Study Of BDS Satellite Orbit Determination Using Kalman Filtering Algorithm

BDS is composed of three different constellations: GEO, IGSO and MEO, which can achieve better navigation and positioning capabilities with fewer satellites. The accuracy of satellite orbit directly affects the precision of navigation and positioning, so it is of great significance to study the theory of orbit determination technology and algorithm based on BDS. The main research contents and results are displayed as follows:1. The basic theory and method of navigation satellite orbit determination were introduced, including the dynamic model of the satellite, the satellite observation model, orbit numerical integration method, parameter estimation method and the track accuracy evaluation method.2. The basic principle and algorithm of the bi-directional Kalman Filtering orbit determination were studied. By analyzing radial accuracy of overlapping segment after calculating the 18 regional stations’ observation data of BDS, we have results such as, radial precision of GEO satellite orbit was 0.2m, that of IGSO satellites orbit was 0.16 m and for MEO satellites, it was 0.08 m. In terms of little gross errors in the data preprocessing stage, robust estimation method adding was added in parameter estimation. As for the situation that some of the gross errors may be distributed to normal observations, which will decrease the observations’ weight, when observation stations or epoches are not enough or when the structure of satellites is not stable, a robust algorithm was proposed based on decreasing the weight of the maximum standardized post-estimation errors in every iterative process.This method would increase the orbit determination accuracy remarkably.3. Three kinds of commonly used stochastic orbit determination models of GPS satellites were introduced, different stochastic models’ accuracy, which were constructed by bi-directional Kalman Filtering, were compared. Because the BDS satellite navigation system contains three different types of satellites, Helmert variance component estimation method was used to adjust the weights between different types of satellites. For observations from regional MEO stations were not enough and Kalman filtering was instable at beginning, an improved algorithm was put forward, in which Helmert variance component estimation was adopted after the filtering prosess stayed still and corresponding strategies were used according to the numbers of different types of satellites. The experimental results showed that the average accuracy of BDS radial orbit improved 2cm with the improved Helmert variance component estimation method.4. The stability of Kalman Filter was analyzed. How the initial state values influenced the Kalman filter’s results of BDS orbit determination were studied; the results showed that the error of initial values had considerable impact for GEO satellites but little for IGSO satellites or MEO satellites. For Kalman Filtering system was unstable and length of segmental arc failed to be long enough to make Kalman Filter keep stable, errors of initial values had a great influence on the results of the filtering process. Therefore, an iteration bi-directional Kalman Filtering algorithm was put forward, in which values worked out by using bi-directional Kalman Filtering were considered as initial values to solve orbit determination of satellites and the other parameters were re-settled with the same variance. The experimental results showed that if the initial prior variances stayed the same, bi-directional iteration Kalman Filtering method improved the precision of orbit determination.5. Three widely used kinds of GPS pressure models were presented in this paper, whose fitting and extrapolation precision were compared in BDS satellite navigation system. It turned out that BERNE model could reach the highest fitting level, yet, extrapolation accuracy varied with various kinds of satellites. According to the development of GPS pressure model and the form of Fourier series, second-order-cycle items were inserted in BENRE 9-parameter model to form a new basic model and parameters’ dimension was reduced by correlated coefficient method for three kinds of satellites, then, separateed pressure models were established. It could be validated through available data: for GEO satellites, 15-parameter BERNE pressure model should be adopted; For IGSO satellites, 13-parameter BERNE pressure model should be adopted; For MEO satellite, 9-parameter BERNE pressure model should be adopted.