Autonomous driving cars has always ignited human imagination with their great potential and countless advantages, it is even enough to mentioned that according to WHO organization the annual of annual road traffic deaths has reached 1.35 million in 2018. if Autonomous driving cars can just aid in solving this melancholic situation then it is already a huge research field; not forget to mention that autonomous driven aspires to have zero road accidents. The main task of this thesis is to analyze autonomous driving concepts, existing solutions, technologies and then demonstrate the use of a state-of-the-art approach to create an end-to-end deep learning model for self-driven vehicles. End-to-end approach learned with minimum training data to preserve lane centering in a traffic simulation without being explicitly trained to detect the lane marks. We have examined different existing model such as ALVINN, DAVE for autonomous driven and then based on this information we try to create our own model. We have trained a CNN for this purpose based on NVIDIA research paper and create a simplified version of it. Different image processing techniques applied to enhance the model i.e. adding a layer of noise to the image. The created model had been tested on Udacity open source simulator, and we conclude to have working model for both tracks available on the simulator. As a result, this thesis gives the necessary tools for everyone to develop his own CNN model for self-driven cars with minimum hardware requirements and basic machine learning knowledge