PaperlayerView exact revision on arXiv

Source-linked AI summary

TherMosaic: Accelerating Perceived Thermal Transitions Through Spatiotemporal Thermal Feedback

Zining Zhang, Jiasheng Li, Myungin Lee, Zeyu Yan, Jin Ryong Kim, Huaishu Peng

arXiv:2609.05347v1cs.HC

TL;DR

Thermal feedback is often too slow to align with interactive events. TherMosaic uses coordinated fingertip subregions and the perceptual mechanisms of spatial summation and thermal adaptation; across perceptual studies and VR evaluation, it reduces perceived transition time and thermal lag while improving temporal alignment.

  • Problem

    Thermoelectric thermal interfaces often change temperature too slowly to match interactive timing, limiting temporal alignment with user actions and virtual events.

  • Method

    TherMosaic coordinates multiple independently controlled fingertip thermal subregions using spatial summation and thermal adaptation to accelerate perceived transitions.

  • Results

    TherMosaic reduces perceived thermal transition time and improves temporal alignment in interactive VR use across three perceptual studies and a wearable evaluation.

  • Takeaways & Limitations

    Thermal latency can be addressed through spatiotemporal thermal design in addition to hardware improvements.

  • Takeaways & Limitations

    The prototype targets a single fingertip, limiting the range of interactions it can support.

Abstract

from arXiv · show

Thermal feedback can enrich immersive interaction, but thermoelectric devices often change temperature too slowly to match interactive timing. We present TherMosaic, a spatiotemporal thermal feedback approach that accelerates perceived temperature transitions by leveraging two perceptual mechanisms: spatial summation and thermal adaptation. Focusing on the fingertip, we first investigate this approach using a custom 2*2 array of independently controlled Peltier modules. Across three controlled perceptual studies, we show that distributed thermal stimulation can preserve stable hot and cold percepts despite local deviations, that adaptation helps maintain these percepts during changing stimulation, and that combining these effects reduces perceived transition time by about 30%-40% for transitions originating from hot or cold states. We then translate the same design principles into a standalone wearable implementation of TherMosaic and evaluate it in virtual reality. Our results show that this approach reduces perceived thermal lag and improves temporal alignment between thermal and visual events in interactive use.

1 Introduction

TherMosaic reframes slow thermal transitions as a perceptual-design problem rather than solely a hardware problem. It coordinates fingertip subregions using spatial summation and thermal adaptation, then validates the approach through perceptual studies and a wearable VR implementation.

  • Thermal actuators often require several seconds or longer to switch states, creating temporal mismatch with user actions and virtual events.
  • TherMosaic coordinates independently controlled fingertip subregions to accelerate perceived transitions without directly increasing physical temperature ramp rates.The approach shifts some skin toward the target while preserving the current overall percept.
  • Spatial summation integrates distributed thermal inputs into a unified percept, while thermal adaptation reduces sensitivity to ongoing local changes.
  • TherMosaic extends heterogeneous thermal stimulation to calibrated transitions across the full 20–40 °C operating range and participant-specific cold, neutral, and hot categories.
  • Three controlled perceptual studies test whether distributed deviations remain categorical, adaptation stabilizes changing stimulation, and combined effects reduce perceived transition time.
  • A standalone wearable implementation demonstrates improved temporal alignment relative to a baseline thermal interface in virtual reality.

2 Related Work

Thermal-interface research has explored many actuation methods, but slow heat transfer still limits temporal responsiveness. TherMosaic addresses this limitation through perceptual design, building on spatial summation, thermal adaptation, and prior heterogeneous-stimulation work.

