Prediction error methods for limit cycle data

Prediction error methods are considered for identification of the forward linear dynamics of nonlinear feedback closed-loop systems which operate in a perturbed stable limit cycle. A model of the signals measured in a neighborhood of the limit cycle is presented and shown to satisfy a quasistationarity property. Quasistationarity is then used to prove that prediction error methods are both convergent and consistent for our data model. (C) 2002 Elsevier Science Ltd. All rights reserved.