KAIST researchers used machine learning to design a 3D-printable rubber-like material. They worked with teams at KIST and Seoul National University of Science and Technology. The material prints on light-curing printers and stretches to more than six times its length. Printable formulations normally flow well but stretch poorly, and the reverse also holds. The team cured many liquid formulations in molds and measured stretch, hardness, curing speed and flow. Machine learning then picked the best combination from that dataset. Training data included thick formulations that print badly, not only workable ones. The team printed air-driven soft actuators that bend like a finger. A hand built from several actuators gripped eggs, glass bottles and a computer mouse. The work appeared in Nature Communications on 4 June.
What changed
Printable formulations flowed well but stretched poorly, and stretchy ones were too thick to print.
What it unlocks
Designing 3D-printable rubber-like materials by screening formulations with machine learning before lab testing.
- stretches over 6x original length
- lifted a 1 kg water bottle
- published 4 June 2026
Sources