It is shown that a simple modification (introducing a dead zone in the
adaptation law) of the known adaptive high-gain control strategy u(t)
= -k(t)y(t), k(t)= parallel to y(t)parallel to(2) yields lambda-tracki
ng in the presence of output corrupted noise for a large class of refe
rence signals and a large class of multivariable nonlinear 'minimum-ph
ase' systems of relative degree one. These results are applied to a re
alistic chemical reactor, showing the practical usefulness of these co
ntrol laws. (C) 1997 Elsevier Science Ltd.