cs.ROApr 1, 2026

A soft and lightweight fabric-based pneumatic interface for multimodal fingertip tactile feedback

Authors: Rui Chen, Daniele Leonardis, Antonio Frisoli

Organizations: Institute of Mechanical Intelligence, School of Advanced Studies Sant’Anna (SSSA), 56127 Pisa, Italy.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Wearable fingertip haptic devices are critical for realistic interaction in virtual reality, augmented reality, and teleoperation, yet existing approaches struggle to simultaneously achieve adequate tactile output, low mass, simple fabrication, and untethered portability. Here we show that fabric-based pneumatic actuation can address this gap. Our device comprises four pneumatic chambers fabricated from thermoplastic polyurethane-coated fabric via computer numerical control heat-sealing, yielding a soft, conformable interface weighing 2.1 g that operates untethered with a wrist-mounted control unit. Mechanical and dynamic characterization confirms that the fabric actuators produce sufficient force, displacement, and bandwidth for fingertip tactile rendering. A psychophysical study with 15 participants demonstrates classification accuracy exceeding 90% across three distinct tactile modes -- contact configuration, directional sliding, and vibrotactile frequency. These findings establish fabric-based pneumatic actuation as a viable technology route for lightweight, low-cost, and multimodal fingertip haptic interfaces.

Figures & tables

Figure 1: Overview of the fabric-based pneumatic fingertip haptic interface. (A) System architecture showing the wrist-mounted control box, pneumatic tubing, and fingertip haptic interface. (B) The 2 ×\times 2 chamber configuration enabling independently addressable spatiotemporal actuation patterns. (C) Fabrication process via CNC heat-sealing and assembly of the soft wearable interface. (D) Performance comparison against prior fingertip haptic devices in normalised force output and displacement. (E) Three stimulation modes supported by the device: contact configuration, directional sliding, and vibrotactile feedback.
Figure 2: Mechanical and dynamic characterisation of the fabric actuator. (A) Geometric parameters of the haptic actuator. (B) Modelled and experimental force–displacement results across actuator widths WW . (C) Pressure–hysteresis curves of the single-chamber actuator. (D) Force output as a function of chamber number. (E) Force output of the four-chamber actuator across three pressure levels. (F) Bode plot demonstrating a −- 3 dB bandwidth of 7.1 Hz. (G) Step response showing 64 ms rise time and 11 ms fall time. (H) Force and pressure output stability over 1,000 actuation cycles.
Figure 3: Portable haptic device and spatial pressure distribution. (A) Haptic device worn on the index finger. (B) Exploded view of the control box showing internal components. (C) The soft haptic interface weighs only 2.1 g. (D) Pump pressure output as a function of PWM duty cycle. (E) Experimental setup for pressure distribution measurement. (F) Pressure distribution under varying actuation pressure levels. (G) Pressure distribution across different chamber actuation patterns.
Figure 4: Human perception study: contact configuration discrimination task. (A) VR environment used for touch simulation. (B) Distinct actuation patterns generated under different contact configurations. (C) Experimental setup for the user study. (D) Nine actuation patterns corresponding to different contact configurations. (E) Confusion matrix illustrating an overall classification accuracy exceeding 94%. (F, G) Accuracy and response time distributions across different stimulus patterns.
Figure 5: Human perception study: sliding and vibrotactile discrimination tasks. (A) VR environment for the sliding task. (B) Schematic of six directional sliding patterns. (C) Confusion matrix for the sliding task, indicating an overall classification accuracy above 96%. (D, E) Accuracy and response time distributions across sliding stimuli. (F) VR environment for vibrotactile-based texture representation. (G) Schematic of three vibrotactile frequency conditions. (H) Confusion matrix for the vibrotactile task, indicating an overall classification accuracy above 98%. (I, J) Accuracy and response time distributions across frequency conditions.
More in this topic

Pneumatic Actuation

See all 29 papers

Explore similar work

CardsList

  1. A Wearable Pneumatic Device for Continuous, Closed-Loop, Bidirectional Tactile Interaction

    Sep 1, 2026Cosima du Pasquier, Aliyah Smith, Serin Huber +5Tactile SensingRobot Teleoperation

    Department of Mechanical Engineering, Stanford University, Stanford, CA, USA

  2. A wearable haptic device for edge and surface simulation

    Apr 1, 2026Rui Chen, Xianlong Mai, Alireza Sanaei +3Haptic FeedbackVirtual Reality

    Institute of Mechanical Intelligence, School of Advanced Studies Sant’Anna (SSSA), 56127 Pisa, Italy. · CAS Key Laboratory of Mechanical Behavior and Design of Materials, Institute of Humanoid Robots, School of Engineering Sciences, University of Science and Technology of China, Hefei, China.

  3. A Reconfigurable Fabric Based Pneumatic Actuator with Button Fastened Constraint Modules for Multi Mode Actuation

    Oct 8, 2026Kentaro Kimura, Hiroki Ishizuka, Yusuke Sakaue +2Soft Pneumatic ActuatorsSoft Robotics

    Graduate School of Engineering Science, Osaka University, Osaka 560-0043, Japan