We stated the main properties of the Boltzmann equation in the first part
of this thesis. After that we modeled the equation on a discrete velocity
grid and we proved, that the modeled equation had the same properties as
the original equation. In the second part we developed numerical algorithms
for solving the (now) discrete equation. We then implemented this
algorithms in C++ for the computation on a CPU and parallelized in CUDA for
the computation on a GPU. These implementations were tested on various well
known test problems and we investigated if a parallelized implementation,
of this problem, on a GPU could be useful.