Robust geodetic parameter estimation through iterative weighting

Abstract

A procedure for robust linear estimation of parameters on the basis of iterative weighting of observations is presented. The approach considers the weights of observations as not just functions of the observational variances, but as functions of both the observational variances and estimates of the observational residuals. However, as the residuals are themselves functions of the estimates of the unknown parameters, the entire estimation procedure is performed iteratively. Three test examples comprising a linear point estimation, a linear regression case, and a geodetic network have been adopted to demonstrate the procedure. The results indicate that the proposed approach is effective in the isolation and management of outliers, and further that the robust estimation is in general a more efficient estimation procedure than the ordinary least squares.