Robotic hand breakthrough: enhanced dexterity with tactile fingertips

The recent innovation in robotics has taken a significant leap forward with the development of a four-fingered robotic hand at the University of Bristol. Led by Professor Nathan Lepora in the field of Robotics and AI, this breakthrough accomplishment showcases a robotic hand with artificial tactile fingertips capable of rotating objects in any given direction and orientation, including when the hand is upside down – a feat never achieved before, SSP reported. The findings have been published on the arXiv preprint server for scientific researchers and enthusiasts to explore.

The potential impact of improving the dexterity of robotic hands is noteworthy for various sectors, particularly automation in areas such as handling goods in supermarkets and precise waste sorting for recycling purposes. While previous research, like that conducted by OpenAI in 2019, achieved human-like feats of dexterity, it required extensive resources, including a cage housing 19 cameras and several thousand CPUs. However, OpenAI eventually disbanded their robotics team due to the high cost involved in this complex setup.

Professor Lepora and his team aimed to explore if similar dexterity results could be achieved through simpler and more cost-effective means. Over the past year, several university teams, including MIT, Berkeley, New York (Columbia), and Bristol, have showcased remarkable accomplishments in robot hand dexterity, such as picking up and passing rods or effortlessly rotating children's toys, all without relying on extravagant setups. Instead, these achievements were realized through simple setups and regular desktop computers.

The key enabling factor in these advancements, as mentioned in the recent Science Robotics article titled "The future lies in a pair of tactile hands," was incorporating a sense of touch into the robot hands. The ability to develop highly sensitive tactile sensors became feasible through technological progress in smartphone cameras, which have become so small that they fit comfortably within a robot's fingertip.

In Bristol, researchers successfully implemented artificial tactile fingertips using a 3D-printed mesh of pin-like papillae beneath the artificial skin. This innovative approach involves replicating the intricate internal structure of human skin. Utilizing advanced 3D printing techniques capable of blending both soft and hard materials, the team managed to construct complex structures akin to those found in biological systems.

Excitingly, by leveraging tactile data, the researchers trained the robot hand to perform tasks even when it's upside down or waved around on a robotic arm – a significant milestone. Initially, the robot hand would struggle and drop objects, but through iterative training, they achieved remarkable success.