Abstract: System designers and engineers frequently rely on brainstorming and intuition to tell them what's likely to work when confronted with a new problem, but this means that new technologies often fail to live up to their full potential as they're constrained to operate within familiar architectures. This presentation introduces GEMINI, the Generative Evolutionary Mixed-Integer Network Interpreter, which leverages advances in high-performance computing, generative neural networks, and Bayesian optimization to formally tackle configuration design problems which are traditionally handled intuitively. This allows exploration and analysis of never-before-considered designs without relying on the status quo as a starting point. Inspiration is drawn from and special attention is given to configuration and system architecture problems for electric vertical take-off and landing (eVTOL) aircraft aimed at urban air mobility (UAM) missions.
About the presenter: Jordan Smart is a Ph.D. Candidate in the department of Aeronautics and Astronautics at Stanford University, advised by Juan J. Alonso in the Aerospace Design Lab. His research focuses on the use of artificial intelligence within aircraft design, particularly for mixed-integer optimization problems. He is a fellow of the Stanford Vice Provost for Graduate Education's DARE program, the National Science Foundation's Graduate Research Fellowship Program, and the Stanford School of Engineering's Graduate Engineering Fellowship Program. His other awards include the James W. Lyons Award, the Robert H. Cannon, Jr. Summer Doctoral Fellowship, the Stanford Community Impact Award, and the JEDI Service Award. Prior to beginning his graduate education he held several engineering roles at Lockheed Martin, working on the Trident D5 missile program and the F-35 Lightning II Joint Strike Fighter.