PofTac: A Robust Plastic Optical Fiber-Based Tactile Sensor for Shared Human-Robot Tactile Sensing

Abstract

Transferring human manipulation skills to robotic hands requires tactile data beyond kinematic demonstrations, including contact location, force magnitude, and force distribution. Existing wearable tactile interfaces, however, are limited by motion artifacts, electromagnetic interference (EMI), restricted sensing range, and insufficient cross-platform compatibility. This paper presents PofTac, a plastic optical fiber (POF)-based wearable fingertip tactile sensor designed to provide a shared tactile sensing domain across human and robotic fingertips, a key enabler for future human-to-robot skill transfer. PofTac uses light intensity modulation based on Cladding Mode Frustrated Total Internal Reflection (CMFTIR), where contact between a polymer soft pad and pre-bent woven POFs induces force-dependent optical loss. The compact 2 × 3 array is 3 mm thick and weighs 1.98 g, enabling integration on both human and robotic fingertips. With multilayer perceptron (MLP)-based calibration, PofTac achieves sub-millimeter contact localization and a normal-force error of 0.424 N over a 15 N sensing range, while maintaining a 12.2 mN force threshold and robustness against motion artifacts and EMI. Synchronized teleoperation experiments with a human hand and a robotic hand demonstrate consistent tactile responses during rigid and soft object manipulation, supporting PofTac as a shared tactile sensing interface, a foundational step toward future human-to-robot skill transfer.