Anshul B's Senior Project Blog

Project Title: ICON: using deep learning to build an autonomous car BASIS Advisor: Kay Yoo Internship Location: GCU Onsite Mentor: Ryan Woodward, Professor of Computer Science

Project Abstract

I will be working with Professor Ryan Woodward at GCU along with his colleagues. We present Intelligent Classification and Optimized Navigator (ICON), an AI-powered autonomous driving system that navigates using real-time image processing. ICON is a new generation of autonomous cars because it solely relies on image input, unlike other autonomous cars that use GPS and LIDAR; this allows ICON to navigate unmapped areas. ICON will use deep learning models such as CNN to process the images and create steer commands. Our main goal is to evaluate the effectiveness of a self-driving car that navigates based on images. We plan to implement ICON on a small car using a deep learning model trained and tested on images. Our main experiment location will be the GCU campus to assess its real-time performance. Key challenges will include lighting variations, occlusions, and generalization across different terrains. Then, we will compare ICON's performance to traditional GPS autonomous cars, which will then highlight the benefits and drawbacks of ICON. Overall, this project contributes to the growing field of vision-based autonomous cars, thus offering insights into the viability of image-only navigation for self-driving cars.