Hapt: Soft Thermal–Haptic VR Interface

• Hapt is a soft, fabric-based thermal–haptic interface that integrates pneumatic actuation with conductive heating to deliver co-located temperature and pressure cues.
• It features ultra-light wearable fingertip modules (2 g per unit) with textile pneumatic chambers and embedded resistive heaters, optimizing user comfort and dexterity.
• Experimental studies demonstrate improved manipulation accuracy and reduced fingertip indentation, positioning Hapt as a promising alternative to rigid haptic devices.

Hapt is a soft, fabric-based thermal–haptic interface for virtual reality and teleoperation that integrates pneumatic actuation and conductive fabric in a fully soft, wearable fingertip module. It delivers co-located normal-pressure and thermal cues to the fingerpad by embedding heating elements within textile pneumatic chambers, and it is organized as two modules for the index finger and thumb on a wrist-strap platform. The device is explicitly positioned as an ultra-light alternative to conventional motorized or rigid thermal interfaces, with a reported mass of 2 g per finger unit, rapid thermal modulation, and experimentally validated effects on thermal identification and VR manipulation performance (Chen et al., 28 Aug 2025).

1. Device concept and physical architecture

Hapt is built around a fabric-based fingertip module, denoted FTHA, that combines a textile pneumatic pouch actuator with a fabric-based resistive heater. The pneumatic chamber uses dual-layer TPU-coated woven fabric (40D), heat-sealed to form a pouch motor. Typical planform dimensions are tailored to finger anthropometrics: approximately 18 mm × 16 mm for the index finger and 20 mm × 18 mm for the thumb. The heating element is a laser-cut woven conductive fabric with a 15 mm × 15 mm active area and 3 mm trace widths, bonded onto the inner face of the TPU-coated fabric. A 50 kΩ @ 25°C NTC thermistor is placed centrally on the same layer to sense the heated area (Chen et al., 28 Aug 2025).

Integration proceeds by bonding the heater and thermistor to one TPU-coated fabric sheet, then heat-sealing a second TPU-coated sheet around the perimeter to form the airtight chamber. A TPU air tube is bonded to the pouch for pneumatic routing, and an elastic fabric interface stitches the module to the finger attachment. Two such modules are mounted on an ergonomic wrist-strap platform that routes air lines and electrical leads. The wrist-strap hub houses an ESP-32 control board, an H-bridge current driver for the heaters, and two proportional solenoid valves (SMC ITV0010) for pressure regulation (Chen et al., 28 Aug 2025).

A central design feature is mass and wearability. The FTHA weighs 2 g per finger unit, whereas conventional motorized fingertip devices are described as 10–50 g. The reported design objective is improved comfort and dexterity over long sessions. This weight reduction is not merely ergonomic; it is tied to the broader attempt to preserve natural finger motion while still providing normal-force and thermal feedback from a single compliant interface (Chen et al., 28 Aug 2025).

2. Actuation principles, control, and governing models

Hapt combines pneumatic normal-force actuation with conductive thermal actuation. In the pneumatic subsystem, compressed air is fed through proportional solenoid valves to generate controllable chamber pressure. In VR tasks, valve commands are proportional to fingertip proxy indentation, and pressure is capped at 20 kPa for interaction forces, although full characterization was performed up to 50 kPa. The ESP-32 outputs a setpoint to the proportional regulators, and pressure is reported as stable and repeatable across trials (Chen et al., 28 Aug 2025).

The thermal subsystem uses Joule heating in the conductive fabric. The heater is driven by a current-controlled H-bridge, while temperature is measured by the co-located NTC and regulated via a PID loop in the 25–50°C range. Heating is active, but cooling is passive through natural convection and conduction to the finger and ambient. The device therefore uses geometry and clearance strategies to accelerate cooling when pressure is released. Pneumatic and thermal outputs are scheduled independently by the microcontroller but are physically co-located, permitting congruent cues such as squeeze plus warmth upon contact (Chen et al., 28 Aug 2025).

