Source-linked AI summary
A Wearable Pneumatic Device for Continuous, Closed-Loop, Bidirectional Tactile Interaction
Cosima du Pasquier, Aliyah Smith, Serin Huber, Joshua Phelps, Ilana A. Cohen, Ava Chen, Monroe Kennedy, Allison M. Okamura
TL;DR
Wearable tactile sensing and haptic display are typically engineered separately, limiting bidirectional touch interaction in teleoperation. This paper unifies both functions in a closed-loop pneumatic device, reducing applied pressures by up to 23.1%, task duration by up to 27.4%, and subjective mental workload by 18.8%.
Problem
Wearable tactile sensing and haptic display are typically engineered separately, limiting matched bidirectional tactile interaction for teleoperation.
Method
The system uses distributed, wearable pneumatic channels in which local closed-loop control enables the same textile pouch mechanism to sense and display pressure.
Results
Applied pressures fell by up to 23.1%, task duration by up to 27.4%, and subjective mental workload by 18.8% during teleoperated manipulation.
Takeaways & Limitations
The device provides a practical foundation for bidirectional tactile communication and future tactile interaction and teleoperation research.
Takeaways & Limitations
Technical interruptions and participant exclusions in the multi-subsystem experimental setup indicate a need for tighter system integration.
Abstract
from arXiv · showhide
We present a system of two wearable pneumatic haptic devices that supports continuous, closed-loop, bidirectional tactile interaction at perceptually relevant force and temporal scales. A single device can contain up to twelve pressure sensing channels connected to textile-based pneumatic pouches. Each channel in a device can be used as a sensor, an actuator, or both. As an actuator with integrated sensing, the channel generates stable skin indentation through local closed-loop control. As a sensor, a channel can be mounted (or worn) on any surface, including on a robot gripper or on the human body, and used to measure touch interactions with the environment or a human user. A distributed architecture supports sustained pressure output, rapid dynamic response, and wireless pairing of identical devices in a system to transmit and reproduce tactile pressure signals in real time. Device-level characterization demonstrates force bandwidth exceeding 30 Hz, rapid and well-damped step responses, and extended pressure retention compared to prior compact pneumatic platforms. Human studies show that pressure-based fingertip feedback enables discrimination of force and stiffness, improves teleoperated manipulation by reducing applied pressures by up to 23.1% and task duration by up to 27.4%, and lowers subjective mental workload by 18.8%, particularly under visually constrained conditions. By unifying tactile sensing and haptic feedback within a single pneumatic modality, the device provides a practical foundation for bidirectional touch interaction in teleoperation.
SUMMARY · KEYWORDS
The paper presents wearable pneumatic haptic devices for continuous, closed-loop, bidirectional tactile interaction, combining multi-channel sensing and actuation. Device characterization and human studies demonstrate responsive operation and benefits for force perception and teleoperated manipulation, especially under visually constrained conditions.
- SUMMARY: The system uses two wearable pneumatic haptic devices to support continuous, closed-loop, bidirectional tactile interaction at perceptually relevant force and temporal scales.
- SUMMARY: Each device can contain up to twelve pressure-sensing channels connected to textile-based pneumatic pouches.
- SUMMARY: Each channel can function as a sensor, an actuator, or both, with integrated sensing enabling stable skin indentation through local closed-loop control.
- SUMMARY: Force bandwidth exceeds 30 Hz, while step responses are rapid and well damped and pressure retention is extended relative to prior compact pneumatic platforms.
- SUMMARY: Pressure-based fingertip feedback enables force and stiffness discrimination and improves teleoperated manipulation under visually constrained conditions.
- SUMMARY: Applied pressures decrease by up to 23.1%, task duration by up to 27.4%, and subjective mental workload by 18.8%.
- KEYWORDS: The paper is characterized by the keywords Soft Robotics, Haptics, and Teleoperation.
INTRODUCTION
Teleoperation needs continuous, interpretable bidirectional tactile feedback, but haptic display and robot tactile sensing are typically engineered separately. Pneumatic textile pouches offer a shared modality because internal pressure both actuates and senses normal force, enabling compact closed-loop, wireless interaction.
- Haptic feedback can lower task completion time1,2, improve manipulation accuracy3, and decrease peak and average interaction forces4.
