Published Paper Details:

The automobile industry has put a lot of effort and innovative work over recent years which has made the automobile industry one of the most exciting engineering disciplines. Self-driving Vehicles are constantly a piece of Sci-fi films and are considered as the eventual fate of the automobile industry. Path-keeping, adaptive cruise control, and automatic braking are the semi-autonomous functions introduced in the market for modern vehicles. The aim of this project is to design and simulate autonomous ground vehicle which is able to move autonomously while detecting and avoiding obstacles with the help of deep learning. Autonomous vehicle development and machine learning both are using deep learning as the most important frontier. This idea can be extended to a wide array of applications such as transportation in malls, hotels or schools. The training is conducted using the CARLA(Car Learning to Act) simulator annually and it achieves good results on recognizing lane lines on road and traffic signs. The goal of the work is to prepare a convolutional neural network for control directions via delineating pixels from a single front-facing camera which is looking straightforwardly. This approach is demonstrated in many autonomous vehicles. The framework explains here how the system learns and what are the different decisions it takes while driving on highways, in heavy traffic on nearby roads, and also on roads with or without lane markings with less training data collected from peoples. It also gives good results in areas with an unclear visual condition, such as in parking areas, and on unpaved streets.