Analog Dual-Tone Multifrequency Signaling for Scalable Tactile Skin With Automatic Data Compression and a Single-Wire Output

Le résumé fourni par la source

Achieving human-scale tactile sensing in robotics has been largely impeded mainly by wiring and latency limitations resulting from the ever-growing number of sensors and associated sampling delays. Several multiplexing schemes have been developed at the sensor level to accommodate more sensors, embedding intelligent communication directly into the sensors via microcontroller interfaces. However, integrating microcontrollers and circuit elements into each sensing unit imposes fundamental constraints on sensor spacing and wiring density, increases fabrication cost, and reduces flexibility and durability. In this work, we propose a novel interfacing scheme that achieves high scalability through frequency multiplexing and automatic data compression, without embedding electronics into each sensor. Our scheme leverages an analog adaptation of dual-tone multi-frequency signaling (aDTMF) for scalable, multiplexed compressed sampling of large tactile arrays. To validate our approach, we developed a 16×16 prototype sensor (256 pixels) that compresses all sensor readings into just 32 row/column frequency magnitudes (16 row + 16 column carriers) summed onto a single output wire, and evaluated its performance in object classification tasks; comparing its response time and accuracy to a conventional raster-scanned sensor. The results demonstrate that the aDTMF sensor improves detection speed by achieving 80% classification accuracy within ~10 ms, compared to ~40 ms for the raster-scanned baseline. While the aDTMF sensor achieved 95.6% classification accuracy using just 32 output values, a raster-scanned sensor, reduced to the same number of outputs (via spatial down-sampling), achieved only 7%, highlighting the effectiveness of aDTMF’s compressed representation. Furthermore, we show aDTMF’s novel applications to sensing and robotic tactile perception systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.