KAIST Researchers 3D-Print a Robotic Hand Using AI-Designed Rubber

Finding the Right Recipe

A research team at KAIST used machine learning to identify a new material formulation for Digital Light Processing 3D printing. The result is an elastomer that stretches to more than six times its original length without tearing, and the team demonstrated it by printing a soft robotic hand.

The challenge in DLP printing has always been a tradeoff. Materials that are durable and stretchable often become too viscous to flow properly during printing. Materials that print easily tend to be weak. The KAIST team fed experimental data from hundreds of material formulations into a machine-learning model, letting the AI find a formulation that balanced printability with stretchability.

A Hand That Can Grasp Eggs and Water Bottles

The researchers printed soft actuators that bend like human fingers when inflated with air. They combined several of these actuators into a robotic hand that lifted a 1-kilogram water bottle, grasped an egg, held a glass bottle, picked up an egg carton, and lifted a computer mouse.

The work, published in Nature Communications, shows how AI can accelerate materials discovery without relying on pure trial and error. Instead of testing thousands of combinations by hand, the model predicts promising formulations that researchers then verify in the lab.

Beyond Robotics

The team sees applications in wearable devices, custom medical equipment, and any system that needs to conform to irregular surfaces while remaining durable. The same machine-learning approach could speed up development of other printable materials with different mechanical requirements.