On June 1, reporters learned from the University of Science and Technology of China (USTC) that the School of Engineering Science and the Institute of Humanoid Robotics of the university have recently developed a new type of giant piezocapacitive sensor patterned after fish scales. The sensor enables robots to detect tiny surface textures, distinguish the hardness of objects, and perform gripping, sorting and human-machine interaction. Relevant research findings have been published in the international journal Advanced Materials.

For robots, visual recognition of an object does not equate to genuine comprehension of its properties. Vision alone often falls short when gripping soft items or objects with fine surface textures such as fruits and fabrics. Judging fruit ripeness, surface roughness of materials and proper gripping force relies on more sensitive and stable tactile sensors.

The research team drew inspiration from fish scales in this work. Fish skin consists of relatively hard scales and soft dermis, which safeguards the fish’s body while sensing external stimuli. Inspired by this structure, the team fabricated an electric-field regulating membrane mimicking fish scales, where rigid scale-like units are arranged on a flexible substrate with micrometer-scale air gaps between adjacent units. Upon slight bending, pressing or sliding of the sensor, these gaps function as electric field gates, converting minuscule physical deformation into distinct variations in electrical signals.

This bio-inspired fish-scale architecture forms the core of the novel giant piezocapacitive sensor. Named GPCS, the sensor built on the above working mechanism serves as highly sensitive electronic skin for robots. Experimental data proves the sensor can identify ultra-small angular shifts under extensive bending with an angular resolution of 0.005° and a rapid response time of only 0.6 milliseconds. Therefore, robots can promptly capture tiny surface undulations and force shifts during grasping and sliding contact to execute sophisticated motion control. In addition, the sensor maintains stable performance after 100,000 cyclic bending tests. Its bending sensitivity is 177 times higher than that of conventional counterparts, and it remains operational even with partial structural damage.

Following performance verification, researchers integrated the sensor onto flexible bionic robot fingers to act as artificial tactile nerves. Sliding the fingertip across diverse material surfaces triggers signal fluctuations corresponding to microscopic surface irregularities, generating unique tactile fingerprints for different substances.

In lab tests, the sensor-equipped flexible fingertip detects microstructured surfaces with varying height and periodicity and identifies printed toner textures as thin as 1.8 micrometers, outperforming human fingertips in tactile discrimination. A single fingertip is capable of distinguishing as many as 16 types of fabric textures.

The sensor is also applicable to robotic gripping tasks. The research team assembled four GPCS sensors into an array and embedded the array inside flexible grippers, the robotic equivalent of human hands. The sensor array monitors subtle real-time deformation during clamping operations. Kiwis of disparate ripeness levels feature varying firmness, leading to differing gripper deformation upon grasping. The sensors capture such discrepancies, and machine learning algorithms analyze the data to assess fruit ripeness with an identification accuracy of 92%.

According to the team, the research group has long focused on developing highly sensitive and robust flexible sensors, whose existing achievements have been deployed in robotic tactile sensing, shape reconstruction of flexible pneumatic fingers and intelligent fragile-object gripping. The latest research expands the application scope of flexible sensors in precision manipulation and intelligent perception for robots and delivers a new technical solution for robots to adapt to complicated real-world scenarios in the future.

Source: Hefei Daily