In context
By early 2025, robotic tactile sensing faced persistent trade-offs between array density and refresh rate, and single-modality sensors struggled with external interference and limited data dimensionality. This work, published in the Chinese journal Robot, addressed these gaps by proposing a multimodal electronic skin inspired by human skin structure.
What was reported
Researchers from Nanchang University and Southeast University developed a multimodal flexible electronic skin that integrates a magnetic-particle-doped silicone layer with a distributed flexible tactile sensing array. This layered design enables simultaneous collection of magnetic tactile and force-tactile array information in both time and space, overcoming the weak spatio-temporal pairing common in traditional systems.
The sensing layer combines an 8×8 flexible piezoresistive array (printed with nano force-sensitive materials) with five Hall sensors beneath the magnetic elastomer. The piezoresistive array offers linear output, high resolution, and no electromagnetic interference, while the magnetic layer provides full-surface continuous sensing. A CNN-SVM-MLP fusion algorithm was built to efficiently merge multimodal haptic data for object recognition.
In experiments with both non-magnetic and magnetic objects, the multimodal e-skin effectively improved the magnetic sensor's resistance to magnetic field interference. The system accurately recognized 24 types of objects with an accuracy of 99.42%.
Why it mattered
This work demonstrated a practical route to robust, high-accuracy tactile perception for robots in unstructured environments, particularly where magnetic interference is a concern. The integration of heterogeneous sensing modalities on a single layer could advance human-robot interaction, adaptive grasping, and surgical applications by providing richer, more reliable tactile data.
“The proposed multimodal flexible e-skin can effectively improve the magnetic haptic sensor's ability to resist magnetic field interference, and accurately recognize 24 types of objects with an accuracy of 99.42%.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-01-15 · “面向机器人交互识别的抗干扰多模态柔性电子皮肤”