×

On the possible divergence of the projection algorithm. (English) Zbl 0835.93015

By means of an example, the authors show that the sequence of estimates generated by the projection algorithm does not necessarily converge. This is in contrast to recursive least squares. In applications such as adaptive control, it is mostly desirable to derive as many properties as possible from the identification part of the design without having to rely on the specific nature of the input. The input depends in a highly nonlinear fashion on the estimates. The construction of the authors shows that convergence is not automatic among the properties that can be derived without additional assumptions on the input sequence.
Reviewer: G.Dimitriu (Iaşi)

MSC:

93B40 Computational methods in systems theory (MSC2010)
93B30 System identification
PDFBibTeX XMLCite
Full Text: DOI Link