  • Peltier modules are widely used because they provide compact, bidirectional heating and cooling across many thermal-interface form factors.
  • Prior systems also use pneumatic, liquid-based, ultrasonic, hydro-optical, chemical, and physiological approaches to produce thermal sensations.
  • Conventional Peltier systems often require several seconds to produce perceptible changes, causing thermal feedback to misalign with interaction events.
  • Hardware approaches such as forced air circulation and mechanically alternating preconditioned Peltiers can accelerate transitions but may require larger or more complex systems.
  • TherMosaic instead investigates whether perceptual design can accelerate thermal transitions without adding hardware complexity.
  • Spatial summation integrates distributed inputs, while thermal adaptation can make gradual changes less perceptible and reduce apparent distance between thermal states.
  • Compared with prior 2–4 °C heterogeneous stimulation, TherMosaic uses continuous-contact, participant-calibrated transitions across the 20–40 °C safe operating range.

3 Perceptual Basis and Study Logic

TherMosaic treats fingertip transitions as category crossings among cold, neutral, and hot rather than precise physical-temperature changes. Its study logic tests whether spatial summation and adaptation preserve the dominant category while subregions move toward a target.

  • TherMosaic aims to reduce the physical change needed for a final perceptual transition by shifting some subregions toward the upcoming target while the overall percept remains stable.
  • The framework defines thermal transitions as crossings between fuzzy cold, neutral, and hot perceptual categories rather than exact physical-temperature changes.
  • The approach depends on spatial summation integrating heterogeneous inputs into one dominant percept and adaptation masking partial local changes.
  • Study 1 tests categorical integration, Study 2 tests category stability during adaptation, and Study 3 tests the combined effect on transition time.
  • The perceptual tests precede translation of the same design logic into a standalone wearable implementation.

4 Benchtop Experimental Setup

The benchtop platform uses a compact 2×2 array of independently controlled Peltier elements to manipulate fingertip thermal subregions. Its sensing, cooling, control, and safety systems support stable, spatially patterned stimulation for the perceptual studies.

  • The benchtop setup uses four independently controlled miniaturized Peltier elements to manipulate local fingertip thermal subregions.
  • The actuator combines four RTDs in copper blocks, four Peltier modules, and a water block for temperature measurement and thermal management.
  • The 2×2 layout provides spatially non-uniform patterns while remaining compact enough for stable fingertip contact.
  • Approximately 3.2 mm × 3.2 mm Peltier modules were selected to match the estimated 3.5 mm subregions of the fingertip contact area.
  • Diagonal grouping distributes two interleaved thermal groups across the array without forming a localized cluster.
  • Embedded RTDs measure interface temperature, while a customized water block independently cools each Peltier in a closed loop.
  • Asymmetric PID control uses separate heating and cooling gains to balance responsiveness and stability.
  • The actuators operate within 20–40 °C with overheating protection that reduces or suspends output when surface temperature is excessive.

5 Thermal Perceptual Studies

Three within-subject perceptual studies tested whether spatial integration and thermal adaptation preserve categorical fingertip percepts during heterogeneous stimulation and support masked temperature changes.

  • Study design: The three studies progressed from spatial integration, to maintaining a dominant percept during partial change, to detecting transitions between thermal categories.All studies used the same participant group and a benchtop fingertip setup with participant-specific thermal category thresholds.
  • Study 1: Spatial integration: Participants reported a single uniform sensation across all Study 1 trials and detected no localized variation among the four modules.Categorical identification remained near ceiling for neutral, cold, and hot stimuli under both uniform and diagonally split patterns.
  • Study 1: Spatial integration: 0.0 mean paired accuracy difference supported equivalence and noninferiority of the diagonally split pattern within the 5% equivalence margin.The 95% bootstrap confidence interval was [−3.1%, 3.3%].
  • Study 2: Perceptual stability: Balanced shifts yielded low change-detection probabilities of 6.3% for cold-origin and 4.2% for hot-origin trials, versus 85.1% and 91.6% in large-change controls.The corresponding odds ratios were 0.012 and 0.004, indicating that participants largely maintained the original dominant percept.
  • Study 2: Perceptual stability: Spatially balanced changes preserved dominant categorical percepts more reliably than unilateral shifts, consistent with spatial integration and adaptation masking gradual local changes.Unilateral shifts produced intermediate detection probabilities of 37.4% for cold-origin and 51.2% for hot-origin trials.

