Résumé : Cosmic voids are promising cosmological laboratories for studying the dark energy phenomenon and alternative gravity theories. They are receiving special attention nowadays in view of the new generation of galaxy spectroscopic surveys, which are covering an unprecedented volume and redshift range. There are two primary statistics in void studies: (i) the void size function, which characterises the abundance of voids, and (ii) the void-galaxy cross-correlation function, which contains information about the density and velocity fields in these regions. However, it is necessary a complete description of the effects of geometrical (Alcock-Paczynski effect, AP) and dynamical (Kaiser effect, RSD) distortions around voids in order to design reliable cosmological tests based on these statistics. Observational measurements show prominent anisotropic patterns that lead to biased cosmological constraints if they are not properly modelled. This thesis addresses this problematic by presenting a theoretical and statistical framework based on dynamical and cosmological foundations capable of describing all the underlying effects involved: the expansion effect (t-RSD), the off-centring effect (v-RSD), the AP-volume effect and the ellipticity effect (e-RSD). These effects can be understood by studying the mapping of voids between real and redshift space. In this way, we lay the foundations for a proper modelling of the aforementioned statistics. In addition, we present a new cosmological test based on two perpendicular projections of the correlation function. The method is fiducial-cosmology free, which allows us to effectively break any possible degeneracy between the cosmological parameters involved. Moreover, it allows us to significantly reduce the number of mock catalogues needed to estimate covariances.