Lietuvos GPS pirmosios klasės išlyginimo kokybės analizė taikant nupjautinio Lp pasiskirstymo dėsnį

Abstract

The article discusses robust procedures using the truncated distribution Lp in the analysis of adjustment of GPS networks. The distribution Lp in the method of least squares involves the iterative re-weightening of observations. The new weights in iterations were calculated by three different methods. For experimental calculations the 1-order GPS network of Lithuania was taken. From the charts presented in the work it is seen that after the 1st iteration the corrections of observations get the values close to zero, what makes the values of new weights in the consequent iterations to be close to zero too. At the same time, the diagonal values of the inverse matrix N-1 in neighbouring iterations varies periodically with a very big amplitude. The unstable variation of the matrix values shows that the truncated distribution Lp, used in calculations, does not conform to the real distribution of the network errors. The reciprocal condition number estimate of the matrices P and N in iterations shows the intention of the matrices to become illconditioned after 1st iteration. Therefore further correction of the adjustment results does not make sense, unless the observations have several standing-out errors.[...].