TransTac: Visuo-Tactile Modality Transition Via Ultraviolet-Encoded Transparent Elastomers
Lingyue Yang, Bin Fang
AI summary
Problem
Existing vision-based tactile sensors are opaque and block visual observation, while RGB-D cameras degrade rapidly at close range, leaving a critical geometric sensing gap during the transition from approach to physical contact.
Approach
The system uses a transparent silicone elastomer embedded with UV-reflective markers and a lightweight detector to track deformation, combined with a prior-guided Delaunay stereo matching algorithm that fuses sparse tactile triangulation with dense RGB-D depth estimation.
Key results
- 83.3% zero-shot recognition accuracy on tactile images
- 21% improvement in stereo correspondence robustness over baselines
- Cross-modal alignment similarity increases from 0.2 to 0.77
- ~2.44 mm mean geometric error maintained in near-contact zones
Why it matters
Enables robots to seamlessly transition from visual approach to tactile contact in unstructured environments, improving grasping and adaptive manipulation.
Abstract
Vision-based tactile sensors (VBTS) recover high- resolution contact geometry but typically rely on opaque elas- tomer layers that prevent visual transparency, while RGB-D cameras provide global depth perception yet degrade signifi- cantly at close range. To address this limitation, we present TransTac, a transparent ultraviolet (UV)-encoded binocular VBTS that integrates visual observation and marker-based tactile reconstruction within a single compact device. The system employs a transparent elastomer embedded with UV- reflective markers and a prior-guided Delaunay stereo matching algorithm for robust sparse triangulation. To reliably detect densely distributed semitransparent mark- ers, we develop a lightweight detector that enables stable lo- calization under contact and deformation. The proposed prior- guided Delaunay matching improves correspondence robust- ness by approximately 21% compared with global assignment baselines while maintaining high reconstruction accuracy. In semantic evaluation, TransTac achieves up to 83.3% zero- shot recognition accuracy on tactile images, exceeding opaque tactile baselines by approximately 50 percentage points. Embed- ding analysis further reveals substantially stronger cross-modal alignment with natural images, with class-center similarity increasing from around 0.2 to over 0.77. Controlled near- distance experiments quantify the degradation of RGB-D depth reliability and demonstrate extended geometric coverage en- abled by visuo-tactile integration. Finally, a compact prototype is implemented with an approximate hardware cost of $70. Code and hardware design are publicly available at https://github.com/87361/TransTac.