Source-linked AI summary
A Durable Vision-Based Tactile Fingertip for Robotic Manipulation
F. Richard Cottrell, Megha H. Tippur, Edward H. Adelson
TL;DR
Commercial vision-based tactile sensors provide rich contact information but remain vulnerable to abrasion and concentrated loading in demanding robotic applications. This work develops a silicone-gel fingertip with a textured TPU protective film, replaceable cartridge, and accelerated durability tests. Under the defined conditions, the sensor lasted substantially longer, degraded gradually, and preserved useful tactile imaging while supporting rapid replacement.
Problem
Vision-based tactile sensors offer rich contact information, but limited durability under abrasion and repeated concentrated loading constrains demanding robotic applications.
Method
The paper develops a rounded fingertip using soft silicone gel, a nonpigmented textured TPU protective film, and a removable cartridge, evaluated with accelerated sanding and repetitive probe tests.
Results
All nine developed sensors remained functionally usable through their respective repetitive-loading test endpoints, while commercial specimens ruptured more than two orders of magnitude earlier under the defined protocols.
Takeaways & Limitations
Gradual damage, preserved tactile imaging, and cartridge replacement support scheduled maintenance for demanding robotic tactile sensing.
Takeaways & Limitations
Commercial comparisons apply to the common laboratory tests rather than a complete assessment of overall sensor durability, and repetitive-loading lifetimes are lower bounds because functional failure was not observed.
Abstract
from arXiv · showhide
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.
1 Introduction
Vision-based tactile sensors provide rich contact information but face durability and maintainability challenges in demanding robotic use. This work addresses those challenges with a durable, replaceable tactile sensing surface evaluated using accelerated comparative tests.
- Motivation: Tactile sensing captures contact location, surface geometry, force distribution, shear, and incipient slip unavailable from external vision alone.These capabilities matter for robotic hands handling objects with unknown geometry, position, compliance, or surface condition.
- Background: Vision-based tactile sensors use internal cameras to observe compliant-surface deformation, enabling high-resolution contact measurement with comparatively simple optical components.GelSight established this approach, followed by compact sensors such as DIGIT and rounded tactile fingertips.
- Design rationale: Nonpigmented, textured surfaces can combine tactile imaging with external visual information while emphasizing nearby objects and reducing distracting distant scenes.The proposed sensor uses total internal reflection and a finely textured outer surface for this purpose.
- Challenge: Durability requires balancing abrasion and loading resistance against compliance and tactile-image quality.Increasing hardness or protective-layer thickness can improve damage resistance but may reduce sensitivity or image quality.
- Maintainability: Gradual degradation and rapid cartridge replacement are treated as complementary requirements because worn sensing surfaces must support scheduled, low-downtime maintenance.Replacement avoids disassembly, electrical disconnection, optical realignment, and recalibration.
- Approach: The proposed fingertip combines a soft silicone gel, thin nonpigmented textured TPU film, removable cartridge, and accelerated abrasion and probe tests.The study also examines damage progression, imaging preservation, and cartridge-replacement practicality.
2 Hardware
The hardware is a modular humanoid-finger platform with interchangeable tactile cartridges and a transparent textured sensing surface. Its imaging design reproduces fine surface detail while the textured interface suppresses distracting external imagery without contact.
- Finger platform: Two rounded interchangeable tactile-finger platforms differ principally in fingertip size, with the smaller human-scale design used for most experiments.The platform supports evaluation of multiple sensing technologies without modifying the vision hardware.
- Hardware components: The finger integrates a miniature color camera, RGB LED boards, rigid housing, and removable sensing cartridge.The cartridge snaps onto the housing and contains the compliant tactile sensor.
- Modularity: The cartridge can be replaced rapidly without disassembling the camera and illumination hardware.The snap-on construction is intended to simplify maintenance after sensor damage.
- Cartridge options: The platform accommodates pigmented and nonpigmented, smooth and textured, and fiducial-marker-based sensor architectures.The present investigation selected the transparent textured cartridge from these compatible options.
- Sensor imaging: Without contact, the textured surface blurs externally illuminated scene images while internal RGB light is piped through the gel and scattered at the texture.This optical behavior reduces distant visual distraction while preserving the sensing configuration.
- Imaging capability: Fine features such as BGA balls, embossed lettering, recycling symbols, threaded fasteners, gauge holes, and textile stitching are reproduced with high spatial fidelity.The compliant silicone gel was selected to preserve sensitive tactile imaging.
