•   When: Friday, March 06, 2020 from 01:30 PM to 02:30 PM
•   Speakers: Julian Panetta
•   Location: Engineering Building 4201
•   Export to iCal

Abstract:
Modern digital fabrication technologies like 3D printers and CNC machines offer great power to produce customized, finely-tuned physical objects with exciting applications in architecture, medicine, robotics, and more. However, searching the vast space of geometries manufacturable by these technologies for the best solution to a given design problem requires sophisticated computational design tools, and this is the main bottleneck preventing us from leveraging these technologies' full potential.

My talk will present my work on inverse design problems for digital fabrication: developing algorithms that take users' high-level specifications of desired functionalities of a deformable object and solve for the optimal manufacturable geometry. These problems require new algorithms for fast, accurate, and differentiable physical simulation and robust numerical optimization to predict the functionalities of candidate designs and determine how to improve them. The highly nonconvex optimization problems involved also necessitate new computational strategies to produce good initial guesses (which we develop using insights from differential geometry) and to escape from local optima. I will discuss the software and algorithms I have created for optimization-based inverse design in the context of two specific applications: elastic metamaterials and deployable structures.

Elastic metamaterials are fine-scale spatial arrangements of fabrication materials that can exhibit properties not found in any natural material and achieve greater strength-to-weight ratios than pure solids. Metamaterials have promising applications in fields like prosthetics, soft robotics, and aerospace, and are a perfect setting to develop new computer graphics and geometric modeling techniques: their properties depend only on the shape and topology of the material arrangement. Deployable structures are shape-shifting objects that can transition from a compact configuration (efficient to fabricate and transport) to a desired 3D shape. They have applications in many settings from emergency shelters to medical implants and satellite antennas.

Bio:
Julian Panetta is currently a postdoc at the École Polytechnique Fédérale de Lausanne in Switzerland supervised by Mark Pauly. He received his PhD in computer science from NYU's Courant Institute, where he was advised by Denis Zorin. Julian is interested in physical simulation, geometry processing, and optimization-based inverse design problems, specifically focusing on applications for 3D printing, metamaterials, and deployable structures. Before NYU, he received his BS in computer science from Caltech and did research at NASA's Jet Propulsion Lab.