Abstract: Future mechanical and aerospace systems must operate both efficiently and safely in dynamic, uncertain environments. Although advances in AI, sensing, and actuation offer new avenues for autonomy, current methods frequently struggle with real-time constraints, limited data, and the need for stringent safety guarantees. This talk introduces a Trustworthy AI framework that integrates geometric methods, control theory, and machine learning to address these challenges in aerospace applications.Specifically, I will highlight Geometric Policy Optimization (Geometric PO), a novel technique that leverages the underlying problem structure to enhance computational and learning efficiency, ensure performance and stability guarantees, and accommodate policy constraints. I will demonstrate its capabilities using an estimation-control duality approach for inferring an aircraft’s wing-wave behavior under unknown gusts. Additionally, I will introduce a risk-aware control framework that ensures resilience against extreme events modeled by heavy-tailed distributions, enabled by a risk-constrained extension of Geometric PO. Finally, I will show how these techniques extend to multi-agent coordination, thereby facilitating scalable autonomy in complex systems. By synthesizing learning, control, and geometry, this integrated approach advances the foundation for safe, efficient, and resilient autonomy in next-generation aerospace systems.

Bio: ShahriarTalebi is a researcher at Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS), Harvard Faculty of Arts and Sciences (FAS), and the NSF AI Institute in Dynamic Systems (Dynamics AI). He earned his Ph.D. in aeronautics and astronautics from the University of Washington in 2023, focusing on control, alongside an M.Sc. in mathematics specializing in differential geometry. His research develops rigorous frameworks that integrate geometry, machine learning, and control to advance data-driven autonomy. Through geometric learning methods for control and inference under uncertainty, his research contributes to resilient and adaptable estimation and decision-making strategies. He has been recognized for his excellence in teaching with the 2022 Excellence in Teaching Award at the University of Washington. He is also a recipient of the William E. Boeing Endowed Fellowship, Paul A. Carlstedt Endowment, and Latvian Arctic Pilot–A. Vagners Memorial Scholarship (2018–2019), as well as the Frank Hubbard Engineering Scholarship (2017).