Research Projects

Coral reefs are among the most biologically diverse and economically essential ecosystems on the planet, yet they are increasingly threatened by pollution, human interaction, and climate change. Traditional coral reef monitoring methods tend to be invasive and limited, prompting the need for innovative robotic solutions.  This project leverages a foundation model to predict small-scale local ocean flows around coral reefs. This enables us to train a robot with reinforcement learning to safely navigate complex reef structures under uncertain conditions.

Flying drones can be challenging for beginners, especially in cluttered environments. This project seeks to develop a drone software stack that assists human pilots by navigating safely around obstacles using only RGBD sensor data, without full state information. Trained through reinforcement learning, the drone employs a task-agnostic safety filter that intervenes only when the pilot's actions put it at risk, enabling smooth and intuitive human-robot collaboration. We aim to further advance this technology for agile and dynamic drone flight applications.

We are developing a modular, autonomous robotics stack that enables seamless integration of various robot dynamics and control algorithms. MiniTrucks serve as an excellent platform for testing this system and are utilized in Princeton University's ECE346 course, Intelligent Robot Systems