Autonomous Car Prototype Design Using Udacity Simulator With Convolutional Neural Network

Abstract

This research presents the process and development of an autonomous vehicle prototype using Udacity Simulator with the Convolutional Neural Network (CNN) method. This research uses NVIDIA CNN Architecture for its convolutional neural network architecture. The objective of this research is to design and generate a machine learning model to run the Udacity simulator in autonomous mode, which allowing car objects on the simulator to move autonomously. The proposed approach involves training a CNN model using a labeled dataset of tracks images captured through training mode. The output of the CNN model is then used to control the steering and acceleration commands of the vehicle. The performance of the machine learning model is evaluated using MSE, RMSE, and MAE parameters. In addition, it is also evaluated on its ability to navigate one of the tracks in the simulator.