Bursting Scenaria and Performance Limitations of Adaptive Algorithms in
the Absence of Excitation
Kostas S. Tsakalis
Supported in part by ARPA under grant F49620-93-1-0062 and in part
by NSF under RIA grant ECS-9111346.
Presented at the 3rd IEEE Mediterranean Symposium on New Directions
in Control and Automation, July 1995
Published in KYBERNETIKA, V.33, 1, 17-40, 1997; related results
in Automatica, V.32, 4, 549-560, 1996; also in Int. J. of Adaptive
Control and Signal Processing, V.8, 173-199, 1994 (Asymptotic Performance
Guarantees in Adaptive Control, with Suttipan Limanond).
A simple, yet general, bursting scenario is developed for a wide class
of parameter estimation and system identification algorithms in the absence
of sufficient excitation. This allows for an analytical derivation of a
lower bound on the worst-case performance of such algorithms in the presence
of perturbations. Simple examples are constructed, illustrating the implications
of these results in adaptive control.
Click here to
see a brief description of the main result, a bursting example and some