Scientists develop human-like robotic hand with smart palm-finger coordination

BEIJING -- Imagine a robotic hand that coordinates its palm and fingers with human-like precision -- not only grasping objects with perfect stability but also "feeling" their 3D contours and distinguishing between different materials: This vision has become reality thanks to groundbreaking research from Shanghai Jiao Tong University.

The innovation solves a fundamental challenge in robotics. While soft robotic hands with sensing capabilities have shown promise, existing designs primarily focused on finger sensitivity while neglecting the palm's crucial role in manipulation.

The palm, as a large and crucial contact region providing mechanical support and sensory feedback, was -- up until this point -- underexplored due to limited sensing density and interaction with the fingers.

The research team overcame these limitations by developing a fully integrated system where the palm and fingers work in perfect harmony, achieving unprecedented levels of sensitivity and control.

At the heart of this advancement lies three groundbreaking features: a high-resolution tactile palm with 181,000 sensing units per square centimeter (754 times more sensitive than human skin), remarkably flexible fingers capable of 204.3 degrees of bending, and intelligent coordination algorithms that enable seamless palm-finger interaction.

The palm's sophisticated visual-tactile system combines micro-cameras with multi-layer elastic sensors, while the fiber-reinforced pneumatic fingers provide both strength and delicacy with 14.6-newton grasping force.

This integration allows the robotic hand to perform tasks with remarkable human-like capability. In demonstrations, it executed complex maneuvers like retrieving cards with perfect precision, detecting microscopic flaws in fabrics, and even pouring tea while automatically adjusting its grip to prevent spills.

The research team has enhanced the system's capabilities by integrating advanced AI technology, dramatically improving its object recognition performance. In challenging blind tests, the robotic hand demonstrated remarkable precision by identifying and retrieving specific target items from opaque bags filled with assorted small objects -- from tiny peas and sunflower seeds to larger nuts and bolts.

Impressively, the system achieved 88 percent overall recognition accuracy, reaching 100 percent performance for distinctive objects like sunflower seeds.

"Our results suggest the promise of fusing rich palm tactile sensing with soft dexterous fingers for advanced interactive robotic operations," the researchers noted.

The study has been published in the journal Nature Communications.