AI-Optimized 3D-Printed Material Powers Soft Robotic Hand for Versatile Grasping.

Researchers in South Korea have successfully engineered an artificial intelligence-driven method to create a novel polymer capable of extreme stretchability while maintaining high 3D printability. The breakthrough, detailed in Nature Communications, was spearheaded by Professor Seungchul Lee of the Korea Advanced Institute of Science and Technology, alongside Dr. Jongbeom Na from the Korea Institute of Science and Technology and Professor Bumsoo Park from the Seoul National University of Science and Technology. The innovation resolves a persistent manufacturing dilemma: balancing fluidity with mechanical resilience. Digital Light Processing 3D printers cure liquid resins into precise geometries, but increasing a material's elasticity typically raises its viscosity beyond workable limits, hindering flow and curing. Diluting the resin improves printability but weakens the final product. To eliminate this trade-off, the consortium developed an AI-based material design framework. Researchers experimentally tested diverse resin formulations, recording cure rates, viscosity, tensile strength, and elasticity, deliberately including highly viscous compounds that traditionally fail printing. Machine learning algorithms processed this dataset to map formulation variables against performance outcomes, ultimately predicting an optimal chemical composition that satisfied both printability and flexibility requirements. The resulting material stretches over six times its original length without tearing. To demonstrate practical utility, the team manufactured soft pneumatic actuators using standard DLP hardware. Under air pressure, these actuators replicated biological muscle contraction, bending naturally. Assembling multiple actuators produced a soft robotic hand capable of securely lifting a one-kilogram water bottle while gently grasping fragile items like raw eggs, glass bottles, and irregularly shaped objects without damage. This research establishes a scalable paradigm for accelerated materials discovery. By substituting traditional trial-and-error experimentation with predictive computational modeling, developers can rapidly identify promising formulations using empirical data, drastically reducing iteration cycles. The approach significantly lowers the barrier to entry for advanced additive manufacturing, enabling broader adoption of soft robotics, adaptive wearable technology, and customized medical devices. By demonstrating how machine learning can bridge experimental data and material engineering, the study positions AI as a foundational tool for next-generation manufacturing and biomedical innovation.