Prediction of Steering Angle of vehicle using Deep Learning Models

Mohammed, Javed (2022) Prediction of Steering Angle of vehicle using Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (628kB) | Preview

Abstract

In today’s advent of technology an exponential rise and enhancement has been observed in the field of computer vision, image processing and deep learning. This renaissance has recently dominated the automation field. The concepts of image processing can be easily applied in driving vehicles wherein the drivers can be replaced with the rising technology. However, the most important factor that comes into picture in automation is the steering angle of a car by which the vehicle is responsible to take the curve. Automotive vehicles (AV’s) have started considering the steering angle prediction and many automotive companies have also invested in it such as Tesla and Udacity. This filed has however attracted multiple researchers and insurance companies to invest in them. Deep learning architectures have been considered to be the apt fundamentals that can be applied in such a scenario. Hence, this project proposes a steering angle prediction in AV’s using DL. The implementation is carried out in two modules namely: image processing and CNN. For every point along the trajectory of the vehicle, a steering angle is calculated, considering the speed of the car and the applied brakes as significant parameters and a set of images are captured by the cameras installed. The first phase of execution involves image processing that utilizes images for the training purpose and data augmentation for resizing the images. In the next phase, CNN concepts are applied and the processed image obtained from first phase is fed as the input to this phased and a predicted steering angle is generated as the output. For this type of automated model, the leading car companies use the architecture called ‘Alex Net’ and ‘Pilot Net’ is the other architecture which was created by Tesla, these types of architectures basically can store the bigger size images or highly resulted images which leads to the complexity and high cost. So, for reducing the price and complexity of model in this project we are going to create a light architecture by embedding big frame architectures into it and named as ‘J-Net’.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > TL Motor vehicles. Aeronautics. Astronautics Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	23 Feb 2023 13:20
Last Modified:	02 Mar 2023 08:50
URI:	https://norma.ncirl.ie/id/eprint/6231

Actions (login required)

View Item