Source-linked AI summary
FlexiTac: A Low-Cost, Open-Source, Scalable Tactile Sensing Solution for Robotic Systems
Binghao Huang, Yunzhu Li
TL;DR
Robotic end-effectors still lack the rich tactile feedback that supports dexterous manipulation, motivating a sensing platform that is easier to deploy and scale. FlexiTac combines flexible FPC-based piezoresistive pads with compact synchronized readout electronics and demonstrates support for tactile learning pipelines. The platform provides a low-cost, manufacturable, versatile interface for contact-rich manipulation, while its tactile simulator omits shear forces to improve sim–real consistency.
Problem
Robotic tactile sensing remains less capable than biological skin, while researchers need sensing that is easier to deploy, reproduce, and scale across robot platforms.
Method
FlexiTac combines thin FPC-Velostat-FPC piezoresistive pads, compact multi-channel readout electronics, and support for tactile learning and simulation pipelines.
Results
FlexiTac improves manufacturing repeatability and scalability while supporting visuo-tactile learning, cross-embodiment transfer, and real-to-sim-to-real training.
Takeaways & Limitations
FlexiTac provides a practical, low-cost, open-source tactile platform that is easy to fabricate, integrate, and scale across diverse robotic systems.
Takeaways & Limitations
The tactile simulator outputs penetration depth and normal force but omits shear forces to reduce modeling complexity and improve sim–real consistency.
Abstract
from arXiv · showhide
We present FlexiTac, a low-cost, open-source, and scalable piezoresistive tactile sensing solution designed for robotic end-effectors. FlexiTac is a practical "plug-in" module consisting of (i) thin, flexible tactile sensor pads that provide dense tactile signals and (ii) a compact multi-channel readout board that streams synchronized measurements for real-time control and large-scale data collection. FlexiTac pads adopt a sealed three-layer laminate stack (FPC-Velostat-FPC) with electrode patterns directly integrated into flexible printed circuits, substantially improving fabrication throughput and repeatability while maintaining mechanical compliance for deployment on both rigid and soft grippers. The readout electronics use widely available, low-cost components and stream tactile signals to a host computer at 100 Hz via serial communication. Across multiple configurations, including fingertip pads and larger tactile mats, FlexiTac can be mounted on diverse platforms without major mechanical redesign. We further show that FlexiTac supports modern tactile learning pipelines, including 3D visuo-tactile fusion for contact-aware decision making, cross-embodiment skill transfer, and real-to-sim-to-real fine-tuning with GPU-parallel tactile simulation. Our project page is available at https://flexitac.github.io/.
1 Introduction
FlexiTac addresses the gap between human dexterous tactile feedback and less capable robotic sensing with a low-cost, open-source, scalable platform. Its flexible pads, compact readout, and unified interfaces support deployment across robots and tactile learning pipelines.
- Motivation: Partial tactile feedback already improves grasp prediction, grasping and regrasping, contact-rich insertion, and visuo-tactile policy learning despite lacking several biological sensing modalities.The paper motivates tactile sensing as complementary to vision and relevant to dexterous manipulation.
- Platform contribution: FlexiTac combines flexible dense-signal sensor pads, compact synchronized readout electronics, and unified interfaces into a low-cost, open-source platform for robot end-effectors.The system is intended as a practical drop-in module for real-time control and large-scale data collection.
- Manufacturing contribution: Its manufacturing-friendly FPC-based design improves sensor repeatability, scalability, and customization across robotic embodiments.The contribution targets easier deployment, reproduction, and scaling of tactile sensing hardware.
- Learning applications: FlexiTac supports visuo-tactile learning, cross-embodiment skill transfer, and real-to-sim-to-real pipelines with tactile simulation.These capabilities extend the platform beyond sensing hardware to contact-rich manipulation research.
2 Hardware
FlexiTac hardware uses thin, conformable piezoresistive pads with scalable FPC electrodes and lightweight readout electronics. Multiple form factors, modular addressing, rapid fabrication, and approximately $30 per unit support deployment across diverse end-effectors.
