Analysis of slow convergence regions in adaptive systems

Abstract

In this thesis, the convergence properties of errors are examined in a class of adaptive systems that corresponds to adaptive control of linear time-invariant plants with state variables accessible. The existence of a sticking region is demonstrated in the error space where the state errors move with a finite velocity independent of their magnitude. It is shown that these properties are also exhibited by adaptive systems with closed-loop reference models, which have been demonstrated to exhibit improved transient performance, as well as those that include an integral control in the inner-loop. Simulation and numerical studies are included to illustrate the size of this sticking region and its dependence on various system parameters. With the existence of sticking regions shown for inner-loop adaptive controllers, the impact on outer-loop control is demonstrated for systems that implement inner-loop adaptation.