Researchers have introduced a new robotic hand, known as the F-TAC Hand, which features high-resolution tactile sensors embedded across 70% of its surface area. Designed to improve dexterous manipulation in changing environments, the system was presented in a study published today in Nature Machine Intelligence.
The F-TAC Hand incorporates tactile sensors with a spatial resolution of 0.1 millimetres, enabling detailed perception of contact surfaces. According to the research team, this level of tactile feedback has not been achieved previously in robotic hands, which have typically struggled with fine manipulation tasks in unpredictable settings due to limited sensory input.
The system combines this tactile sensitivity with generative algorithms that replicate human-like hand configurations, aiming to improve adaptability during object handling. In testing, the F-TAC Hand completed 600 real-world trials involving complex grasping tasks. Performance analyses indicated that the system outperformed robotic hands without tactile sensing, with statistically significant results reported (p<0.0001).
Professor Kaspar Althoefer, Director of the Centre of Excellence in Advanced Robotics at Queen Mary University of London, described the technology’s potential to improve manipulation accuracy and object understanding during contact. He stated that the combination of broad tactile coverage and improved perception algorithms contributes to an enhanced capability in interpreting object properties.
The research traces its origins to earlier work at Queen Mary University, where camera-based tactile sensors with high spatial resolution were developed under the supervision of Professor Althoefer. The sensors and methodologies refined during the doctoral research of Wanlin Li form the technical basis of the new system.
The study suggests applications for the F-TAC Hand in areas requiring precise object handling, including manufacturing, assistive robotics, and collaborative human-robot environments.
The article, titled Embedding high-resolution touch across robotic hands enables adaptive human-like grasping, is available through Nature Machine Intelligence.
