Our new approach to model-based estimation, control, and learning fuses ensemble simulations of state, parameter, and control input perturbations with the available measurements, labels, or performance variable precision to update parameters, states, inputs, and their uncertainties in an adjoint-free manner. Fully nonlinear black-box or gray-box” models are admissible, and the quantified posterior uncertainties enables the maximization of information gain for efficient control, estimation, or learning. In this SuperUROP, we develop an embedded GPU-based solution for ensemble simulations for differential equations and apply the technique to real-time aircraft control. We anticipate that the computational core will support other applications in source localization, and coordinated control.”

I look forward to participating in SuperUROP because I have enjoyed my work thus far with ESSG on related projects. I am interested in signal processing and control, and the knowledge I will gain through this program will be invaluable to my continuing education in these areas. As a career, I would like to participate in industrial R&D and getting training in scientific communication through SuperUROP will help me thrive in such environments.