5.5 Study 3: Perceived Transition Time Between Thermal Categories

Study 3 tested whether diagonal preconditioning shortens perceived transitions between thermal categories. It reduced detection times for transitions beginning in established cold or hot states, but not for neutral-origin transitions.

  • Experimental design: Study 3 compared uniform baseline stimulation with diagonal preconditioning before six ordered transitions among cold, neutral, and hot categories.Trials began with a stable starting pattern, applied the test manipulation during Phase 2, and transitioned fully during Phase 3.
  • Overall analysis: The Group × Transition interaction was significant, F(5, 514) = 8.21, p < .001, η2_p = .074, indicating that effects differed across transition types.Group and Transition also showed significant main effects.
  • Non-neutral-origin transitions: For cold-origin transitions, detection time fell by 1.85 s for cold-to-hot and 1.94 s for cold-to-neutral transitions.Cold-to-hot decreased from 5.60 s to 3.75 s; cold-to-neutral decreased from 4.69 s to 2.75 s.
  • Non-neutral-origin transitions: For hot-origin transitions, detection time fell by 2.09 s for hot-to-cold and 1.29 s for hot-to-neutral transitions.Hot-to-cold decreased from 6.12 s to 4.03 s; hot-to-neutral decreased from 4.59 s to 3.30 s.
  • Neutral-origin transitions: Neutral-origin transitions showed no reliable benefit: Δ = −0.40 s for neutral-to-cold and Δ = 0.17 s for neutral-to-hot.Neither comparison was significant.
  • Non-neutral-origin transitions: 1.29–2.09 seconds: test conditions reduced detection time for cold- and hot-origin transitions by approximately 28%–41% relative to baseline.The benefit occurred when the initial thermal state was categorically established as cold or hot.

6 Wearable Implementation of TherMosaic

The validated TherMosaic thermal core was translated into a standalone fingertip-worn device for interactive evaluation while preserving the benchtop system’s independently controlled 2×2 architecture.

  • Thermal interface: The wearable preserves the validated 2×2 architecture, including four independently controlled 3.2 mm × 3.2 mm Peltier modules and a fingertip contact area of approximately 7 mm × 7 mm.Copper blocks contain embedded RTD sensors and water-cooling blocks.
  • Interactive deployment: The resulting device enables standalone fingertip wear for interactive evaluation while preserving the validated thermal interface.The implementation was used for the wearable evaluation described in the following section.
  • Standalone hardware: Standalone operation adds a fingertip-sized PCB, onboard control electronics, rechargeable battery, and custom 3D-printed enclosure.The device retains the benchtop water-cooling system and matches its ramp performance.

7 VR Evaluation

The VR evaluation compared TherMosaic with a uniform baseline during visually triggered hot-to-cold and cold-to-hot magic changes. TherMosaic improved perceived synchronization, made thermal transitions feel more abrupt, and reduced perceived thermal lag.

  • Apparatus and Procedure: Participants wore a fingertip TherMosaic device while switching between fire and ice magic in a VR scarecrow game.The study included 12 participants and repeated visually triggered thermal transitions in both directions.
  • Conditions: TherMosaic preconditioned one diagonal module group toward the upcoming temperature while the other preserved the current thermal category.The baseline rendered a uniform temperature across all four modules.
  • Synchronization: 19.58 points: TherMosaic increased synchronization VAS ratings over baseline, with a significant paired comparison and large effect size.The 95% confidence interval was [11.01, 28.15], with t(11) = 4.48, p = .001, and Cohen’s d = 1.29.
  • Transition Dynamics: TherMosaic made thermal transitions seem more sudden than baseline for both fire and ice changes.Baseline ratings were 6.00 ± 0.21 and 5.92 ± 0.19, versus 3.08 ± 0.43 and 3.33 ± 0.43 under TherMosaic, both p < .001.
  • Thermal Lag: 2.50 points: perceived thermal lag decreased under TherMosaic, from 5.67 ± 0.31 at baseline to 3.17 ± 0.53.The difference was significant at p = .003 and was consistent across transition directions.
  • Qualitative Feedback: Participants described TherMosaic transitions as better matched to visual feedback and less disruptive than slower baseline transitions.Qualitative comments linked thermal delay to uncertainty about whether the magic had been caught.

