A mechanochromic tactile skin for ultra-precise, high-dimensional sensing with minimal dataset requirements

Tactile sensing is fundamental to both human interaction and robotic perception, yet existing sensors face challenges in accurately capturing high-dimensional inputs—including 3D force, pressure, contact area, and torsion. Here, we introduce a soft tactile sensor based on structural color, incorporating a photonic crystal architecture for ultra-sensitive detection of nanometer-scale deformation. The sensor achieves a force resolution of 0.018 N over a range of 0.1-1.5 N and a spatial accuracy of 0.015 mm, along with pixel-level precision in contact and pressure mapping. Unlike conventional methods that require large-scale training datasets, our system directly extracts interpretable physical features from polarized color images, reducing data requirements by over three orders of magnitude (to ~1 KB). We demonstrate the sensor’s versatility across tasks such as high-resolution force mapping of irregular surfaces, intuitive single-stick multi-axis control for virtual environments and high dynamic robotic grasping with simultaneous force and pressure feedback. This work offers a scalable tactile sensing strategy and paves the way for integration in data-efficient, high-performance robotic systems.