Autonomous driving is an emerging industry, garnering sponsorship and attention from almost all big-name car brands and attracting scientists as well. However, most current autonomous driving approaches rely on highly detailed prior maps, which are extremely time-consuming to generate, maintain, and extend to other regions. This project aims to create a system capable of real-time application while actively quantifying uncertainty in any autonomous decision or segmentation through the use of Bayesian deep learning. Coupled with light detection and ranging (LIDAR), the pipeline will allow for a quickly trainable neural network capable of operating on any properly equipped vehicle.

SuperUROP appeals to me because of the intense learning I will achieve through research, experimentation, and execution even more so than the courses at MIT that have taught me the fundamentals. What prepared me for research is a growing drive to apply what I know and extend it to enhance autonomous driving. I am excited for the constant struggles, sporadic victories, but most important, the potential of this project and how it could impact the world.