- 2.1 Sensor Principle and Pad Manufacturing: FlexiTac offers modular configurations ranging from 32×12 pads to 32×32 mats and compact 8×16 and 16×16 forms for different end-effectors.The pads can be mounted on rigid, compliant, and deformable platforms without significant mechanical redesign.
- 2.1 Sensor Principle and Pad Manufacturing: Each pad uses a sealed FPC-Velostat-FPC laminate with a 12 × 32 matrix, 2 mm pitch, and fabrication time of approximately 5 minutes per pad.The latest design uses patterned copper openings in the FPCs as electrodes, replacing manually aligned conductive threads.
- 2.1 Sensor Principle and Pad Manufacturing: Slots between electrodes increase conformability and concentrate deformation on Velostat, while supporting beams and polyimide stiffeners preserve structural reliability.These features keep the active sensing area flexible while limiting unwanted warping during assembly and use.
- 2.3 Readout Board PCB Design: The readout board uses multiplexers and shift registers to address high-dimensional matrices with minimal wiring, streams synchronized measurements at 100 Hz, and supports lower-resolution pads through channel subsets.An Arduino Nano provides the microcontroller, and standard 0.5 mm FFC connectors provide low-profile interfacing.
- 2.4 Cost Breakdown and Scalability: A complete FlexiTac unit costs approximately $30, while the design supports scalable fabrication and further cost reduction by integrating the microcontroller onto the PCB.At 1000-unit volume, an FPC pair costs $1.36 and the readout PCB costs $2.61, excluding the Arduino Nano.
3 System-Level Integration for Robot Learning
FlexiTac provides a modular tactile interface for robot learning that combines dense contact sensing with synchronized multimodal data streams. The system supports shared 3D visuo-tactile representations, cross-embodiment transfer, and calibrated real-to-sim-to-real training.
- System-Level Integration: FlexiTac integrates thin tactile pads and synchronized readout into a practical interface for vision, proprioception, and control pipelines.The platform is presented as supporting visuo-tactile representation learning, cross-embodiment skill transfer, and simulation-based fine-tuning.
- Visuo-Tactile Representation and Decision Making: Multi-view RGB-D and taxel readings are transformed into a unified 3D point representation, with tactile magnitude appended as a feature for learning-based control.Tactile points receive 3D positions through forward kinematics and calibrated pad geometry, then are merged with visual points and modality indicators.
- Visuo-Tactile Representation and Decision Making: A diffusion-policy module conditions on the fused point set and outputs manipulation actions, supporting closed-loop correction under visual occlusion or precise contact-regulation demands.
- Cross-Embodiment Tactile System: The same FlexiTac sensing stack records synchronized visual, action, and dense fingertip tactile data on a portable human device and an xArm platform for shared-format transfer.The shared interface addresses differences in sensor instances, mounting geometries, and collection sessions while supporting repeated contact and sliding.
- Cross-Embodiment Tactile System: Dense, spatially organized contact measurements support transferred skills such as grasp stabilization, pressure modulation, and fine corrective adjustments.These skills are learned from demonstrations collected on a portable human-operated device and transferred through the shared tactile interface.
- Tactile Simulation with Real-to-Sim-to-Real Pipeline: GPU-parallel tactile simulation models each pad as taxel contact points matching real sensor resolution, enabling simulation-based fine-tuning and sim-to-real transfer.The pipeline omits shear forces, and calibration tunes normal stiffness and damping by matching simulated and real force-response curves.
4 Conclusion
FlexiTac is presented as a practical tactile sensing platform that lowers deployment barriers across robotic systems. Its hardware, electronics, calibration, and simulation support target contact-rich manipulation research.
- 4 Conclusion: FlexiTac combines manufacturable sensor hardware, lightweight readout electronics, calibration, and simulation support in a low-cost, open-source platform.The platform is described as easy to fabricate, integrate, and scale across diverse robotic systems.