- Continuous, interpretable bidirectional tactile feedback is increasingly essential as teleoperation supports control and demonstration collection for imitation learning.
- Conventional teleoperation separates human haptic displays from robot tactile sensors, while bilateral force-control loops can become unstable under hard contacts, stiff settings, and communication latency5,6.
- Pneumatic textile pouches can both display and measure normal force because internal pressure is simultaneously the actuation variable and sensed signal.Inflated contact also reproduces the fingerpad’s natural growth of contact area with force against a compliant object.
- The proposed compact closed-loop controller locally regulates each pouch, removes compressor and valve burdens, and wirelessly transmits sensed pressure with latency below the threshold of immediate tactile perception.
RESULTS
The wearable pneumatic system achieved rapid, sustained, bidirectional tactile interaction, with force and stiffness perception approaching direct touch but degraded weight discrimination in harder comparisons. Haptic feedback reduced applied pressure, improved selected manipulation outcomes, and lowered workload, especially under visual occlusion.
- System architecture: The same textile pouch operated as a sensor or actuator through local closed-loop pressure regulation with integrated valves, sensors, and microcontrollers.The untethered architecture supports distributed sensing, actuation, and wireless pairing of identical devices for real-time tactile transmission.
- Device performance: The device sustained pressure for 30 min, settled in 39 ms, provided a 34.2 Hz -3 dB force bandwidth, and mirrored pressure between paired devices in 64 ms.It retained pressure above 1700 mbar after 16 min, whereas comparable compact pneumatic systems lost pressure within seconds.
- Perceptual evaluation: Participants discriminated force at 92.2% accuracy for two elements and 92.1% for three elements, while stiffness feedback matched direct touch for two elements and remained within six points for three.Device-mediated stiffness accuracy was 91.4% versus 90.1% by hand for two elements, and 90.8% versus 96.7% for three.
- Perceptual evaluation: Weight discrimination fell from 95.7% to 87.5% in two-element comparisons and from 100% to 53.8% in three-element comparisons with teleoperation.The results indicate that mediated feedback preserved simpler weight ordering but lost substantial performance when a third weight increased cognitive and motor complexity.
Discussion
The discussion presents a wearable pneumatic platform that unifies closed-loop tactile sensing and haptic display, enabling stable tactile cues, bidirectional communication, and improved teleoperation. Results support reliable force and stiffness discrimination with reduced cognitive demand, while highlighting weight-discrimination challenges and system-integration limitations.
- The shared physical mechanism unifies tactile sensing and haptic display in a body-worn, closed-loop pneumatic design.This addresses the tradeoff between sustained pressure systems confined by external hardware and untethered open-loop systems limited to transient pulses.
- Closed-loop pressure regulation and rapid, well-damped response enable stable, continuous fingertip indentation at perceptually relevant force and temporal scales.These properties support prolonged contact and fine force regulation, especially in teleoperation where force cues must remain stable over time.
- Stiffness discrimination approached direct manual performance, while force and stiffness tasks overall showed high accuracy with haptic feedback.Indentation-based cues can convey continuous contact information, including force magnitude and material properties.
- Weight discrimination declined in harder teleoperated 3-element comparisons because participants had to integrate fingertip pressure cues with simultaneous robot control.The result highlights how task-level cognitive demands shape perceptual performance.
- Pairing two identical devices enables real-time bidirectional tactile communication and pressure-signal reproduction without intermediate abstraction.The twin-system architecture also supports recording tactile interaction data with the same sensing modality under haptic and non-haptic conditions.
- Teleoperated manipulation showed reduced applied force, improved efficiency, and lower mental demand and frustration, particularly under visually constrained conditions.The findings indicate that tactile cues reduce the cognitive effort needed to infer contact state and regulate interaction forces.
- Technical interruptions, participant exclusions, and confusion between robotic and haptic latency reveal the need for tighter subsystem integration in future studies.The setup combined separate robot-control, visual-feedback, and haptic-feedback subsystems, creating practical limitations.
METHODS
The system uses distributed, locally closed-loop pneumatic channels with wireless coordination to provide bidirectional fingertip feedback during hand-tracked teleoperation. Its evaluation compares haptic and non-haptic operation through perception and teleoperation tasks involving force, stiffness, and weight discrimination.
