Optimal Control Under Uncertainty
Sarah Gady, Princeton University
Robust optimization involves determining the best parameters or engineering design given uncertainties in measurements or manufacturing tolerances. The typical formulation minimizes cost, for example, while guaranteeing performance as measured by one or several metrics over the uncertainty set. When there is one performance metric, the robust optimization problem is equivalent to a min-max. As part of the DOE CSGF program, the PDE-constrained optimization test problems are expanded to account for the previously mentioned uncertainty present within measurements. Additionally, the scalability of these optimization tools are examined, specifically for the computing resources available at Argonne National Laboratory. These uncertainties may be due to noisy measurements, control settings, or uncertainty within the given state. Groundwork for this project was formed during first-year graduate courses at Princeton University, including but not limited to Optimal Control and Estimation, Mathematical Methods for Engineering, Automatic Control Systems and Advanced Orbital Mechanics. The combination of this work and additional courses drives the future research of obtaining minimum radiation orbital transfers for all-electric satellites.
Abstract Author(s): Sarah Gady