Shear-Based Grasp Control Enables Human-Like Dexterity in Robotic Hands

Researchers from MANiBOT partner University of Bristol, with the Istituto Italiano di Tecnologia and University of Pisa, have developed a novel control framework for robotic hands that enhances dexterity and safety in handling delicate objects.

Using the Pisa/IIT SoftHand equipped with soft biomimetic tactile sensors—miniaturised versions of the vision-based TacTip—they created a system that senses contact pose and force at all five fingertips.

The team trained deep learning models to interpret tactile data across sensors and fed this into a grasp control scheme that adapts in real time. This approach allows the hand to react reflexively to disturbances, retaining grip without damaging fragile items.

Demonstrations included dynamic tasks like holding a flexible cup, pouring while its weight shifted, and following human-led object movement.

This advancement brings robotic hands closer to human-like manipulation, with potential applications in service robotics, prosthetics, and delicate automation tasks.

The study was conducted by Christopher J. Ford, Haoran Li, Manuel G. Catalano,Matteo Bianchi, Efi Psomopoulou, and Nathan F. Lepora

Read it in detail in pre-print here.