating the parameters which define a given system for the purposes of implementing modern control strategies is important. Using an Autoregressive-Moving Average (ARMA) representation for the unknown system, a model is identified by processing input/output data to estimate the coefficients associated with the ARMA equation. Identification of unknown system parameters using Kalman filtering methods was accomplished by augmenting the state vector. In this thesis the Kalman filter is formulated so that parameters can be identified explicitly. We call this approach the Adaptive Kalman Identifier (AKI). It is shown that the Adaptive least mean square (LMS) Adaptive Recursive LMS and Adaptive Lattice filters are special suboptimal cases of the AKI. The convergence and modeling properties are compared with those of the AKI by simulation using various types of data. With minor modification, the AKI algorithm was used to identify the linear and non-linear ARMA models of the phase locked loop (PLL). A discrete PLL using a forward Euler integration scheme was used as a source of non-linear data. The AKI technique appears to enable one to discern when a potential non-linear system enters into its non-linear mode of operation. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1981
- Accession Number
- ADA102707
Entities
People
- Leopoldo M. Mayoral
Organizations
- Naval Postgraduate School