3 Materials
Material selection targeted optical transparency, abrasion resistance, elastic recovery, durable bonding, and manufacturability. Platinum-cured silicone and TPU were selected, while a simplified single-chemistry TPU architecture retained durability and thermal conditioning improved sanding lifetime.
- Material requirements: The gel and outer skin were required to combine optical transparency, abrasion resistance, elastic recovery, durable bonding, and reproducible manufacturing.These requirements reflect the simultaneous mechanical and optical demands on a vision-based tactile sensor.
- Gel selection: Platinum-cured silicone was selected over TPE-S because it showed excellent elastic recovery and negligible compression set under identical touch testing.TPE-S exhibited unacceptable permanent deformation despite good optical clarity.
- Protective skin: TPU was selected for its combination of abrasion resistance, elasticity, optical transparency, and chemical durability.Compared with TPE-S systems, TPU also avoided compression set but required coupling chemistry for durable bonding to silicone.
- Architecture simplification: Six sequential coatings of a single TPU grade provided comparable durability to mechanically graded multilayers while substantially simplifying manufacturing.The adopted architecture uses one TPU chemistry, one silicone formulation, and commercially available materials.
- Process optimization: Thermal conditioning of the TPU coating solution increased sanding lifetime by approximately a factor of two to three.The optimized conditioning procedure was incorporated into the final fabrication process.
- Interfacial bonding: Durable service life depended on strong bonding among the cartridge, silicone gel, and TPU skin, achieved using corona activation and silane coupling treatment.The cartridge material was selected to provide structural rigidity while retaining snap-in replacement compliance.
4 Sensor Preparation
The sensor was fabricated by depositing a textured TPU film in a mold, bonding it to soft silicone gel and a polymer cartridge, then curing and demolding the assembled structure. The process directly replicated the tactile texture without additional finishing.
- Sensor construction: The completed sensor comprised an approximately 85 µm polyurethane film bonded to 3 mm platinum-cured silicone gel and a rigid polymer cartridge.The gel had approximately Shore A hardness 6, and the textured mold defined the sensing surface.
- Fabrication sequence: Fabrication proceeded through mold preparation, polyurethane deposition, cartridge preparation, silicone casting, and final demolding.These stages formed the complete sensor assembly.
- Mold preparation: The aluminum mold contained the surface texture replicated onto the finished tactile sensor.Mold release was used to preserve texture fidelity during sensor removal.
- Protective-film deposition: The TPU protective film was produced by applying six sequential wet-on-wet coatings with partial solvent evaporation between applications.The coating used BASF Elastollan 1170A10 polyurethane in a THF/MEK solvent mixture.
- Adhesion preparation: Corona and silane treatments promoted adhesion between the TPU film, silicone gel, and polymer cartridge.The treatments included controlled reaction and thermal-curing conditions before assembly completion.
- Casting and curing: The mold cavity was filled with platinum-cured silicone gel, the prepared cartridge was inserted into the uncured gel, and the assembly was cured at 180°F (82°C) for 3.5 hours.Insertion displaced excess gel while positioning the cartridge within the mold.
- Finishing: Demolding directly transferred the textured mold surface into the polyurethane film without additional processing.This produced the final tactile sensing surface in the completed cartridge assembly.
5 Testing
The study uses two accelerated tests to compare tactile-surface durability under sustained abrasion and severe concentrated mixed-mode loading. The repetitive probe configuration produces confined shear, while the sanding setup provides controlled sliding abrasion.
- Test objectives: Two accelerated tests evaluated sustained abrasive wear and repeated concentrated loading with lateral displacement.The developed sensor was compared with commercial GelSight Mini and DIGIT sensors under common laboratory conditions.
- Repetitive probe test: 39.2 N (4 kgf) loading produced approximately 3.5 mm probe penetration, less than approximately 0.5 mm separation from the cartridge, and approximately 1.1 mm lateral displacement.The probe cycled at 45 cycles per minute with a 0.3 s peak-force hold and 0.5 s unloaded interval.
- Repetitive probe test: A lateral-displacement-to-compliant-thickness ratio exceeding approximately 2 indicated severe confined shear deformation, not direct local material strain.The gel becomes confined between the probe and underlying cartridge at high penetration depths.
- Test interpretation: The RPT was designed as a deliberately severe comparative endurance test rather than a simulation of typical service contact.It accelerates damage associated with high-strain contact, interfacial delamination, and concentrated deformation.
