On a boundedness conjecture for output error adaptive algorithms

Abstract

A boundedness conjecture is presented which states that all signals within a specified adaptive system including the output error and parameter estimates remain globally bounded for all time despite any locally unstable behavior. Results suggested by computer simulations are verified by an exact analysis of a linearized periodic version of the adaptive system.<>