GelSphere: An Omnidirectional Rolling Vision-Based Tactile Sensor for Online 3D Reconstruction and Normal Force Estimation

*Equal contribution
IROS 2026
A spherical, rolling tactile sensor for continuous 3D surface capture and online normal force estimation.
GelSphere teaser overview

Abstract

We present GelSphere, a spherical vision-based tactile sensor designed for real-time continuous surface scanning. Unlike traditional vision-based tactile sensors that can only sense locally and are damaged when slid across surfaces, and cylindrical tactile sensors that can only roll along a fixed direction, our design enables omnidirectional rolling on surfaces. We accomplish this through our novel sensing system design, which has steel balls inside the sensor, forming a bearing layer between the gel and the rigid housing that allows rolling motion in all axes. The sensor streams tactile images through Wi-Fi, with online large-surface reconstruction capabilities. We present quantitative results for both reconstruction accuracy and image fusion performance. The results show that our sensor maintains geometric fidelity and high reconstruction accuracy even under multi-directional rolling, enabling uninterrupted surface scanning.

Highlights

  • Omnidirectional rolling contact instead of single-axis motion.
  • Wireless tactile image streaming for online inference.
  • Continuous large-surface reconstruction from tactile observations.
  • Normal force estimation from the same sensing pipeline.
1.07-1.25 mm
Planar drift MAE
1.3-2.7 deg
Angular drift
Wi-Fi
Streaming interface
3D + Force
Joint sensing output
Fabricated GelSphere sensor

The fabricated GelSphere prototype integrates the rolling gel shell, internal imaging stack, and compact wireless hardware into a self-contained tactile sensing platform.

Sensor Design

The sensing skin rolls over a bearing layer while the optical module remains magnetically stabilized, preserving a consistent internal view during omnidirectional motion.

GelSphere sensor design

System Results

Single and Hex Reconstruction

Single-contact observations remain geometrically consistent with the reference hex target, showing that GelSphere preserves local shape accuracy while rolling across the surface.

Single frame and hex geometry comparison

Large-Surface Reconstruction

Rolling contact images are fused into larger tactile maps in real time, enabling continuous surface reconstruction over a 75 x 32.5 mm2 scan region.

Large surface reconstruction
Online fusion maintains geometric consistency during multi-directional rolling.

Force Estimation

The same tactile observations also support normal force regression, so the sensor measures shape and contact load simultaneously during rolling interaction.

Force estimation results

Live Reconstruction

Project Video

BibTeX

@inproceedings{gelsphere2026,
  title     = {GelSphere: An Omnidirectional Rolling Vision-Based Tactile Sensor for Online 3D Reconstruction and Normal Force Estimation},
  booktitle = {2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  year      = {2026},
}