Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GelSlim 4.0: Focusing on Touch and Reproducibility

Andrea Sipos, William van den Bogert, Nima Fazeli

TL;DR

GelSlim 4.0 addresses the reproducibility and manufacturability gap in visuotactile sensing by delivering a modular, affordable hardware redesign with planar lenses, plus open-source perception algorithms for depth and shear estimation. It provides extensive CAD data in OnShape, a complete manufacturing manual, and a dataset of GelSlim 4.0 images across multiple sensors and objects. The depth estimator uses a learning-based RGB-to-depth mapping via a U-Net, while the shear field is derived from dot-pattern optical flow and represented as a 2D vector field. A human-subject study with 17 novices demonstrates reproducibility across tasks and documents changes to improve usability, aiming to democratize tactile sensing in education and research.

Abstract

Tactile sensing provides robots with rich feedback during manipulation, enabling a host of perception and controls capabilities. Here, we present a new open-source, vision-based tactile sensor designed to promote reproducibility and accessibility across research and hobbyist communities. Building upon the GelSlim 3.0 sensor, our design features two key improvements: a simplified, modifiable finger structure and easily manufacturable lenses. To complement the hardware, we provide an open-source perception library that includes depth and shear field estimation algorithms to enable in-hand pose estimation, slip detection, and other manipulation tasks. Our sensor is accompanied by comprehensive manufacturing documentation, ensuring the design can be readily produced by users with varying levels of expertise. We validate the sensor's reproducibility through extensive human usability testing. For documentation, code, and data, please visit the project website: https://www.mmintlab.com/research/gelslim-4-0/

GelSlim 4.0: Focusing on Touch and Reproducibility

TL;DR

GelSlim 4.0 addresses the reproducibility and manufacturability gap in visuotactile sensing by delivering a modular, affordable hardware redesign with planar lenses, plus open-source perception algorithms for depth and shear estimation. It provides extensive CAD data in OnShape, a complete manufacturing manual, and a dataset of GelSlim 4.0 images across multiple sensors and objects. The depth estimator uses a learning-based RGB-to-depth mapping via a U-Net, while the shear field is derived from dot-pattern optical flow and represented as a 2D vector field. A human-subject study with 17 novices demonstrates reproducibility across tasks and documents changes to improve usability, aiming to democratize tactile sensing in education and research.

Abstract

Tactile sensing provides robots with rich feedback during manipulation, enabling a host of perception and controls capabilities. Here, we present a new open-source, vision-based tactile sensor designed to promote reproducibility and accessibility across research and hobbyist communities. Building upon the GelSlim 3.0 sensor, our design features two key improvements: a simplified, modifiable finger structure and easily manufacturable lenses. To complement the hardware, we provide an open-source perception library that includes depth and shear field estimation algorithms to enable in-hand pose estimation, slip detection, and other manipulation tasks. Our sensor is accompanied by comprehensive manufacturing documentation, ensuring the design can be readily produced by users with varying levels of expertise. We validate the sensor's reproducibility through extensive human usability testing. For documentation, code, and data, please visit the project website: https://www.mmintlab.com/research/gelslim-4-0/
Paper Structure (23 sections, 2 equations, 6 figures, 2 tables)

This paper contains 23 sections, 2 equations, 6 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Robotic Skins and Taxel-Based Sensors
  4. Visuotactile Sensors
  5. METHODOLOGY
  6. GelSlim 4.0 Updates
  7. GelSlim Algorithms
  8. Depth Estimation
  9. Shear Field Estimation
  10. Human Study on GelSlim 4.0 Reproducibility
  11. Experimental Design
  12. Participants
  13. Procedure
  14. EXPERIMENTS AND RESULTS
  15. GelSlim Algorithms
  16. ...and 8 more sections

Figures (6)

  • Figure 1: a) GelSlim 4.0 with shear tracking markers mounted on a WSG-50 gripper. b) 3D printed Block M next to a dime for scale. c) Block M indented into a GelSlim 4.0 with shear tracking markers. d) Block M indented into a GelSlim 4.0 without shear tracking markers.
  • Figure 2: Exploded diagram of the GelSlim 4.0 for the WSG-50 Gripper.
  • Figure 3: Results of depth estimation using our proposed method. The rows in each section consist of a photo of the real object, the distorted RGB difference image obtained from the GelSlim 4.0 sensor, and the estimated rectified depth image. Left: 3D printed button, ping pong ball, 3D printed asterisk shape. Middle: 3D printed Y shape, 3D printed T shape, 3D printed cylindrical peg. Right: Marble, necklace pendant, small zip tie.
  • Figure 4: Distorted RGB images and their shear fields obtained while a robot performs a peg insertion task: first contact, hole exploration, and insertion.
  • Figure 5: Prior experience of study participants across all versions
  • ...and 1 more figures