Copyright 0 IFAC Nonlinear Control Systems Design,
Enschede, The Netherlands, 1998
DEVELOPMENTS IN TIME-VARYING FEEDBACK
STABILIZATION OF NONLINEAR SYSTEMS
Pascal Morin, Jean-Baptiste Pomet, Claude Samson *,1
* INRIA
2004 Route des Lucioles
06902 SOPHIA-ANTIPOLIS, FRANCE
E-mail: (Firstname.Lastname}@inria.lr
Abstract: Over the last ten years, time-varying feedback has emerged as a new
approach for the asymptotic stabilization of highly nonlinear control systems. This
paper gives an overview of the developments made in this area. In particular, relations
between theoretical developments and practical motivations are emphasized.
Copyright © 1998 IFAC
Keywords: Nonlinear control systems, stabilization, time-varying feedback
1. INTRODUCTION
We consider control systems of the form x =
l(x,1J.), x E Rn, E Rm, with 1(0,0) = 0.
To simplify, it is assumed that 1 is analytic in
a neighborhood of (x = 0,1J. = 0). The general
problem under consideration is the asymptotic
stabilization, in the sense of Lyapunov, of the
origin x = ° via the use of a feedback control
law. Let us just recall that asymptotic stabilization
means the satisfaction of two properties: i) stabil-
ity of the origin of the closed-loop system, and ii)
convergence to this point of all the solutions of the
closed-loop system.
Within this large set of systems, and with respect
to the problem 01 asymptotic leedback stabilization,
one may try to distinguish several subsets of in-
terest. The first one (obvious and important) is
composed of systems the tangent approximation
of which, at the equilibrium point (x = 0,11.= 0),
is stabilizable, i.e. which can be locally asymptoti-
cally stabilized by means of a linear feedback 11. =
Kx. This set contains, in particular, all control-
lable systems which can be exactly linearized via
either static or dynamic feedback, since controlla-
bility of the tangent approximation is a necessary
condition for this (Charlet et al., 1991). Most of
1 The authors are with the "GDR/PRC d'Automatique"
565
the existing feedback control literature has been,
and still is, devoted to this class of systems. Al-
though the function 1 (., .) may be nonlinear, the
system may be viewed as almost linear in the sense
that any feedback control technique developed for
stabilizable linear systems successfully applies in
order to ensure i) local stability, ii) fast (expo-
nential) convergence of the system's trajectories
to zero, and iii) some amount of robustness with
respect to modeling errors.
Another subset of potential interest is composed
of systems which are controllable at the origin
-in a ''reasonable'' sense-, but whose tangent
approximation at (x = 0,1J. = 0) is not sta-
bilizable, so that none of the classical feedback
techniques based on some type of linearization
applies to them. In this respect these systems are
truly nonlinear, by contrast with those belonging
to the first category. For most of these systems,
there exist open-loop controls which rapidly drive
the system's state to zero, when the initial error
is not too large and in the absence of modeling
errors. However, this does not imply, contrary to
the case of linear systems, that asymptotically
stabilizing feedbacks exist. As a matter of fact,
studying this class of systems has revealed that
non-existence of asymptotically stabilizing contin-
uous pure-state feedbacks 11.( x) is a rather common
situation, as first pointed out in (Sussmann, 1979).