Track fitting with non-Gaussian noise Original Research Article

This paper is a continuation of a previous note on robust track fitting. The nonlinear filter which has been investigated there is extended by running several Kalman filters in parallel. The number of filters is kept small by a suitable collapsing procedure. We report results on the precision of the estimated track parameters and on the computational load required by the algorithm with non-Gaussian observation errors. We also develop the correct smoothing algorithm and discuss the treatment of non-Gaussian process noise.