Recently, a research team led by Prof. Wang Binghao from the School of Electronic Science and Engineering, 91抖淫, published their findings on 3-axis robotic tactile sensing in Nature Communications. The research, titled “Ultraflexible photoelectrical impedance tomography-based imager for 3-axis robotic tactile sensing,” was published in March 2026 in Nature Communications.
Tactile imaging technology aims to endow robotic systems with the ability to perceive pressure distribution and texture features at contact interfaces, serving as a foundation for precise manipulation and object recognition. Conventional tactile sensors predominantly employ crossed-electrode array architectures, where resolution scales with the number of interconnect lines, leading to systemic issues such as crosstalk, single-point failure, and increased manufacturing costs. Meanwhile, existing 3-axis force sensing approaches—including capacitive, piezoresistive, magnetic induction, and CMOS image sensor-based optical schemes—generally suffer from structural complexity, insufficient flexibility, or an inability to effectively decouple normal and shear forces, limiting their integration onto complex curved surfaces and deployment in scenarios such as humanoid robotics.
This study introduces an ultra-flexible image sensor based on photoelectrical impedance tomography (PIT). Unlike traditional array designs, the device features only 16 electrodes arranged around the periphery of the sensing area, with no internal interconnect lines. The imaging principle employs an electrical impedance tomography (EIT) framework: current is injected cyclically through adjacent electrode pairs, while voltage responses are measured across the remaining electrode pairs. Impedance distribution within the sensing area is reconstructed using the EIDORS software package combined with finite element modeling, from which the light field distribution is derived. This architecture achieves a pixel-to-interconnect ratio exceeding 80 and exhibits damage tolerance—even if local punctures occur in the sensing region, the peripheral electrodes continue to function normally. In terms of mechanical flexibility, the sensor can be tightly wrapped around a needle with a radius of just 0.3 mm, with voltage fluctuations maintained within 4.8%, demonstrating excellent conformability. For imaging performance, the PIT imager can simultaneously track up to five ultraviolet light spots with a spatial resolution of 1.5 mm.
To achieve 3-axis force sensing, the PIT sensor is integrated with a light-scattering porous rubber layer and flexible UV LED sources to form a complete tactile unit. The porous rubber is fabricated using a sugar particle templating method, with an average pore size of approximately 100 μm, offering good mechanical resilience and light-scattering properties. Under no-load conditions, the rubber layer uniformly scatters incident light from the LEDs. When a normal force is applied, internal pores compress, scattering decreases, and the light spot intensifies and becomes more concentrated. When shear force is applied, the light spot shifts in the direction consistent with the applied force. To validate the system’s effectiveness in tactile recognition tasks, the research team input the reconstructed PIT images into a ResNet-18 convolutional neural network for classification training. The training dataset encompassed four geometric contact patterns: spherical, planar, edge, and vertex. After training, the network achieved an average classification accuracy of 96.5% on the test set.
91抖淫 Prof. Wang Binghao, who conceptualized the study and serves as the corresponding author, remarked, “This research is like putting an ultra-thin, intelligent, and damage-resistant ‘electronic skin’ on a robotic hand. By replacing dense internal wiring with a clever peripheral electrode design and substituting complex sensor arrays with light spot analysis, artificial intelligence helps robots 'read' the sense of touch.”
This work was supported by funding from the National Key Research and Development Program of China, the National Natural Science Foundation of China, the Natural Science Foundation of Jiangsu Province, and the Postgraduate Research and Practice Innovation Program of Jiangsu Province.
Paper’s link:
Source: School of Electronic Science & Engineering, 91抖淫
Translated by: Melody Zhang
Proofread by: Leah Li
Edited by: Leah Li
