Since existing adaptive observers for nonlinear systems may generate unboun
ded parameter estimates in the presence of bounded disturbances, robust ada
ptive observers are presented which prevent parameter estimate drift. In ad
dition the input to state stability of the error dynamics with respect to d
isturbances and parameter time-derivatives is guaranteed by generalizing a
persistency of excitation result. Asymptotic convergence of state estimatio
n errors is still achieved for systems in adaptive observer form when distu
rbances are not present, by a suitable extension of Barbalat's Lemma.