Dexterous, tactile robot hands are bringing robots closer to one of the most difficult human abilities to reproduce: using our hands to manipulate objects through a combination of movement and touch. Unlike simple industrial grippers that primarily open and close, human-like robotic hands use multiple articulated fingers and joints to grasp, rotate, reposition, and manipulate objects within the hand.
Just as importantly, researchers are covering fingers and palms with tactile sensors that function somewhat like artificial skin, measuring contact, pressure, force, slip, and other information that vision alone cannot provide. Recent research has demonstrated hands with high-resolution tactile sensing distributed across much of the palm and fingers, allowing robots to adjust their grip as an object moves or begins to slip.
Combining this sense of touch with cameras, joint-position sensing, artificial intelligence, imitation learning, and reinforcement learning could allow robots to learn delicate skills such as turning a small object between their fingers, handling fruit without crushing it, using tools, folding materials, or picking up unfamiliar objects without knowing their exact shape beforehand.
Research also shows that combining touch, vision, and proprioception—the robot’s awareness of its own joint positions—can improve dexterous manipulation compared with relying on a single type of sensing. As these technologies improve, tactile human-like hands could become a key part of humanoid and service robots, enabling them to perform the enormous variety of everyday tasks that were originally designed around the capabilities of the human hand.
The video above is intended for educational purposes and does not constitute or imply an endorsement by FutureForAll.org
