Interpretable And Safe Deep Reinforcement Learning Control In Automated Driving Applications