We present a new on-line multi-model algorithm for parameter estimatio
n of time-varying nonlinear systems. The time variation is captured by
assuming that the system parameters change according to a Markovian m
echanism. The algorithm postulates a finite number of possible values
of the system parameter and computes recursively the credit function o
f each parameter value, according to its predictive accuracy. A conver
gence analysis of the algorithm is presented which indicates that the
algorithm estimates correctly the parameter value in the time interval
s between source switchings. This conclusion is corroborated by numeri
cal experiments. (C) 1998 Elsevier Science Ltd. All rights reserved.