We are interested in the identification of an unknown time varying additive
component of a controlled nonlinear autoregressive model, a problem of int
erest in the modeling and control of uncertain systems, such as those met i
n biotechnological processes. A kernel-based nonparametric estimator is pro
posed whose almost sure convergence is studied by means of a Lyapunov stabi
lizability assumption and laws of large numbers for martingales. We then ad
apt the general result to several classes of deterministic or random functi
onal model uncertainties.