A methodology which performs the robust updating of complex uncertain dynamical systems with respect to modal experimental data in the context of structural dynamics is constructed. Since both model uncertainties and data uncertainties must be considered in the computational model, then the uncertain computational model is constructed by using the nonparametric probabilistic approach. An extension to the probabilistic case of the input error methodology for modal analysis is presented. Such robust updating formulation leads to solve a mono-objective optimization problem in presence of inequality probabilistic constraints. A numerical application is presented.