Robust estimation in geodetic networks

Abstract

The application of robust estimation to geodetic networks is analysed versus the classical least-squares approach. In case of gross or systematic errors appearance either in the mathematical model or in the observations to be adjusted, least-squares estimation along with detection statistical tests over the results present considerable problems for isolating them and avoiding their influence. Conversely, robust estimation provides a maximum-resistance solution and therefore the capability of identifying and quantifying them. Finally, we show the advantages of dealing with robust estimation as a global optimization problem rather than as an iteratively reweighted least squares scheme.