Toward Trusted Autonomous Space Guidance: Meta-Reinforcement Learning, Certification, and Adaptive Lunar Landing
Instructions:
- Add this to your calendar for a convenient 15-minute reminder.
- Enjoy!
Abstract:
Future space missions will require guidance systems that can operate autonomously while remaining robust to uncertainty, disturbances, modeling errors, and off-nominal conditions. Reinforcement learning (RL) and meta-reinforcement learning (meta-RL) offer promising approaches for developing such capabilities, with meta-RL providing the additional ability to adapt guidance behavior as vehicle dynamics or environmental conditions change. Recent results have demonstrated improved generalization across uncertainties and actuator and sensor failures.
This webinar will present recent developments in learning-enabled space guidance, focusing on three complementary areas. First, applications of RL and meta-RL to planetary landing, rendezvous, proximity operations, and interception will be reviewed. Second, approaches for the verification and certification of black-box learned guidance policies will be discussed, including the use of Sparse Identification of Nonlinear Dynamical Systems (SINDy) to derive interpretable models that support analysis and validation. Finally, recent work on adaptive lunar powered-descent guidance will be presented, where meta-RL dynamically schedules the guidance gains and time-to-go of an analytical fractional-polynomial guidance law while retaining computational efficiency, physical constraints, and interpretability.
Together, these efforts illustrate complementary pathways toward adaptive, robust, and certifiable autonomous space guidance.
About the Speaker:
Dr. Roberto Furfaro is a Full Professor in the Departments of Systems and Industrial Engineering and Aerospace and Mechanical Engineering at the University of Arizona, where he also serves as Deputy Director of the Space, Security, Safety & Sustainability Center (S4C). He received his Laurea Degree in Aeronautical Engineering from the University of Rome “La Sapienza” and his Ph.D. in Aerospace Engineering from the University of Arizona. Dr. Furfaro is internationally recognized for his work in space systems engineering, guidance, navigation and control, artificial intelligence for aerospace systems, and Space Situational Awareness. His research spans autonomous guidance and control for planetary landing and hypersonic vehicles, close-proximity operations around small bodies, and machine-learning-enabled space autonomy and space domain awareness. He has served as Principal Investigator or Co-Principal Investigator on numerous NASA, AFRL, and other federally sponsored programs, with total awarded funding exceeding $80M, and has authored more than 120 peer-reviewed journal articles and over 300 conference papers and abstracts.
From 2011 to 2016, Dr. Furfaro served as Systems Engineering Lead for the Science Processing and Operations Center of NASA’s OSIRIS-REx Asteroid Sample Return Mission. He currently leads the Target Follow-up Working Group for NASA’s NEO Surveyor Mission and serves as Chair of the International Neural Network Society section on Artificial Intelligence for Space Systems Engineering. He is a Fellow of the American Astronautical Society and an Associate Fellow of the American Institute of Aeronautics and Astronautics. In recognition of his