W. Waltenberger (HEPHY VIENNA)
State of the art in the field of fitting particle tracks to one vertex is the Kalman technique. This least-squares (LS) estimator is known to be ideal in the case of perfect assignment of tracks to vertices and perfectly known Gaussian errors. Experimental data and detailed simulations always depart from this perfect model. The imperfections can be expected to be larger in high luminosity experiments like at the LHC. In such a context vertex fitting algorithms will have to be able to deal with mis-associated tracks and mis-estimated or non-Gaussian track errors. We present a vertex fitting technique that is insensitive to outlying observations and mis-estimated track errors, while it retains close-to-optimal results in the case of perfect data; it adapts to the data. This is realized by introducing weights that are associated to the tracks and reflect the compatibility of the tracks with the vertex. Outliers are no longer simply discarded - as is done in most of the classical robustification techniques - but rather downweighted. The algorithm will be presented in detail, and comparisons with classical methods will be shown in relevant physics cases.