- System architecture: Each channel combines a proportional piezoelectric valve, pressure sensor, and dedicated microcontroller for independent closed-loop regulation of textile pneumatic actuators.A system-level microcontroller coordinates channels and wireless communication, while a single CO2 cartridge or wall-air source supports the system.
- Closed-loop control: Local proportional-integral control continuously samples pressure and adjusts valve opening, enabling low-latency setpoint tracking with minimal overshoot across scalable channels.The architecture also minimizes pneumatic dead volume and leakage to support sustained pressure output during prolonged interaction.
- Teleoperation feedback: During teleoperation, gripper normal forces were linearly mapped with an empirically determined scaling factor of four to continuously updated fingertip pressure setpoints.Thumb and index actuators produced normal indentation, while wireless paired devices transmitted force measurements and commands bidirectionally.
- Experimental protocol: Participants controlled a Kinova Gen3 7 DoF arm through unencumbered HoloLens 2 hand tracking, with wrist motion mapped to Cartesian velocity and thumb-index distance controlling gripper opening.The study included 25 able-bodied participants and two phases: Perception and Teleoperation.
- Experimental protocol: The perception phase comprised five tasks spanning direct and robot-mediated discrimination of fingertip force, stiffness, and weight under restricted visual and whole-body motion cues.Participants completed a five-minute training session using an empty aluminum soda can before formal evaluation.
DECLARATION OF INTERESTS
The authors disclose financial, commercial, advisory, contractual, and patent interests related to Haptica, Inc., while the remaining authors report no competing interests. They also disclose using ChatGPT and Claude to improve writing clarity and readability, with subsequent author review and responsibility for the publication.
- Declaration of interests: Several authors report ties to Haptica, Inc., including company leadership, shareholding, scientific advising, equity ownership, and later paid engineering services.C.d.P. is a co-founder, officer, and shareholder; A.M.O. is a scientific advisor with equity; and J.P. later provided paid contract engineering services.
- Declaration of interests: C.d.P. and S.H. hold a provisional patent on technology related to this work, while the remaining authors declare no competing interests.The patent is identified as submission #S25-344-PROV.
- Declaration of interests: The authors used ChatGPT and Claude to enhance manuscript clarity and readability, then reviewed and edited the content and retained full responsibility for the publication.The disclosure names OpenAI’s ChatGPT and Anthropic’s Claude as the writing-support tools.
SUPPLEMENTAL INFORMATION INDEX
The supplemental information contains additional methods, Figures S1–S9, and Tables S1–S7.
- Supplemental methods, Figures S1–S9, and Tables S1–S7 are provided in Document S1.
Supplemental Information A Wearable Pneumatic Device for Continuous, Closed-Loop, · Bidirectional Tactile Interaction
The supplied passage lists contributors to the “Bidirectional Tactile Interaction” section.
- Bidirectional Tactile Interaction: The listed contributors are C. du Pasquier, A. Smith, S. Huber, J. Phelps, I. Cohen, and A. Chen.
- Bidirectional Tactile Interaction: Additional listed contributors are M. Kennedy III and A. M. Okamura.
1 Controller Design and Architecture
The device uses up to twelve independently controlled pneumatic channels with local pressure sensing and distributed closed-loop control, enabling scalable bidirectional tactile interaction. Characterization shows high force bandwidth, rapid settling, and substantially longer pressure retention than a comparable open-loop platform.
- System architecture: Each device contains up to twelve identical channels, each combining a proportional valve, pressure sensor, and channel microcontroller for local closed-loop control.Two identical devices can pair wirelessly so one senses and transmits pressure signals while the other reproduces them.
- Controller architecture: Distributed per-channel control avoids centralized bottlenecks and supports scaling to multiple channels.The collocated pressure sensor provides feedback for each channel’s local controller.
- Actuator characterization: 34.2 Hz average force bandwidth and 76.2 N maximum force demonstrate dynamic and high-strength actuator performance, exceeding Airport-controlled actuators’ 14.5 Hz bandwidth.At 2650 mbar, maximum force was 76.2 ± 2.8 N; force bandwidth was 34.2 ± 1.2 Hz versus 14.5 Hz for Airport.