8 Discussion and Limitations

The discussion interprets TherMosaic as a perceptual alternative to faster hardware, while identifying scheduling, contact-area, and within-category intensity limits. Its effect is plausibly explained by spatial integration, categorical perception, and thermal adaptation.

  • Discussion: TherMosaic suggests that thermal latency can be addressed through spatiotemporal thermal design rather than hardware improvements alone.The interpretation is based on three perceptual studies and a VR evaluation.
  • Perceptual Mechanism: Spatial integration, categorical perception, and thermal adaptation may mask local changes while part of the fingertip shifts toward a target.Together, these properties may reduce the remaining physical change needed to cross into a new perceptual category.
  • Scheduling: TherMosaic works best when upcoming thermal transitions can be anticipated, limiting its use in dynamic or spontaneous interactions.More sophisticated scheduling may be required when transition timing is not explicitly controlled.
  • Scalability of Contact Area: The current prototype targets a single fingertip, which limits the range of supported interactions and motivates scaling to larger contact areas.Larger coverage may provide a more solid sensation, but wearable scaling introduces weight, cooling, and power challenges.
  • Intensity Invariance: TherMosaic does not guarantee perfect within-category intensity invariance, so future work should quantify subtle intensity changes directly.Suggested measures include continuous intensity ratings, JND thresholds, and pairwise discrimination tasks.

9 Conclusion

TherMosaic accelerates perceived fingertip thermal transitions by combining spatial summation and thermal adaptation. Across perceptual studies and VR evaluation, it reduced perceived transition time and improved temporal coherence in interactive use.

  • Conclusion: TherMosaic uses heterogeneous thermal stimulation to reduce perceived thermal transition time and improve temporal coherence during interaction.The findings position thermal latency as an opportunity for perceptually informed interface design, not solely a hardware limitation.

A Ramp-Rate Data

The ramp-rate appendix characterizes the physical behavior of the four-module system during heating, cooling, and representative thermal transitions. It reports averaged responses, module-level variability, and baseline-versus-TherMosaic profiles using repeated trials.

  • Step Response: 25–35 °C: repeated square-wave setpoints characterized the average four-module step response after steady-state holds.Heating and cooling rates were estimated from approximately the 10%–90% step-amplitude region using linear fits.
  • Multi-module Transient Behavior: High-to-low-to-neutral sequences measured individual module temperatures to assess inter-module consistency and variability.The four module responses were recorded and plotted separately.
  • Representative Transitions: 2 °C/s: representative cold-to-hot and hot-to-cold profiles used the same fixed ramp rate as Study 3.These profiles correspond to the transitions shown in Figures 11 and 12.
  • Experimental Design: Five repetitions per condition and transition type produced mean module curves with SEM shading for baseline and TherMosaic.The design enables direct comparison of ramp rate and spatial temperature distribution over time.

B Thermal Calibration Procedure

The study calibrated four participant-specific thermal references to define reliable cold, neutral, and hot categories before testing TherMosaic. Calibration used discrete fingertip trials, adaptive temperature adjustments, and randomized verification.

  • Four reference temperatures—cold anchor, cold–neutral boundary, neutral–hot boundary, and hot anchor—defined each participant’s thermal categories.These references established the category thresholds used in the experiment.
  • Participants began at approximately 30 °C, with all stimuli constrained to a safe 20–40 °C operating range.
  • Each discrete trial lasted 4 s, after which participants classified the sensation as cold, neutral, or hot; researchers adjusted temperature in 1 °C increments.
  • Researchers located the cold–neutral boundary by increasing or decreasing temperature until participants consistently judged the lowest neutral temperature.
  • The cold and hot anchors were selected as temperatures consistently judged within their category and positioned beyond the corresponding category boundary.
  • Before the main experiment, each calibrated temperature was presented repeatedly in random order, and participants had to classify it consistently.
Loading 2609.05347v1…