This thesis aims at studying phase transition in vehicular traffic with a Boltzmann-type kinetic approach. Phase transition naturally emerges from the derivation of macroscopic quantities as a consequence of binary microscopic interactions. No ansatz is needed, contrary to macroscopic models. First, the kinetic traffic model is presented referring to the literature. It is based on a follow-the-leader approach and analytical results are obtained in the quasi-invariant interaction regime. Both the case with and without autonomous vehicles are considered. In particular, the introduction of autonomous vehicles is investigated as a tool to mitigate road risk and relies on a Model Predictive Control approach.