- Sanding test: The rotating-drum sanding test used 400-grit abrasive paper on a 3-inch drum at 80 rpm, with abrasion performance normally evaluated at approximately 1.96 N.The abrasion endpoint was the first visually detectable rupture exposing the underlying silicone gel.
6 Results
The developed sensor substantially outlasted commercial references in both accelerated tests. Its damage progressed locally and gradually, while tactile imaging retained useful detail at the reported endpoints.
- Abrasion resistance: 180 minutes was the mean time to protective-film rupture at 400-grit paper and a 200 g load, compared with approximately 0.5 minutes for DIGIT and 0.4 minutes for Mini.At the standard condition, the developed-sensor range was 176–184 minutes across two specimens.
- Abrasion resistance: Lifetime decreased approximately tenfold as normal force increased from 200 g to 400 g.The developed sensor still substantially outperformed the commercial reference sensors under the defined sanding conditions.
- Damage progression: Localized TPU-film rupture left most of the sensing area intact, so the abrasion endpoint marked damage onset rather than complete functional failure.Tactile images retained useful detail at the defined endpoint and beyond it.
- Repetitive probe testing: Eight days and 520,000 contacts produced a faint circular probe imprint while tactile images changed little during repetitive testing.The test used a 6-mm-diameter probe at 39 N (4 kgf).
- Commercial comparison: Nine developed sensors remained functionally usable for 5–8 days, whereas commercial Mini and DIGIT specimens ruptured after approximately 35 and 25 minutes, respectively.Three developed sensors formed small localized TPU–gel delamination blisters, while imprints generally produced little detectable imaging interference.
- Interpretation: The commercial comparison is limited to common laboratory conditions because the sensors differ in geometry, elastomer construction, protective-surface design, and intended operating conditions.The measurements therefore do not constitute a complete assessment of overall sensor durability.
7 Discussion
The proposed construction improves durability while preserving useful tactile imaging during gradual damage development. A replaceable cartridge supports scheduled maintenance, although the accelerated comparisons do not establish field lifetime or a standardized test.
- Discussion: More than two orders of magnitude improvement was observed under the defined accelerated conditions, while repetitive-loading lifetime remained an incomplete estimate because functional failure was not reached.The developed sensor endured approximately 2–3 hours of sanding and remained usable for 5–8 days in repetitive probe testing.
- Damage behavior: Localized film rupture, probe imprints, and small delamination blisters developed gradually while useful tactile information remained available.This separates physical damage onset from complete loss of tactile function.
- Maintainability: The removable cartridge can be replaced without sensor disassembly, electrical disconnection, optical realignment, recalibration, or specialized tools.Together with gradual degradation, this architecture supports scheduled rather than emergency maintenance.
- Limitations and development use: The accelerated tests helped identify promising fabrication conditions but cannot quantify individual process-variable effects without independent investigation.Additional interlaboratory testing and statistical characterization would be required before standardization.
- Scope: The durability concepts may apply beyond the nonpigmented optical architecture to sensors using pigmented elastomers or embedded surface markers.This broader applicability is stated as a supported scope of the developed concepts.
8 Conclusion
The work demonstrates a vision-based tactile sensor that combines increased laboratory durability with gradual damage and rapid sensing-surface replacement. These properties support maintainable deployment while leaving interfacial delamination resistance, abrasion-test repeatability, and graded protective structures for future improvement.
- 2–3 hours: the developed sensor reached protective-film rupture in the accelerated sanding test.
- All nine developed sensors remained functionally usable when repetitive probe testing was discontinued after 5–8 days.Seven lasted 5 days, one 6 days, and one 8 days; functional failure was not observed, so these results are lower bounds under the defined conditions.
- Damage developed gradually rather than catastrophically, while useful tactile imaging was preserved through the respective test endpoints.Abrasion caused localized TPU-film rupture, whereas repetitive loading produced probe imprints and occasional small delamination blisters.
- The replaceable sensing cartridge can be exchanged without disassembly, electrical disconnection, optical realignment, or recalibration.Removal requires only a simple plastic prying tool, supporting scheduled maintenance and economical replacement.
- The maintainable construction may also apply to other vision-based tactile sensor architectures beyond the demonstrated nonpigmented sensor.
- Future work targets interfacial delamination, abrasion-test repeatability, and graded TPU structures that preserve tactile sensitivity while reducing interfacial stress.
- The study presents maintainable sensor construction as a practical path toward vision-based tactile sensing for demanding robotic applications.Its approach reduces wear, allows gradual damage development, and enables rapid replacement rather than attempting to eliminate wear entirely.