Energy-constrained robots, such as those used in environmental monitoring and space robotics, must balance actuation and computation costs, which are often of similar magnitude. Prior work in the Low-Energy Autonomy and Navigation (LEAN) lab has reduced these computation costs through motion-planning and perception algorithms, including Compute Energy Included Motion Planning (CEIMP) and DecTrain for efficient deep learning. Building on this foundation, we enabled vision-based perception on robotic car test platforms and refined experiments to evaluate algorithmic efficiency. This project extends that work by co-designing hardware and software: deploying a metareasoning perception algorithm, accelerating it with FPGA-based hardware, and validating energy savings through hardware-in-the-loop experiments on miniature robotic vehicles.

Research at the intersection of hardware design, machine learning, and energy efficiency excites me. I am eager to apply and expand my experience across low- and high-level systems to optimize crucial tasks end-to-end in the fast-paced field of autonomous navigation. The SuperUROP program offers a unique opportunity to deepen my research skills and make meaningful contributions to the field, and I am excited for the journey ahead.