In a significant advancement for robotic manipulation, researchers at EPFL's Computational Robot Design & Fabrication (CREATE) Lab have developed a robotic hand that can grasp objects with remarkably human-like movements that emerge spontaneously from its design rather than from complex programming.
The ADAPT hand (Adaptive Dexterous Anthropomorphic Programmable sTiffness) utilizes a strategic distribution of compliant materials—primarily silicone strips wrapped around a mechanical structure and spring-loaded joints—to create what researchers call "self-organized" grasps. Unlike traditional robotic hands that require precise information about an object's position and properties, the ADAPT hand can adapt to various objects with minimal input.
"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 Kai Junge, a PhD student working under Professor Josie Hughes at the CREATE Lab.
The hand's design is remarkably efficient, using only 12 motors housed in the wrist to control its 20 joints. The remaining mechanical control comes from springs that can be adjusted for stiffness and from the silicone 'skin' that can be added or removed. In testing, the ADAPT hand successfully picked up 24 different objects with a 93% success rate, using movements that mimicked natural human grasping with 68% similarity.
What makes this development particularly noteworthy is the hand's programming simplicity. It moves through just four general waypoints to lift an object, with any further adaptations occurring automatically without additional programming—what roboticists call 'open loop' control. This allows the hand to adapt its grasp to objects ranging from a single bolt to a banana without reprogramming.
The EPFL team is now building on this foundation by integrating closed-loop control elements, including pressure sensors in the silicone skin and artificial intelligence. This approach could lead to robots that combine compliance's adaptability with precise control, potentially revolutionizing how robots interact with unpredictable environments or spaces designed for humans.