GelSlim: A High-Resolution, Compact, Robust, and Calibrated Tactile-sensing Finger
TL;DR A GelSight-style tactile finger made slim for grasping in clutter by routing light through acrylic guides and imaging the gel via a mirror, with a fabric-covered gel and an image calibration that offsets wear. After 3000 grasps fewer than 0.06% of its pixels had died.
What they do
Earlier GelSight sensors point a camera straight at the gel and light it with several colors for photometric 3D reconstruction, which makes them bulky cuboids. GelSlim drops precise 3D reconstruction and captures only texture and contact shape, so the optics fold into a thin, pointed finger that can squeeze between objects in a bin. It is also built tougher, and a calibration keeps its images consistent as it wears.
How it works
- Optics. A front-surface glass mirror reflects the gel image to a Raspberry Pi Spy Camera set back in the finger, giving a larger pad without a thicker finger. Mirror and camera angles trade finger thickness against gel length, within the camera’s field of view and depth of field.
- Illumination. On each side, two white LEDs sit at the focus of a parabolic reflection inside acrylic wave guides, so light enters the gel nearly parallel and then crosses it after a mirror-painted 90° bounce.
- Gel. A slightly harder silicone (15:1 XP-565) with semi-specular paint, covered by a loose-weave fabric that protects it and concentrates pressure, so even flat objects leave a contact patch.
- Calibration. Once after fabrication: warp the perspective using a four-square pattern and subtract the background illumination. During use: re-subtract the current background, restore the reference brightness and apply adaptive histogram equalization for contrast.
Results
- Setup. Two GelSlim fingers on a WSG-50 gripper and ABB arm ran over 3300 grasp-lift-vibrate trials, imaging calibration targets every 100 grasps. One prototype only.
- Signal. Raw brightness and signal strength both drop after about 1750 grasps; after correction both stay more consistent.
- Wear. After 3000 grasps fewer than 0.06% of pixels (around 170) are dead from worn paint. Calibration only marginally evens out signal strength across a 5x5 ball array.
Limitations
No 3D reconstruction. The finger is slimmer but not smaller, and the camera’s field of view and depth of field make smaller versions hard. The tip, where contact is richest, is rigid and unsensed, which also hurts durability. The sensor still ages, and clustered dead pixels mean replacing the gel. Grasp control with the sensor is proposed but not demonstrated.
Figures
Abstract
This work describes the development of a high-resolution tactile-sensing finger for robot grasping. This finger, inspired by previous GelSight sensing techniques, features an integration that is slimmer, more robust, and with more homogeneous output than previous vision-based tactile sensors. To achieve a compact integration, we redesign the optical path from illumination source to camera by combining light guides and an arrangement of mirror reflections. We parameterize the optical path with geometric design variables and describe the tradeoffs between the finger thickness, the depth of field of the camera, and the size of the tactile sensing area. The sensor sustains the wear from continuous use -- and abuse -- in grasping tasks by combining tougher materials for the compliant soft gel, a textured fabric skin, a structurally rigid body, and a calibration process that maintains homogeneous illumination and contrast of the tactile images during use. Finally, we evaluate the sensor's durability along four metrics that track the signal quality during more than 3000 grasping experiments.
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.