The field of autonomous vehicles is quickly emerging and becoming more prevalent in technological industries and everyday life. However, autonomous vehicles continue to need improvement. Currently, research and development is still being done, since in theory autonomous vehicles would have many benefits regarding road safety and infrastructure. Thus, this project aims to control a simulated vehicle using convolutional neural networks. Using a convolutional model, the program controls a vehicle using predicted steering angles from image inputs at specific points in the simulated environment. The simulator used for the project was built using the Unity game engine, while the convoluted models were made using Keras and Tensorflow libraries. It was discovered that models generated with a high enough accuracy can perform well in the simulation, since it can run for a few laps without crashing. Models that had the highest accuracy used either ReLU or ELU activation functions, without dropout layers. They also had preprocessed images using a Gaussian blur and resizing. The training data consisted of both original data taken from the simulator and augmented images, since a wide variety of training data helps the model to generalize its results. Finally, the loss function used was the mean squared error, with an adaptive movement estimation optimizer.