TY - THES A1 - Schneider, Thilo T1 - Mathematical Methods for the Approximation of Radar Traces N2 - Most major airports collect recordings of the position of aircrafts at specific times. Those data typically requires extensive smoothing and corrections before it can be used for later analysis. Conventional smoothing approaches fail to model the movement physically correct, i.e. do not take standstills of aircrafts into account. In this thesis we develop a method to detect standstills, employ robust smoothing splines for data fitting, add adequate boundary conditions for the detected standstill periods (i.e. force the function to be constant and to entry- and exit-direction for the standstills to be identical) and give an algorithm to solve those approximation problems efficiently. In the progress we give an explicit proof for the convergence of the IRLS algorithm proposed by Huber to solve M-type estimates for non-linear approximation problems. Furthermore we derive a blueprint for a method to solve separable, quadratic least squares problems with very few quadratic variables. KW - Non-Linear Least-Squares KW - Non-linear IRLS KW - Huber-Approximation KW - DBSCAN KW - Robust Smooth Regression KW - Spline-Approximation KW - Numerisches Verfahren KW - Nichtlineare Optimierung KW - Radar Y1 - 2015 UR - https://opus4.kobv.de/opus4-uni-passau/frontdoor/index/index/docId/326 UR - https://nbn-resolving.org/urn:nbn:de:bvb:739-opus4-3269 ER -