Authors:
T. Larkowski
;
J. G. Linden
;
B. Vinsonneau
and
K. J. Burnham
Affiliation:
Control Theory and Applications Centre, Coventry University, United Kingdom
Keyword(s):
Bias compensation, Bilinear systems, Errors-in-variables, Recursive estimation, Regularization, System identification.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Signal Processing, Sensors, Systems Modeling and Control
;
System Identification
Abstract:
The paper investigates a recursive approach for the bias compensating least squares (BCLS) technique. The method presented is applied to the problem of on-line identification of single-input single-output bilinear models in the errors-in-variables framework. Within this framework the recursive bilinear BCLS algorithm is realized when a bilinear Frisch scheme (BFS) is iteratively applied for the estimation of the parameters of an exemplary bilinear system, giving rise to the exact recursive BFS (ERBFS) method. Moreover, a further extension of the ERBFS incorporating Tikhonov regularization with variable exponential weighting is considered and this is shown to be beneficial in the initial period of the identification procedure.