Seminars

Space engineering and science applications inherently rely on noisy and incomplete data, necessitating joint design of data acquisition and post-processing strategies. Effective experimental design and tailored computational algorithms must be developed together, with information theory bridging these processes and providing broadly applicable insights. This dissertation contributes to these areas through three applications.

First, it develops a closed-loop, information-driven sensor control strategy for rapidly acquiring newly detected anthropogenic space objects—a capability of significant interest to the United States Space Force. The approach is built around a random finite set statistics framework that considers negative information from empty measurement scans, multiple potential targets, and measurements corrupted by false positives and missed detections.

Second, this research advances trajectory planning for low-thrust spacecraft operating within a peer-to-peer navigation framework, reducing reliance on Earth-based communication relays. The method optimizes observer spacecraft trajectories by jointly considering information gain (observability) and control effort. The resulting optimal control problem is efficiently solved using sequential convex programming.

Finally, the dissertation presents an image post-processing algorithm using autoencoder neural networks to enhance the detection sensitivity of faint substellar companions in high-contrast astronomical imaging. The method leverages a large library of archival data to effectively predict and remove spatially correlated quasi-static speckle noise, enabling planetary scientists to access deeper contrasts and smaller separations, and thus refine exoplanet population statistics.

Collectively, these contributions demonstrate the utility of information-theoretic principles across distinct yet interconnected aspects of space engineering and science, facilitating advancements in operational efficiency, autonomy, and observational capability.