Skin-integrated stretchable actuators toward skin-compatible haptic feedback and closed-loop human-machine interactions

The study addresses the gap in human–machine interfaces (HMIs) where feedback from machines to users—particularly haptic feedback—remains underexploited compared to sensing and control. Rigid electromagnetic actuators offer strong performance but are bulky and hinder natural movement; soft actuators promise better wearability but face challenges in stretchability and actuation performance. Existing devices also fall short in achieving the skin’s force and spatial perception requirements, including the two-point discrimination (2PD) thresholds at highly sensitive skin sites. The research aim is to develop skin-compatible, stretchable haptic actuators that exceed tactile perception thresholds, achieve high spatial resolution compatible with human 2PD, and maintain performance under skin-like strains, thereby enabling effective closed-loop HMIs.

Prior haptic systems largely rely on rigid electromagnetic actuators that impede dexterity. Soft actuator approaches include: (1) Electrostimulation—simple structures but variable skin impedance and risks of discomfort or lesions; (2) Externally force-driven (pneumatic/hydraulic)—capable of large deformations/forces but bulky due to pumps, valves, and tubing; (3) Stimuli-responsive material actuators (e.g., dielectric elastomers)—thin and light but needing high voltages; advancements include hydraulically amplified self-healing electrostatic actuators (HASEL/HESEL) and hydraulically amplified taxels (HEXEL), which trade off pitch size, displacement, and force. Reported arrays often have pitch ≥10–13 mm, insufficient for areas with 2–4 mm 2PD (e.g., fingertips). Overall, existing cutaneous haptic devices do not adequately meet skin’s force and spatial perception simultaneously in compact, stretchable formats, limiting real-world adoption.

Device concept: A vertical hydraulically amplified electrostatic actuator that separates the high-voltage oil-filled pouch from the skin. The actuator comprises: (1) an oil-filled TPU pouch with printed Ag electrodes on both inner surfaces containing dielectric silicone oil; (2) a central opening capped by a stretchable silicone membrane (PDMS:Ecoflex0030 blend) as the activation region; (3) a soft pressure-amplification pillar on the membrane to increase delivered pressure without raising voltage; and (4) a neck structure to reduce flow resistance and improve hydraulic conversion efficiency while minimizing device height. Operating principle: Electrostatic attraction between pouch electrodes initiates a bottom-up zipping motion, expelling oil upward into the activation region; the membrane bulges out-of-plane, and the pillar concentrates pressure at the skin interface. Design modeling: Maxwell stress governs pouch pressure P = ε0 εeq U^2 / (2 d^2), with εeq determined by dielectric layers and oil gap. Fluid conversion rate α = Vbump / Vpouch, with Vbump = π r^2 D. Materials and fabrication: TPU films (HM65 shells; HM75 seal) form the pouch via plasma activation, screen-printed Ag electrodes, and heat lamination. The stretchable membrane is a PDMS:Ecoflex0030 mixture (weight ratio 1:5), spin-coated and cured, bonded to TPU after plasma and APTMS surface treatment. Pressure-amplification pillars are fabricated by magnetic field-assisted self-assembly of NdFeB microparticles in PDMS (7:1) forming pillar arrays via Rosensweig instability, then cured and demagnetized. Final assembly includes neck/seal integration, oil injection, and pillar bonding. Sensor development: Resistive pressure sensors use porous silicone skeletons (PDMS or Ecoflex0030) created via sacrificial sugar templates; the skeletons are coated by dip-coating with CB/Ecoflex to render conductivity and pressure sensitivity, and sandwiched between Ag-coated PET electrodes. Optimization studies: - Membrane composition and thickness: tested PDMS:Ecoflex ratios and thicknesses (40–100 µm). Ratio 1:5 and 80 µm thickness selected to balance displacement, durability, and bonding. - Pouch parameters: electrode distance optimized via oil volume; 170 µm selected to balance displacement vs fluid conversion/pressure. Pouch height optimized at 4 mm (vs 3 or 5 mm). Pouch width optimized at 10 mm (vs 6 or 8 mm). - Opening diameter: 2 mm chosen over 1.5 mm to maximize displacement with similar pressure. - Pillar geometry: selected contact area diameter ~130 µm, base diameter ~660 µm, height ~800 µm to reduce buckling and device height. Mechanical testing: Measured out-of-plane displacement, output pressure, and computed fluid conversion under varied geometries. Evaluated performance under 15% tensile strain. User perception study: Assessed tactile detectability at multiple skin locations (e.g., finger, hand back, others) with and without pillars; 20 trials per site; participants indicated perceived stimuli. Closed-loop demonstrations: - Surface texture recognition: A porous pressure sensor on a robot finger scanned 1-, 2-, and 4-wedge surfaces at 10 mm/s. Resistance peaks triggered actuator drive on a user’s finger. Small variations (ΔR < 0.2) were ignored; larger changes normalized to 100% of 4 kV; voltage applied during rising phases only. Blindfolded, acoustically shielded users underwent training then 15 randomized test groups each (total 135 trials across 3 users). - Object shape recognition: A sensor array on a robot palm detected spatial resistance patterns from sphere, bar, and cube. A matched-layout actuator array on the user’s upper arm reproduced spatiotemporal cues via voltage modulation. Each shape presented 30 times (90 total tests). Safety: The skin-contact materials (PDMS, Ecoflex, TPU) are insulating; current limited below 0.5 mA by the amplifier, with automatic shutdown above threshold.

