Abstract
Large-area, high-speed tactile skins could improve robotics, prosthetics, and human-machine interfaces, but scaling them is limited by wiring complexity and readout bandwidth. We introduce Single-Pixel Tactile Skin (SPTS), a flexible, daisy-chainable tactile array that implements compressive sampling directly in hardware. Each sensing element uses a miniature low-cost microcontroller to apply programmable analog weights, while all pixel currents are summed into a single output channel. Repeating this process yields global projections from which tactile images are reconstructed using sparse recovery, reducing the measurements required relative to raster scanning. In a 10×10 array, SPTS achieved ≥98% object classification accuracy with 20 measurements, corresponding to an effective 3500 FPS, and captured an 8 ms projectile impact in 23 reconstructed frames. Because reconstruction quality improves progressively with measurement count, SPTS can rapidly localize contact from sparse data and refine tactile images over time, enabling scalable, responsive tactile sensing for physical interaction and control.