The reported first-order models are standard but operationally important. Normal force is approximated as

$F = P A,$

where $P$ is chamber pressure and $A$ is the effective contact area. A lumped membrane model is also given:

$F_n \approx P A - k_m x,$

with $k_m$ as effective membrane stiffness and $x$ as normal displacement, while chamber compliance is

$C = \frac{dV}{dP}.$

For the thermal side, electrical power is

$P_\text{elec} = I^2 R,$

and the lumped transient response is

$T(t) = T_\infty + (T_0 - T_\infty)e^{-t/\tau}.$

Cooling from the fabric surface follows

$q = h A (T_s - T_\infty).$

These expressions are used to explain the near-linear force–pressure trend and the clearance-dependent cooling behavior (Chen et al., 28 Aug 2025).

A distinctive control variable is fingerpad–actuator clearance $P$0. Larger $P$1 reduces conductive coupling to the finger and allows ambient airflow, increasing effective $P$2 and accelerating cooling when deflated. This produces a direct cooling–force trade-off: larger clearance shortens cooldown to baseline but modestly reduces peak force. The design implication reported in the source is that $P$3 mm is an effective compromise (Chen et al., 28 Aug 2025).

3. Measured performance and optimization

Thermal and force characterization define the quantitative profile of Hapt. The device achieved a peak heating rate of up to 3°C/s in the unloaded initial phase. The average heating rate to a 40°C setpoint was approximately 1.79°C/s unloaded and approximately 0.79°C/s in-contact over 34→40°C. On the pneumatic side, the maximum normal force was 8.93 N at 50 kPa with zero clearance, $P$4 mm (Chen et al., 28 Aug 2025).

Clearance optimization was carried out with a calibrated fixture using anatomically curved plates and dual precision screws. Initial contact at 0.05 N defined $P$5 mm, and additional clearances of 1, 2, and 3 mm were introduced. Force was measured with a high-precision load cell while pressure ramped from 0 to 50 kPa. Near-linear $P$6–$P$7 behavior was observed across all clearances. At 50 kPa, the quantified force trade-off was 8.93 N at $P$8 mm, approximately 8.5 N at $P$9 mm, approximately 7.7 N at $A$0 mm, and approximately 6.6 N at $A$1 mm, corresponding to reductions of 5%, 14%, and 26% relative to baseline (Chen et al., 28 Aug 2025).

The same optimization was motivated by thermal behavior. Thermal characterization showed faster cooldown when the module was not in contact, and deflation-induced separation increased convective cooling while reducing skin conduction. The preferred operating point, $A$2 mm, retained approximately 86% of peak force while materially enhancing passive cooling responsiveness. This suggests that the device was designed not around absolute peak force alone, but around a coupled multiphysics criterion that balances return-to-baseline time with usable interaction force (Chen et al., 28 Aug 2025).

Safety and comfort constraints are also part of performance characterization. Recommended skin-contact operating ranges were reported as prolonged contact generally at or below 43–45°C, with brief cues up to 50°C under careful monitoring. The device enforces a 50°C ceiling, uses typical rendering within 25–44°C, and provides closed-loop temperature control with NTC feedback at ±1°C precision. The heater drive is low voltage, conductive traces are insulated, and strain relief and wiring are described as robust (Chen et al., 28 Aug 2025).

4. Perceptual validation and task-level effects

Two user studies were reported. The participant pool comprised 11 healthy subjects, 7 male and 4 female, aged 28.8 ± 3.1 years, all right-handed, under SSSA protocol 412023. The first study examined thermal discrimination. Stimuli were cool at 25°C, warm at 40–41°C, and hot at 43–44°C. The procedure used 18 randomized trials, six per level, after a 5-minute familiarization period, with constant 10 kPa contact pressure and approximately 20 s deflation between trials to promote cooling (Chen et al., 28 Aug 2025).

Thermal identification accuracy was 0.98 overall across the three levels, specifically 0.98 for cool, 0.95 for warm, and 1.00 for hot. Mean response times were 10.2 s for cool, 13.0 s for warm, and 11.0 s for hot, with an overall mean of 11.41 s. These values indicate that the interface supported categorical thermal discrimination under controlled contact pressure rather than only subjective warmth perception (Chen et al., 28 Aug 2025).

The second study evaluated a VR pick-and-place manipulation task using Ultraleap hand tracking and a virtual proxy–spring method. Indentation was mapped to chamber pressure up to 20 kPa, and two conditions were randomized: Haptic Feedback (HF) and No Feedback (NF). Each participant completed 15 trials per condition. Success required stable placement for at least 1 s, and metrics included success rate, time to success, and fingertip indentation during contact (Chen et al., 28 Aug 2025).

