Modeling human driving behavior

Abstract

The goal of this thesis paper is to explore models that can predict and anticipate driver behaviors on the road and give probabilities on future actions of neighboring vehicles, while being lightweight enough to be formally verifiable. This thesis starts with looking into related work and doing a short literature review on previous work on driver models. We then talk about the available datasets used to perform such work, different models used (from classic regressions to neural networks) and finally present my approach and my results.