Algorithmic innovations in extended unbiased FIR filtering of nonlinear models

Two algorithms of extended unbiased FIR (EFIR) filtering are proposed for nonlinear state estimation. The first algorithm is basic and the second one employs the nonlinear-to-linear observation conversion obtained by the batch EFIR filter with minimum memory. Unlike the extended Kalman filter (EKF), both EFIR algorithms ignore the noise statistics and demonstrate better robustness, but require the optimal horizon. Applications are given for robot indoor self-localization utilizing radio frequency identification tags.