The reported outcomes showed that success rate improved from 88.5% in NF to 96.4% in HF, with $A$3. Time to success was 8.8 s in NF versus 9.2 s in HF, with $A$4, indicating no significant difference. Indentation, used as a behavioral proxy for force control, was reduced under HF from 9.3 mm to 6.4 mm, with $A$5. The interpretation in the source is that haptic feedback improved force control precision and manipulation success, even though completion time did not significantly change (Chen et al., 28 Aug 2025).

5. Position within the haptics literature

Hapt is situated within a larger shift toward lightweight, soft, and fabric-integrated interfaces. The underlying contrast in the source material is threefold. Compared with vibrotactile devices, Hapt adds true normal-pressure cues and real thermal sensations, rather than only high-frequency cues and motion illusions. Compared with kinesthetic force-feedback devices, it is markedly lighter and more compliant, whereas motorized fingertip interfaces are described as precise and strong but heavy, rigid, and power-hungry. Compared with thermal-only systems such as Peltier-based devices, it emphasizes ultralight compliant integration and rapid heating, while relying on passive cooling aided by clearance control (Chen et al., 28 Aug 2025).

This places Hapt in dialogue with other soft haptic platforms that prioritize manufacturability and distributed wearability. "FingerPrint" is a fully 3-D printed, soft, monolithic fingertip device based on an origami waterbomb base that delivers four degrees of freedom of cutaneous stimulation, including approximately 7 N normal force and approximately 1.2–1.3 N shear, but it is a vacuum-actuated printed thimble-scale mechanism rather than a thermal–pneumatic textile interface (Zhakypov et al., 2022). "Hoxels" extends a related logic to the wrist and forearm through fully 3-D printed, soft, wearable haptic voxels that produce x–y shear and z pressure, with blocked forces up to 1.6 N laterally and 20 N normally, but it addresses distributed wrist displays rather than fingertip thermal realism (Zhakypov et al., 2022).

At a broader systems level, haptics is defined as tactile feedback technology that applies forces, vibrations, and/or motions to the user, and haptic rendering is typically framed as collision detection, collision response, and actuator control in a tight servo loop (Yadav et al., 2013). Hapt follows that general formulation but narrows it toward co-located thermal and pressure rendering at the fingerpad. Its distinguishing contribution is not generic force display, but congruent multimodal cueing in a fully soft textile form factor (Chen et al., 28 Aug 2025).

6. Applications, scaling, and unresolved problems

The application domains explicitly identified for Hapt are VR and teleoperation. In VR, the device is proposed for material identification and immersion, such as distinguishing “warm ceramic vs cool metal” together with squeeze, as well as training simulations that require gentle force modulation and gaming interactions with thermal context. In teleoperation, the cited scenarios involve remote handling where thermal cues indicate process states, such as warm workpieces or heated tools, while pressure cues guide grasp control and may reduce slip and overgrip through tactile feedback loops (Chen et al., 28 Aug 2025).

The device is also described as scalable. The pouch-plus-heater module is fabric-printable and can be replicated across multiple fingers or body sites, while wrist hubs can multiplex valves and drivers for multi-module systems. Integration can be driven by vision-based or glove-based tracking, with indentation mapped to proportional pressure control and thermal cues rendered as event-driven levels or setpoints. Wireless data via UDP/WiFi and compact control hardware are noted as supporting mobile use (Chen et al., 28 Aug 2025).

Open challenges are explicitly identified. Bidirectional thermal rendering remains unresolved, with active cooling via microblowers, high-efficiency heat sinking, or fluidic systems named as possible directions. Latency and control bandwidth in the pneumatic loop require further characterization, and embedded pressure sensors together with model-based control are proposed for tighter haptic loops. Materials optimization is also open, including conductive yarns, copper foils, and liquid metal composites for heating uniformity and durability. Additional challenges include scaling to more fingers and cues such as vibration and shear while preserving comfort and energy efficiency, and long-session thermal management with adaptive ramp profiles responsive to skin state and user sensitivity (Chen et al., 28 Aug 2025).

Taken together, these constraints show that Hapt is less a finished endpoint than a specific design point within soft multimodal haptics: an ultra-light textile interface that demonstrates co-located heat and pressure at the fingertip, validates measurable perceptual and task-level benefits, and frames future work around active cooling, higher-bandwidth control, and broader multimodal scaling (Chen et al., 28 Aug 2025).