This paper analyzes the robustness of deep learning models in autonomous driving applications and discusses the practical solutions to address that.