The content of this paper is the adjustment of positional geodetic networks by robust estimations. The techniques (basedon the unconventional estimations) of repeated least-square method which have turned out to be suitable and applicable in the practisehave been demonstrated on the example of the local geodetic network, which was founded to compose this thesis. In the thesisthe following techniques have been chosen to compare the Method of least-squares with those many published in foreign literature:M-estimation of Biweight,M-estimation of Welsch and Danish method. All presented methods are based on the repeated least-squaremethod principle with gradual changing of weight of individual measurements. In the first stage a standard least-square method wascarried out in the following steps – iterations we gradually change individual weights according to the relevant instructions/ regulation(so-called weight function). Iteration process will be stopped when no deviated measurements are found in the file of measured data.MatLab programme version 5.2 T was used to implement mathematical adjustment.