The Methods Of Star Map Recognition And Filtering For Integrated Navigation

With the development of aircraft technology, requirements on the performance of navigation systems become more and more stringent. Therefore, it is difficult to satisfy these requirements by a single navigation mode in real world application, and integrated navigation methods become more popular. One of these methods is the SINS (Strapdown Navigation System)/CNS (Celestial Navigation System)/GPS (Global Positioning System) integrated navigation. This dissertation focuses on the methods of star map marking, star map recognition, and information fusion of SINS/CNS/GPS integrated navigation. The main contents of this dissertation are listed as below.Firstly, we focus on the problem of star map marking, and propose a three-order-star-map-marking algorithm. On the basis of the characteristics of star distribution on the sky and the principle of capturing stars by star sensor, the geometric relation of stars on the sky is investigated in detail in this algorithm to avoid"hole"sky regions. Also, the theory of celestial geometry is used, and the third-order star map marking model is built. Thus, star sensors can avoid"thin"regions on the sky, and open properly to aim at dividing celestial regions. Furthermore, three stars can be captured successfully in one time. The third-order-star-map-marking algorithm not only can solve the"hole"problem of star capturing, but also can acquire high star-map recognition ratio.Secondly, we focus on the problem of star map recognition, and propose two star map recognition algorithms. The first algorithm is proposed on the condition of vehicle's slow-attitude maneuvering, an all-sky star map recognition algorithm based on reducing star recognition regions. In this algorithm, based on the principle of star sensor's image motion in the shutter time, the correlation of star sensor's projection positions in multiple star maps is analyzed. Further more, the theory of star map pattern recognition and the principle of coordinate transformation are employed to build the algorithm model. However, the second algorithm is proposed on the condition of multiple captured stars, an all-sky star map recognition algorithm based on star eigenvalues. In this algorithm, based on the principle of capturing stars by star sensors, star neighboring regions'characteristic information is analyzed. Further more, as star eigenvalues are selected as star recognition factors, the model of all-sky star map recognition can be built.Thirdly, we focus on the problem of information fusion in the SINS/CNS/GPS system described by the nonlinear/Gaussian error models. With the framework of hybrid nonlinear error models, UT transformation in the UKF algorithm is partly simplified, and an improved UKF algorithm is proposed. For the SINS/CNS/GPS integrated navigation system, a federated UKF filtering algorithm is proposed, which combines the improved UKF with the federated filter. According to this algorithm, the improved UKF serves as the local filters, and the federated filter is used to fuse the information from the local filters. Further, the federated UKF filtering algorithm is applied on SINS/CNS/GPS integrated navigation system, and the numerical simulation is performed. The effect of the proposed star map algorithm on the integrated navigation system is investigated. The federated UKF filter together with the all-sky star map recognition based on reducing star recognition regions is applied on the SINS/CNS/GPS integrated navigation system, and the numerical simulation analysis is done. Also, the effect of the celestial navigation on the SINS/CNS/GPS integrated navigation is investigated. The SINS/CNS/GPS integrated navigation system is compared with the SINS/GPS integrated navigation system, and the numerical simulation analysis is done.Finally, we focus on the problem of information fusion in the SINS/CNS/GPS system described by the nonlinear/non-Gaussian models. In view of the nonlinear/ non-Gaussian characteristics of the system, a federated unscented particle filtering algorithm is proposed. In this algorithm, UPF serves as the local filters, the federated filter is used to fuse outputs of all local filters, and the global filter result is obtained. Because the algorithm is not confined to the assumption of Gaussian noise, it is of great significance to integrated navigation systems described by the non-Gaussian noise. This algorithm is applied on the SINS/CNS/GPS integrated navigation system described by the nonlinear/non-Gaussian error models, and the numerical simulation analysis is done.