Researchers from the Computational Robot Design & Fabrication (CREATE) Lab at EPFL have achieved a major breakthrough in biomimetic robotics with their ADAPT Hand (Adaptive Dexterous Anthropomorphic Programmable sTiffness) that can grasp objects with human-like dexterity without requiring precise environmental information or complex programming.
The ADAPT Hand, developed by Kai Junge and Josie Hughes, successfully picked up 24 different objects with a remarkable 93% success rate in experimental tests. What makes this achievement particularly significant is that the hand's movements emerge spontaneously through the interaction between its compliant materials and the objects being manipulated, rather than through explicit programming.
"As humans, we don't really need too much external information to grasp an object, and we believe that's because of the compliant—or soft—interactions that happen at the interface between an object and a human hand," explains Junge, whose research was published in Nature Communications Engineering on May 13, 2025.
Unlike traditional robotic hands that require motors for each joint, the ADAPT Hand uses only 12 motors for its 20 joints, with the remaining mechanical control coming from springs and silicone 'skin' that can be tuned to different levels of stiffness. The hand is programmed to move through just four general positions to lift an object, with any further adaptations occurring without additional programming—what roboticists call 'open loop' control.
The distributed compliance throughout the hand—in the skin, fingers, and wrist—allows it to self-organize its grasp based on object geometry, automatically exhibiting different grasp types depending on what it's picking up. When compared to human grasping patterns, the ADAPT Hand achieved a direct similarity of 68% to natural human grasps.
This approach bridges the gap between artificial and biological systems, potentially revolutionizing applications in prosthetics, healthcare robotics, and industrial automation. By mimicking the biomechanical properties of human hands through materials rather than complex algorithms, the ADAPT Hand demonstrates how robots can develop more natural interactions with the physical world—essential for the next generation of assistive technologies.