Adaptive and learning systems: Theory and applications

THIS VOLUME is a selection of 28 papers taken from the Fourth Yale Workshop on Applications of Adaptive Systems Theory held in May 1985 at Yale University. The Yale workshops on adaptive control were first organized in 1979. The period 1980-1985 is especially significant in adaptive system theory, for it was during this period of time that the major results in what is now called robust adaptive control were first developed. Most of the focus in adaptive systems prior to 1980 was on the design of adaptive control laws for systems with unknown parameters, but known order and disturbance free (now commonly referred to as the ideal case). The main analysis tools prior to 1980 were Lyapunov functions, and hyperstability theory. The theory available prior to 1980 is well summarized in the text of Landau (1979). The discovery that additive disturbance signals and unmodeled dynamics could cause instability in adaptive systems in the late seventies and early eighties generated a great deal of interest in adaptive systems which were robust with respect to disturbance signals and unmodeled dynamics. The resulting robust adaptive control theory is the main focus of the first section, Adaptive Control Theory, of this volume. The volume contains four other sections titled: Adaptive Control Applications, Learning Systems, Control of Flexible Space Structures, and Robotics. There are currently very few books on the subject of robust adaptive control. Even the recent IEEE reprint volume (Gupta, 1986) contains only one paper on a robust adaptive problem. There are two books, Egardt (1979) and loannou and Kokotovic (1983), which deal with the disturbance-signal problem, and one book, Anderson et al. (1986), which does have a comprehensive treatment of the problem of robust stability of adaptive systems, using passivity and averaging methods of analysis. There is some overlap in the robust adaptive control theory covered in Narendra and Anderson et al. In fact both books have two common authors. However, the theory is covered in much more detail in Anderson et al. On the other hand, in Narendra a number of other topics, i.e. learning systems, adaptive control applications, flexible space structures, and robotics are included which are not included in the more specialized volume of Anderson et al. The lead paper by Narendra and Annaswamy in the first section of the book summarizes most of the theory that has been applied to robust adaptive control up to 1985, i.e. Lyapunov stability theory, theory of persistent excitations, total and practical stability theory, method of averaging, a-modification and dead-zone methods, passivity methods, etc. The second paper in this section, by Kosut, covers in more detail the method of averaging for adaptive systems, which is also the main focus in Anderson et al. The third paper, by Kreisselmeier, discusses the use of a dead-zone in the adaptive control law to deal with the problem of unmodeled plant dynamics. The fourth paper, by Praly, uses graph topology concepts to study the robustness of adaptive systems, also with respect to unmodeled plant dynamics. The fifth paper, by Ioannou and Tsakalis, deals with discrete-time systems and robustness with respect to additive plant uncertainties. The final paper in this section, by Kumar, discusses identification and adaptive control of linear stochastic systems. Unfortunately this is the only paper in