Abstract:Based on the existing weighted symmetric similarity transformation, only the case where the observation value contains random error is considered, and the case where the observation value contains the gross error is not considered. This paper further verifies that the weighted symmetric similarity transformation is not robust. Based on the weighted symmetric similarity transformation, the method of selecting weight iteration is adopted to make the robust weighted symmetric similarity transformation. The model obtains unit weight mean square error with robustness by the median method and utilizes the standardized residual to construct the weight factor function, to obtain a reliable parametric solution. Comparative analysis shows that: when the observations contain 4-6 gross errors, the method of this paper can be used to detect more data that may have gross errors, the weighting factors given are more reasonable, and the obtained parameter solution has higher accuracy and stronger stability.