particle and sigma -point kalman filters: a comparison of performance

aiming to track the full state of a nonlinearly evolving structural system and simultaneously calibrate its constitutive model in this paper we compare the performances of the sigma - point kalman filter S-PKF , of a standard particle filter PF, and of an extended kalman -particle filter EK- PF. to ensure that structural responses do not affect the filters performances we focus on a single degree - of - freedom DOF system and show that the three filters are all capable to track possible failure mechanisms as far as model calibration is concerned the newly proposed EK-PF performs better than the other two filters providing unbiased parameter estimates even when the system is diverging because of a failure event.

Keywords:

nonlinear systems sigma - point kalman filtering , particle filtering , state tracking , parameter identification.