A multi-model algorithm for parameter estimation of time-varying nonlinear systems (Q1295142)

The paper is devoted to a new on-line multi-model algorithm of parameter estimation of time-varying nonlinear systems. The time variation is modeled by assuming that the system parameters are changing according to a Markovian mechanism. In the developed algorithm, a finite number of possible values of the system parameter is allowed. The credit function of each parameter value is recursively computed, according to a predictive accuracy. A convergence analysis of the algorithm is performed. It indicates that the algorithm correctly estimates the parameter value in the time intervals between source switching. This conclusion is validated by numerical examples.