Page 171 - Mathematical Techniques of Fractional Order Systems
P. 171

Chapter 6





             Fractional Order Error Models


             With Parameter Constraints



             Norelys Aguila-Camacho, Manuel A. Duarte-Mermoud and
                               ´
             Matı´as G. Mayol-Suarez
             University of Chile, Santiago, Chile


             6.1  INTRODUCTION

             Error models (EM) were first introduced in adaptive control literature by
             Narendra and Kudva (1974) and then in a unified form in Narendra (1976),
             for the analysis of integer order (IO) adaptive systems. They have become an
             important approach to face numerous adaptive control and identification pro-
             blems. Four EM have been recognized in the IO case and completely stud-
             ied, denoted as IOEM1 to IOEM4, where the adaptive laws (AL) are of IO
             type. For details on these IOEM, the reader is referred to Chapter 6 of the
             book by Narendra and Annaswamy (2005) where a complete study of them
             is performed. This viewpoint has been successfully used for continuous-time
             systems (Narendra and Kudva, 1974; Narendra, 1976; Narendra and
             Annaswamy, 2005) (which will be the focus of this study), but it has also
             been applied for discrete-time systems (Narendra and Lin, 1980) and even
             for hybrid systems (systems with two different time-scales) (Narendra et al.,
             1983, 1985). Its importance lies in the fact that the solution of any adaptive
             control or identification problem can be readily obtained if the problem is
             stated within this framework. The conditions under which stability and con-
             vergence of IOEM are attained, were established in Narendra and
             Annaswamy (2005) by defining suitable AL for each case.
                A further step in the study of IOEM was done in 1988 when the analysis
             of two IOEM, whose constant but unknown parameters are related through a
             linear algebraic relationship, was completely solved in Duarte and Narendra
             (1988a, 1996a), where conditions for stability and convergence of the con-
             trol/identification error and parameter error were derived. It is well known
             that in adaptive systems the ideal controller parameters and the true plant
             parameters are related by linear algebraic equations through the certainty
             equivalence principle or the Bezout identity. The importance of this subject



             Mathematical Techniques of Fractional Order Systems. DOI: https://doi.org/10.1016/B978-0-12-813592-1.00006-4
             © 2018 Elsevier Inc. All rights reserved.                   159
   166   167   168   169   170   171   172   173   174   175   176