We are working on a high-performance safe reinforcement learning algorithm that uses its knowledge of the system to adapt its behavior while maintaining near-optimal performance. Our approach aims to achieve a higher level of safety across entire trajectories, rather than relying solely on instantaneous safety guarantees.

We are developing deep learning-based 3D reconstruction pipelines using Neural Radiance Fields (NeRF) for uncooperative satellites. Using synthetic datasets from NVIDIA Isaac Sim and real-world satellite mockups, we investigate optimal viewpoint selection, reconstruction of tumbling satellites, and improved reconstruction accuracy under challenging lighting conditions.

We are exploring multi-agent reinforcement learning (MARL) to improve autonomous spacecraft inspection by optimizing image capture for 3D reconstruction. Using NVIDIA Isaac Sim, COLMAP, and Instant-NGP, we evaluate simulated inspection trajectories and reconstruction quality to refine reward functions, enabling flexible and effective 3D reconstruction of target spacecraft.

We are working with machine learning based methodologies to develop data-driven system identification and control techniques with applications to aerospace systems such as UAVs and spacecrafts. Our approach lets us linearize highly non-linear dynamical systems and thus provide a robust linearization scheme compared to traditional techniques such as Taylor series expansion.