February 19, 2025
We will resume the seminar this week. Again, please thank Prof. Ed Habtour for connecting us with Dr. Tenavi Nakamura-Zimmerer from NASA Langley. Dr. Nakamura-Zimmerer will present his work on “Research at NASA on data-driven aerodynamic modeling, uncertainty quantification, and adaptive guidance“. There will be food!
Date: Friday Feb 21
Time: 12:30PM – 1:30PM(ish)
Location: SIG 224
Zoom: Zoom Link
Title: Research at NASA on data-driven aerodynamic modeling, uncertainty quantification, and adaptive guidance
Abstract: We discuss two recent fundamental research efforts at NASA related to data-driven dynamics and control. In the first, we present work from the AEROFUSION Early Career Initiative, which aimed to modernize the construction of aerodynamic models at NASA by applying modern machine learning and uncertainty quantification techniques. Among the novel methods developed with AEROFUSION is the Structured Covariance Gaussian Network (SCGN). This algorithm is based on a pair of neural networks which parameterize mean and covariance functions of a multivariate Gaussian process, trained together to maximize the log-likelihood of observing the given data. We also propose a sampling approach that produces viable surrogate function realizations from the Gaussian process. We illustrate the use of SCGNs for learning an aerodynamic response surface with built-in uncertainty for the Orion crew module. We compare results to a baseline Gaussian process regressor and observe that the SCGN provides comparable uncertainty descriptions with improved scalability to dataset size.
In the second part of this talk, we discuss a framework for adaptive guidance in the context of Urban Air Mobility (UAM). In the proposed approach, parameters of a reduced-order performance model are estimated in real time so that trajectories and guidance commands are replanned with more accurate knowledge of the system’s response and performance capabilities. To apply this methodology to flight control systems, we introduce a simple yet expressive performance model modified from a reference linear design model. With a simulation of a UAM-class aircraft, we show that by optimizing performance model parameters in real time, guidance commands can be intelligently adjusted to recover system stability and the performance of a full-order vehicle model, even in the event of effector failures.
Bio: Tenavi Nakamura-Zimmerer is a Research Aerospace Engineer with the Flight Dynamics Branch at NASA Langley Research Center. He currently works with the Intelligent Contingency Management subproject, where he conducts advanced flight controls research for vertical takeoff and landing aircraft. His research interests include optimal feedback control, scientific machine learning, uncertainty quantification, and aerospace applications. He received his Ph.D. in Applied Mathematics and Statistics from UC Santa Cruz in 2022.