- Pneumatic performance: Pressure retention substantially exceeds the FlowIO open-loop comparison, with pressure remaining at 2300 mbar after 5 minutes and full depressurization occurring only after 30 minutes.The comparable FlowIO system lost pressure in under 10 seconds under the same procedure.
- Controller performance: 29 ms average rise time and 39 ms total settling time provide rapid, stable pressure control suitable for wearable haptic feedback.Rise times ranged from 5 to 34 ms, and response times below 50 ms are described as sufficient for clear, timely tactile feedback.
2 Review of Pneumatic Controllers for Wearables and Soft Robotics
Pneumatic actuation is widely used in wearable robots, soft manipulators, and haptic interfaces, but system performance depends strongly on controller design. Existing platforms increasingly use closed-loop pressure control, yet typically lack sustained, bidirectional tactile signal exchange in untethered wearable form factors.
- Controller Design: Pneumatic actuation dominates wearable robots, soft manipulators, and haptic interfaces because of its compliance and high power density.Controller size, control strategy, sensing capabilities, and dynamic response fundamentally shape system performance.
- Existing Platforms: Most reviewed platforms use external pneumatic supplies or compact micropumps, with closed-loop pressure control increasingly adopted for stability and repeatability.These systems generally emphasize unidirectional actuation rather than two-way tactile communication.
- Existing Platforms: Existing systems typically do not support sustained, bidirectional tactile signal exchange between remote locations, especially in untethered and wearable form factors.This limitation distinguishes the targeted device from the prevailing pneumatic platforms reviewed in this section.
3 Teleoperation Control Implementation
The teleoperation system used HoloLens 2 hand tracking to control a Kinova Gen3 robot through wrist-motion velocity mapping and calibrated finger-based gripper commands. Top-down gripper orientation and update conditions constrained manipulation for consistent contact and stable control.
- Tracking: HoloLens 2 provided six-degree-of-freedom wrist and fingertip tracking using onboard depth, infrared, and inertial sensing without external cameras or additional calibration.Its see-through display preserved direct views of the physical environment, while the head-mounted form factor supported free movement with consistent tracking.
- Robot control: Wrist position and quaternion orientation changes were converted into linear and angular velocities in the HoloLens world frame to control the robot end-effector.Commands were sent as Cartesian velocities through Kinova’s ROS driver, with joint velocities computed by Kinova’s inverse kinematics solver.
- Constraints: Gripper orientation was restricted to a top-down grasp, permitting only angular velocity about the gripper’s Z axis to reduce task complexity and standardize contact geometry.The coordinate frames were only approximately aligned because pose-delta control did not require precise collocation.
- Gripper control: The gripper opening was controlled by the normalized thumb–index distance after session-specific calibration to each participant’s comfortable maximum span.The normalized command mapped gcmd = 0 to fully open and gcmd = 1 to fully closed.
- Command filtering: Gripper commands updated only for changes above 0.05, intervals of at least 0.01 s, and end-effector speeds below 0.05 m/s, filtering noise and motion-related false activations.The speed condition restricted finger commands to periods when the arm was near rest.
4 Detailed User Study Procedures and Additional Results
The supplementary study used randomized perception and teleoperation procedures with 25 participants to evaluate haptic feedback under varied visual conditions. Per-participant analyses showed more consistent force regulation and faster transport with haptic feedback, while usability gains were positive but nonsignificant.
- Outcome Measures: Perception accuracy was computed from confusion matrices, while teleoperation outcomes included completion time, path length, applied pressure, usability, and six NASA-TLX workload subscales.Completion time spanned initial object contact to release, and trials exceeding 5 minutes were classified as failures.
- Per-participant Teleoperation Results: Haptic feedback consistently maintained lower or comparable applied pressure across teleoperation tasks, independently of condition-order effects.Per-participant results also showed that condition order influenced task duration and path length in PnP1 and PnP2.
- Pick vs. Place Time Allocation: Haptic feedback shifted time allocation toward grasping because participants transported objects faster, spending proportionally less time in the Place phase.This timing pattern provides a supplementary explanation for the task-efficiency improvements reported in the main text.
- SUS Results: Participants rated the haptic-feedback system as more usable overall, but the usability difference did not reach statistical significance.Statistical comparisons used paired t-tests after Shapiro–Wilk assessment, with missing logging-error trials excluded symmetrically.