This research project proposal aims to enhance the capabilities of autonomous vehicles in challenging driving scenarios by developing innovative algorithms that dynamically incorporate safety constraints. The focus is on merging real-world safety considerations with theoretical models and testing them on physical robotic platforms. The study employs a combination of supervised and self-supervised learning techniques to improve perception algorithms, enabling autonomous vehicles to navigate complex and uncertain environments more effectively. By advancing autonomous driving technology, the research seeks to enhance safety, accessibility, and efficiency while addressing issues related to human error and variable driving contexts.

I am engaging in a SuperUROP to apply creative problem-solving skills to cutting edge research in my field. My previous work with AVs at NVIDIA (backend) and at the Autonomous University of Barcelona (research on mono-ocular depth perception) has prepared me for this project. I hope to gain a deeper understanding of interconnected machine learning systems and am most excited to gain an better intuition with self-driving vehicles.