Recently with the advancement of deep learning and simulation technologies, the pace of deep learning is much faster the development of driverless cars. In this paper, we suggest the virtual environment in which an autonomous vehicle can navigate, perceive and control using machine learning algorithms. We use CARLA simulator for implementing deep learning approaches for object detection, lane keeping, and surveillance of environmental changes in the driving area. The simulation enables representation of autonomous driving systems in a risk-free environment (while training and testing their functionalities, making it a cost-effective solution that does not contain the real-world dangers. Moreover, we employ reinforcement learning methods to facilitate terrain and traffic responsive adaptive behaviour. Our findings suggest the effectiveness of machine learning enhancements on the performance and safety of Autonomous vehicles could be potentially fruitful when applied in real life situations, especially taking into account the results achieved from simulated environments.