Inverse Design of Kirigami Sheets

Given a target shape and a chosen actuation, find where to cut a flat sheet so that it opens onto that target without tearing its own hinges.

What changed

Neural Form-Finding and Tesseract are snapshots of this work. Two things moved since.

The cut pattern is now parameterised so that deployability holds by construction: every value the optimiser can reach is already a cuttable, deployable sheet, with no penalty and no repair. And the hinges are no longer three springs picked by hand. A neural surrogate supplies the energy each one stores, fitted offline to elasto-plastic solves of a single ligament, about eight orders of magnitude faster than the solve it stands in for. Cut pattern and hinge geometry are now set by one gradient.

One design was cut from polyester and deployed by hand. Its measured shape departs from the prediction by 0.70% of the sheet length.

The defense

I defended on 9 September 2026 at École des Ponts, before a jury chaired by Alain Pecker. He called the work outstanding, and the jury awarded it 19/20.

Thanks

The work was done at the Form Finding Lab at Princeton University between April and August 2026. Four people made it possible.

• Sigrid Adriaenssens, who heads the lab, my professional tutor.
• Rafael Pastrana, Princeton University.
• Isabel Moreira de Oliveira, Princeton University.
• Robin Oval, École des Ponts, my academic tutor.

Thank you for your precious advice.