Abstract There are disadvantages of the BDS Navigation Satellite System (BDS), such as complicated constellation structure and large differences in residual errors of observed values from different types of satellites. In order to precisely determine the weight matrices of observed values between different systems, this paper studies the observed values from IGSO, MEO and GEO satellites in the BDS system through classification and weighting. It proposes an exponential weighting Helmert variance component estimation method that takes historical weight ratios into account. Static and dynamic navigation tests are carried out through a point positioning model to verify the effectiveness of this improved method in GPS/BDS combined positioning, and the results of this improved method are then compared with those of the equal weighting model, as well as of the Helmert variance component estimation method without exponential weighting. Results show that with the improved method, positioning precision in east, north and zenith directions is improved significantly. Under static conditions, compared with the equal weighting model, this improved method results in 50.0%, 51.0% and 420% improvement in positioning precision in the above three directions respectively; under dynamic conditions, relevant positioning precision is improved by 10.0%, 8.0% and 9.0% respectively. Especially under situations of fewer satellites and weaker intensity of satellite geometry, improvement in positioning results is even more significant.
