Autonomous Car Behavioral Training Using Deep Neural Network

Abstract

Autonomous driving is becoming increasingly prevalent nowadays. With the help of a number of images of car movement from the Kaggle self-driving dataset, we explore the feasibility of utilizing the images obtained to train a deep neural network to detect and predict the steering angle, which is the critical part of the car behavior. Since deep neural networks have emerged as powerful tools for training autonomous cars and learning about and improving their driving behaviors, we incorporate convolutional layers and additional layers in the deep neural network architecture so that it can capture the behaviors appropriately and provide effective results. We demonstrate that the implementation of this approach is successful and that the corresponding implementation highlights the potential of deep neural network in advancing autonomous car technology. Our comprehensive evaluation suggests that further research should concentrate on refining the network architecture and enhancing perception capabilities in order to deliver promising advances to the field.