Project Abstract

The rapid advancements in autonomous vehicle technology coupled with the limitations of current driver-assistance systems under variable conditions such as lighting and weather highlight the urgency for this research. Our project uniquely integrates convolutional neural networks (CNNs) with the Robot Operating System (ROS) to develop a highly adaptive object detection system within a virtual environment, addressing a critical gap in automotive safety technologies. By employing a hybrid methodology that combines real-world driving footage and high-fidelity simulations through ROS and Gazebo, we refine the accuracy and response times of detecting pedestrians, traffic signs, and other vehicles. Preliminary findings suggest that this integration significantly enhances object detection capabilities under simulated conditions. This research paves the way for bridging the gap between current assistance systems and fully autonomous driving solutions, contributing substantially to the future of vehicular safety and efficiency.

Keywords: Object Recognition, Convolutional Neural Networks, Advance Driver-Assistance System

 Conference Details

Session: Presentation Stream 21 at Presentation Slot 8

Location: GH022 at Wednesday 8th 09:00 – 12:30

Markers: C�cilia Pradic, Joe Macinnes

Course: MSc Data Science, Masters PG

Future Plans: I’m looking for work