Automatic Uncertainty Propagation Based on the Unscented Transform
Dailys Arronde Pérez, Harald Gietler, Hubert Zangl
Abstract
Automatic uncertainty propagation reduces the effort for computation of uncertainty and is thus a useful tool in a variety of applications. Typically, such tools utilize Taylor series approximations (in particular linearization) or Monte Carlo Methods to perform the calculations. In this paper, we propose the use of the Unscented Transform for automatic uncertainty propagation. A comparison between the approaches - realized in a toolbox for the MATLAB environment and illustrated in two application examples - shows that the Unscented Transform overcomes some of the limitations of linearization and Monte Carlo methods, providing reliable estimates of the output expectation and standard deviation in nonlinear problems evaluating a reduced number of sigma points.