Lausanne, Switzerland – Researchers at EPFL have developed a robotic hand capable of picking up 24 different objects with human-like movements, using soft materials and self-organized grasps.

The robotic hand, named ADAPT (Adaptive Dexterous Anthropomorphic Programmable Stiffness), can pick up items without needing to know their exact position in space. Unlike traditional robots that require precise programming, ADAPT uses compliant materials like silicone and spring-loaded joints to enable spontaneous grasping.

In a series of tests, the ADAPT hand successfully picked up 24 objects with a 93% success rate. The grasping movements were found to resemble natural human grasps, with a 68% direct similarity. These “self-organized” grasps emerge automatically, without any additional programming.

The ADAPT hand is powered by just 12 motors and 20 joints, with the rest of its mechanical control coming from the silicone material and springs. The robot uses only four basic positions to lift objects, adapting its grasp based on the shape of the item.

The research, published in Nature Communications Engineering, demonstrates that robots can perform complex tasks using compliance alone, without the need for sophisticated external programming.

Researchers are now working to enhance the ADAPT hand with sensory feedback and artificial intelligence to further improve its ability to handle unpredictable environments.

Podcast: Play in new window | Download