Implicit Fractional Model Order Estimation Using Interacting Multiple Model Kalman Filters

Abstract

AbstractThe paper deals with implicit and explicit approaches for fractional nonlinear model order estimation using a benchmark model relating applied angular rate and neuron's firing intensity within the vestibular system. The implicit approach is based on an interacting multiple models scheme, where several extended Kalman filters with fixed fractional order nonlinear models are running in parallel. An alternative method based on augmented Unscented Kalman filter is proposed where, where the fractional order of the model is estimated explicitly within the filter state. The performance of the estimators is compared with respect to parameters of estimators and fractional derivative approximation.