cs.ROAug 25, 2026

A Durable Vision-Based Tactile Fingertip for Robotic Manipulation

Authors: F. Richard Cottrell, Megha H. Tippur, Edward H. Adelson

Organizations: MIT CSAIL

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Currently available commercial vision-based tactile sensors provide rich contact information but remain vulnerable to abrasion and repeated concentrated loading, limiting their use in demanding robotic applications. This work presents a durable tactile fingertip comprising a soft silicone gel with a nonpigmented, textured, thin thermoplastic-polyurethane protective film and a replaceable sensing cartridge. Durability was evaluated using two accelerated laboratory procedures: a rotating-drum sanding test and a repetitive probe test applying 39.2 N (4.0 kgf) at 45 cycles per minute. Under the defined sanding conditions, the developed sensor reached the protective-film rupture endpoint after approximately 2-3 hours. During repetitive probe testing, all nine developed sensors remained functionally usable when testing was discontinued: seven after 5 days, one after 6 days, and one after 8 days. Commercial GelSight Mini and DIGIT specimens exhibited initial surface-film rupture after approximately 24-30 seconds of sanding and 25-35 minutes of repetitive loading. Damage to the developed sensor progressed gradually and produced little interference with tactile imaging at the test endpoints. These observations establish durability improvements of more than two orders of magnitude under the defined accelerated conditions. Combining increased durability, gradual degradation, and rapid cartridge replacement offers a practical approach to maintainable vision-based tactile sensing for demanding robotic applications.

Figures & tables

Figure 1: Humanoid vision-based tactile robot fingers.
Figure 2: The components of the humanoid vision-based tactile finger.
Figure 3: Assembly of the finger housing with the LEDs and the camera, (right side) the sensor is snapped onto the finger housing.
Figure 4: Examples of interchangeable vision-based tactile sensing cartridges compatible with the humanoid finger platform. The present work focuses on the transparent textured sensor (upper left).
Figure 5: No contact image.
Figure 6: Left image - object photograph, right image - finger camera image
Figure 7: Sensor manufacturing components (mold and cartridge) and completed sensor.
Figure 8: Repetitive probe test (RPT) apparatus.
Figure 9: Probe position as a function of applied load for the 6-mm-diameter probe having a contact face angled at 20°.
Figure 10: Probe force as a function of penetration depth into the tactile sensor. A rapid increase in force is observed as penetration approaches approximately 3.5 mm, corresponding to near – complete compression of the 3.5 mm gel thickness. The non-linear increase reflects geometric confinement of the gel against the underlying cartridge.
Figure 11: Sectional representation of probe position at 39.2 N (4.0 kgf) applied force. The remaining separation between the probe surface and the underlying cartridge is less than approximately 0.5 mm. Under this condition lateral probe displacement of approximately 1.1 mm is observed producing a confined shear deformation state.
Figure 12: Lateral displacement of the probe tip as a function of applied normal force. The approximately linear relationship reflects compliance within the load-cell carriage assembly. This lateral motion imposes additional cyclic shear deformation on the sensor during RPT operation.
Figure 13: Rotating-drum abrasion apparatus used to evaluate surface durability of the TPU stack. The 3 inch diameter drum is wrapped with 400 grit abrasive paper and operated at 80 rpm. Normal force is applied via calibrated weights on pivot arm assembly.
Digital-scaleDeveloped sensorDIGITMini
reading (g)400 grit150 grit400 grit400 grit
200180 ( n=2n=2 )31 ( n=2n=2 )0.5 ( n=1n=1 )0.4 ( n=1n=1 )
30064 ( n=1n=1 )18 ( n=2n=2 )——
40018 ( n=2n=2 )4 ( n=2n=2 )——
Table 1: Sanding-test time to first protective-film rupture, in minutes.
Figure 14: Sanding results.
Figure 15: The results of a sanding test. Surface photographs and tactile sensor photographs vs time during sanding test. (The specimen shown is representative of the progression of wear and was not one of the two specimens used to calculate the 180-minute value in Table 1.)
Figure 16: Terminal sanding images - Top: Surface photos with time of sanding, Below: tactile camera images of textured textile sample.
Figure 17: GelSight Mini and DIGIT Sanding Test - photographed through the sensor during testing (200 gram force, 400 grit sandpaper, 80 rpm on sanding drum, 3 inch diameter drum)
Figure 18: Repetitive Probe Test Results.
Figure 19: Terminal test images of repetitive probe test.
Figure 20: DIGIT and Mini RPT Results. Top surface photographs

Explore similar work

CardsList

  1. GlowTact: Simple and Compact Vision-Based Tactile Sensing with High Sensitivity and Spatial Resolution

    Sep 26, 2026Yuxiang Ma, Megha Tippur, Pengfei Ye +4Vision-Based Tactile SensingTactile Sensing

    Massachusetts Institute of Technology, Cambridge, MA, USA · Harvard University, Cambridge, MA, USA

  2. Vision-Based Tactile Intelligence for Robotics: Sensing, Learning, and Embodied Manipulation

    Aug 16, 2026Peng Zhou, Jun Hu, Sihan Chen +12Vision-Based Tactile SensingTactile Sensing for Robotic Manipulation

    Great Bay University · The University of Hong Kong · Tsinghua University +5

  3. FingerEye: Learning Dexterous Manipulation with Continuous Vision-Tactile Sensing

    Apr 22, 2026Zhixuan Xu, Yichen Li, Xuanye Wu +2Dexterous ManipulationRobotic Tactile Sensing

    National University of Singapore · RoboScience · Huazhong University of Science and Technology +1