Square-root unscented Kalman filter for state estimation of permanent magnet synchronous motor

Abstract

This paper investigates a square-root unscented Kalman filter for state estimation of a motor drive without position sensor. The square-root filter has advantages of the numerical stability, and memory and code size reduction. A combination of the square-root filter and Kalman filter is an effective way for preventing the performance degradation due to round-off error. The appearance of an unscented Kalman filter was caused by the linearization process error of a conventional Kalman filter type. Using an unscented transformation, the unscented Kalman filter can estimate a accurate state values without the linearization process. This paper investigates the design, implementation and performance of the Potter square-root filter combined with the unscented transformation. A code size and computation time are reviewed and compared to the unscented Kalman filter. The experimental results prove the validity of the designed filter, and show a powerful performance compared to a square-root extended Kalman filter the state estimation of a sensorless permanent-magnet synchronous motor.