Improved Filtering-Smoothing Algorithm for GPS Positioning

The well-known Unscented Kalman Filter (UKF) is widely applied to nonlinear system, while smoothing algorithm shows advantage of accuracy improvement in post processing application. But conventional filter is harassed by roundoff error due to processor´s finite-wordlength in practice. This paper proposes an improved filtering-smoothing algorithm which replaces UKF with Square-Root UKF(SR-UKF) in the forward filtering pass. Two smoothers, fixed-interval smoother and fixed-lag smoother, are incorporated in the backward smoothing pass respectively to form two iterative filtering-smoothing algorithms for GPS positioning estimation. System model is addressed first, then SR-UKF and smoother implementations are described respectively, effectiveness of new algorithm is evaluated by analyzing experiment results, future work will also be discussed.