Abstract
Traditional schemes of non-linear estimation includes extended Kalman filter (EKF). However due to several shortcomings caused by Jacobian linearization the usage of EKF is problematic. To avoid the problems linked with Jacobian linearization, this paper presents Kalman filtering technique based on statistically linearization. The derivation of this nonlinear estimation scheme has been achieved by steps similar to standard Kalman filter (KF) techniques. The system is linearized through statistical linearization rather than Taylor series. This statistically linearization is implemented to obtain the state of two important models, namely two phase permanent magnet synchronous motor (PMSM) and univariate non stationary growth model. It has been shown that the schemes has generated improved performance than EKF. Various performance indeces have been shown for performance comparison. Results obtained through two estimation techniques are compared with the actual state values. The results obtained through proposed scheme are significantly improved compared to the results obtained for existing schemes. In consequence, the error linked with proposed estimation techniques has been greatly minimized through the use of statistically linearized KF.