UNIVERSAL LINEARIZATION CONCEPT FOR EXTENDED KALMAN FILTERS

A careful evaluation of the performance of a universal linearization c oncept-based extended Kalman filter is made by experimentally comparin g its Performance to that of a classical, linearization based, extende d Kalman filter, in the case of a simple nonlinear dynamical system. I n this work, instances of superior universal linearization concept-bas ed extended Kalman filter estimation performance, compared with its cl assical, linearization by differentiation-based extended Kalman filter counterpart, have been discerned. Thus, in the case of nonlinear dyna mics and linear measurements, the estimation advantage of the universa l linearization concept-based extended Kalman filter, compared with th at of the classical extended Kalman filter, increases when the process noise intensity decreases; conversely, in the case of linear dynamics and nonlinear measurements, the estimation accuracy advantage of the universal linearization concept-based extended Kalman filter, compared with that of the classical extended Kalman filter, increases when the process noise intensity increases. Furthermore, compared with the cla ssical extended Kalman filter, the universal linearization concept-bas ed extended Kalman filter is more robust with respect to variations in the dynamics' parameters, in both linear and nonlinear dynamics cases . The advantage of the universal linearization concept-based extended Kalman filter, compared with the classical extended Kalman filter, is more pronounced in the case of small process noise intensity. Encourag ed by the promising experimental results, in this paper a general and novel universal linearization concept-based extended Kalman filter is derived.