- Spatial resolution and stretchability: Actuator arrays achieved 3 mm pitch, meeting fingertip 2PD requirements, and remained functional under 15% tensile strain (skin-like elastic range). - Output performance: Optimized devices delivered ~500 µm out-of-plane displacement and ~1 MPa output pressure, exceeding tactile perception thresholds (≈100 µm displacement; ≈150 kPa pressure depending on site). - Geometry optimization: Selected parameters—electrode gap 170 µm, pouch height 4 mm, pouch width 10 mm, opening diameter 2 mm, membrane PDMS:Ecoflex0030 (1:5) at ~80 µm thickness, pillar geometry with ~130 µm contact diameter, ~660 µm base diameter, ~800 µm height. - Pressure amplification effectiveness: With pillars, output pressures increased substantially versus without pillars, enabling reliable perception across less sensitive skin areas (e.g., hand back). Without pillars, actuators generally failed to trigger perception except at the most sensitive sites (e.g., finger). - Under 15% stretch: Displacement and pressure decreased relative to unstretched state but remained above tactile thresholds. - Sensor performance: Porous Ecoflex with larger pore size yielded higher pressure sensitivity; resistance vs pressure exhibited monotonic changes suitable for modulation. - Surface texture recognition: Closed-loop system enabled users to identify 1-, 2-, and 4-wedge textures with ≥97.8% accuracy over 135 trials. - Object shape recognition: Users correctly identified sphere and bar at 100% accuracy; cube at ~86.7% accuracy over 90 trials. - User safety and comfort: No pain reported during studies; current limited to <0.5 mA; voltage can be reduced to avoid approaching pressure-pain thresholds.

The developed pressure-amplified, stretchable electrostatic actuators directly address skin-relevant constraints—force/displacement thresholds, fine spatial acuity, and conformal stretchability—enabling practical cutaneous haptics in closed-loop HMIs. The vertical pouch architecture enhances safety by distancing high-voltage components from the skin, while the neck improves hydraulic conversion efficiency in a compact form factor. Magnetic self-assembled pillars concentrate pressure without increasing voltage, expanding effective stimulation to less sensitive skin regions. Optimization of membrane mechanics and pouch geometry balances displacement, pressure, and durability, and supports reliable performance under skin-like strains. Integrating highly sensitive porous resistive sensors provides an effective sensing-to-stimulation pathway. The high accuracies in surface texture and object shape recognition demonstrate that the system can faithfully transmit tactile information from robots to users, facilitating bilateral interaction in applications such as prosthetics, teleoperation, and VR/AR. Remaining variations in perception thresholds across users and sites underscore the need for adaptive calibration, but the system’s performance margins (e.g., ~1 MPa pressure) offer flexibility to tune stimuli within comfortable ranges.

This work introduces pressure-amplified, skin-integrated stretchable electrostatic actuators paired with porous resistive sensors to realize skin-compatible haptic feedback and closed-loop human–machine interaction. The actuators achieve ~500 µm displacement, ~1 MPa pressure, 3 mm spatial resolution, and 15% stretchability, meeting or surpassing skin tactile thresholds and enabling deployment on highly sensitive skin areas. User studies validate effective transmission of surface texture and object shape information with high accuracy. These results position the approach for applications in robotics, prosthetics, telemedicine, and immersive environments. Future work can focus on scaling array sizes for richer haptic patterns, further reducing operating voltages while maintaining output, enhancing robustness under long-term cyclic strains, and implementing adaptive control to personalize stimuli across users and body sites.

- Performance under strain: Out-of-plane displacement and output pressure decrease under 15% tensile strain, though remaining above tactile thresholds. - Membrane durability trade-off: Very thin membranes (≈40–50 µm) are fragile; thicker membranes reduce displacement, necessitating a balance (≈80 µm chosen). - User/site variability: Tactile pressure thresholds vary significantly across skin locations and individuals, affecting predictability and requiring calibration. - Shape recognition variance: Lower recognition accuracy for the cube (~86.7%) indicates shape-dependent perceptual limits at current resolution and mapping. - High-voltage drive: Although current is limited for safety and the design isolates the skin, the system requires kilovolt-level actuation, motivating further reduction of operating voltage. - Minor gravity effects were tested and found insignificant, but overall performance depends on careful geometric optimization (e.g., pouch angle θ, height) and may be sensitive to fabrication tolerances.