Skip to main content

Exploiting Convex Structure in Aircraft Design Optimization

Speaker: Warren Hoburg, Dept. of Aeronautics and Astronautics, Massachusetts Institute of Technology

Presentation

Seminar Slide Deck (PDF-3.7MB)

Abstract

There is a deep connection between the mathematical notion of convexity, and practical tractability of solving an optimization problem. Thanks to new primal-dual interior point methods developed and refined over the past two decades, many classes of convex optimization problem can now be solved in polynomial time on a laptop computer. One such problem class, the geometric program (GP), can be leveraged to reliably and efficiently explore conceptual aircraft design spaces. This approach provides a rich formalism for casting engineering design problems, including guarantees of global optimality, sensitivity analysis via Lagrange duality, and fitting of GP-compatible surrogate models to data. Limitations and extensions to non-convex design spaces will also be discussed.

Biography

Photo of Warren Hoburg
Prof. Warren Hoburg is an assistant professor of Aeronautics and Astronautics as MIT. He completed his Ph.D. in Electrical Engineering and Computer Science at Berkeley in 2013 under the supervision of Pieter Abbeel. He received his BS in Aerospace Engineering from MIT. He spent 2013-2014 at Boeing Commercial Airplanes Product Development working on optimization problems in composite manufacturing. His current research interests include convex optimization, operations research, and decision-making in aerospace and manufacturing systems.