Comparison of statistical methods for the alignment of strapdown inertial systems
MetadataShow full item record
AbstractIn the thesis, different numerical methods for the alignment of a Strapdown Inertial Surveying System (SISS) are investigated. Kalman filtering is used as the standard real-time processing method. A new wave filtering method proposed by Salychev and Bykovsky  is presented and successfully used as an alternative to the conventional real-time Kalman filter. For post-mission processing, a recursive least squares method suitable for linear dynamic systems is developed. Repeated Kalman filtering is used as an important means to improve the estimation of the azimuth. Optimal smoothing is also used for refining the azimuth estimation of the SISS. Shown through experimental results using the LTN-90-100 strapdown system is that the different methods can be useful for different purposes. Kalman filtering and wave filtering can be used as real-time alignment algorithms. The wave method reaches a stable state within three minutes while the Kalman filter needs ten minutes. On the other hand, Kalman filtering provides an estimate of the filtering accuracy and this allows statistical testing which is not possible with the wave method. Repeated Kalman filtering, optimal smoothing and least squares methods give comparable results in post-mission processmg. They are accurate within about one arc minute and are stable to about ten arc seconds over a period of ten minutes.
Bibliography: p. 78-80.