Matt Strong

Matt Strong

CS PhD at Stanford University

← Paper Digest

ReSkin: versatile, replaceable, lasting tactile skins

Raunaq Bhirangi, Tess Hellebrekers, Carmel Majidi, Abhinav Gupta

Seminal October 2021

TL;DR A 2-3 mm magnetic skin you can peel off and replace: magnetized particles in elastomer over a few magnetometers, with learned (and self-supervised) models that stay accurate across replacement skins and 50k+ contacts.

What they do

Soft tactile skins wear out, and every replacement responds slightly differently, so a model calibrated on one skin fails on the next. ReSkin separates the cheap, wearable part (a magnetized elastomer) from the electronics (magnetometers on a board), and uses machine learning to make readings consistent across skins and over time.

How it works

  • Magnetic microparticles are mixed into an elastomer, which is cast and magnetized. Pressing the skin moves the particles and changes the field seen by five magnetometers under a 20x20 mm patch.
  • The skin is 2-3 mm thick and samples at up to 400 Hz. They show it on a robot gripper, an arm sleeve, a glove and a dog shoe.
  • A network maps the 15 magnetic readings to contact location and force. A self-supervised triplet loss and a small adaptation set let a model transfer to a new skin with little recalibration.

Why it matters

ReSkin made magnetic tactile skins practical and cheap, and led to follow-ups such as AnySkin. It is the standard reference for the replaceable, low-profile alternative to camera-based sensors like DIGIT and GelSight.

Limitations

Spatial resolution is limited by the number of magnetometers (about 1 mm localization at 90% accuracy). Readings are sensitive to nearby ferromagnetic objects and external fields.

Videos

Full video (YouTube)
Full video (YouTube)

Figures

Figure 1: A) ReSkin is easy to fabricate and the size of a penny, enabling a wide range of applications. B) Robot gri...
Figure 1: A) ReSkin is easy to fabricate and the size of a penny, enabling a wide range of applications. B) Robot gripper using tactile feedback from ReSkin sensors to hold a blueberry without squishing it. C) Dog shoe with an embedded ReSkin sensor; (inset) visualization of sensor measurements. D) Contact localization on a new ReSkin sensor using our self-supervised adaptation procedure. E) Contact localization on a ReSkin curated into a fabric sleeve as a 2in x 4in contiguous skin. F) ReSkin sensor as a fingertip sensor to record forces and contacts while folding a dumpling
Figure 3: A) Experimental setup for data collection with Dobot Magician, ATI Nano 17 (inset), and six sensor boards s...
Figure 3: A) Experimental setup for data collection with Dobot Magician, ATI Nano 17 (inset), and six sensor boards streaming to a control computer. B) Mold for curing elastomer along with magnet holders. C) Two types of circuit boards – rigid and flexible – designed to be used with ReSkin.
Figure 2: ReSkin is replaceable!
Figure 2: ReSkin is replaceable!
Abstract

Soft sensors have continued growing interest in robotics, due to their ability to enable both passive conformal contact from the material properties and active contact data from the sensor properties. However, the same properties of conformal contact result in faster deterioration of soft sensors and larger variations in their response characteristics over time and across samples, inhibiting their ability to be long-lasting and replaceable. ReSkin is a tactile soft sensor that leverages machine learning and magnetic sensing to offer a low-cost, diverse and compact solution for long-term use. Magnetic sensing separates the electronic circuitry from the passive interface, making it easier to replace interfaces as they wear out while allowing for a wide variety of form factors. Machine learning allows us to learn sensor response models that are robust to variations across fabrication and time, and our self-supervised learning algorithm enables finer performance enhancement with small, inexpensive data collection procedures. We believe that ReSkin opens the door to more versatile, scalable and inexpensive tactile sensation modules than existing alternatives.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.