Atomic collisions have been very important for the development of our present understanding of the composition and behaviour of matter at the atomic level. They are also relevant in a wide range of applications; from physical chemistry to fusion reactors, from cancer treatment to research in astrochemistry. In the thesis we apply ab-initio methods to investigate the charge transfer processes when atoms enter in collision with other atoms and molecules. The focus of this work is concentrated in processes that due to their size are not amenable to the accurate methods used in atomic physics. Calculations are based on real time propagation using time-dependent density functional theory (TDDFT) a theory that uses an auxiliary set of non-interacting electrons with an interacting density to describe the state of the electrons in the system. In this work, the application of TDDFT is explored for atomic collisions in a set of problems of interest. First, we demonstrate that TDDFT can be applied in the simple case of a collision of a H+ with a He atom, with agreement with experimental results. Also we show the feasibility of simulate larger systems such as an atom of gold in collision with a butane molecule. When the particles involved in the collision has a velocity, the static ground state of the electrons associated to moving nuclei is not the ground state of the same nuclei in movement. The proper initialization of the wavefunctions is an issue. We introduce a velocity phase factor to account for the Galilean transformation from the static electronic ground state to the dynamic framework of the collision. Several alternatives are proposed in the case of multiple moving particles and a special discussion is devoted for the case of non-local pseudopotentials. Finally, a more challenging process was modelled, the collision of a 2 MeV Li+ impacting a C60 fullerene molecule. The capture cross-section was computed and compared with experimental data available for such process.