Input Error versus Output Error Model Reference Adaptive Control

Algorithms for model reference adaptive control were developed in recent years, and their stability and convergence properties have been investigated. Typical algorithms in continuous time involve strictly positive real conditions on the reference model, while similar discrete time algorithms do not require such conditions. We show how algorithms differ by the use of an input error versus an output error, and present a continuous time input error adaptive control algorithm which does not involve SPR conditions. The connections with other schemes are discussed. The input error scheme has general stability and convergence properties that are similar to the output error scheme. However, analysis using averaging methods reveals some preferable convergence properties of the input error scheme. Several other advantages are also discussed.