FiDTouch: 3-DoF Wearable Finger Haptic Display

• FiDTouch is a wearable 3-DoF cutaneous haptic display that uses a tiny inverted Delta robot to deliver precise tactile cues on the finger pad.
• It renders multidimensional sensations—including contact, pressure, skin stretch, and vibrotactile feedback—to enhance user interaction in VR, teleoperation, and rehabilitation.
• Experimental results show recognition rates up to 83% for directional skin stretch, demonstrating its effectiveness in conveying spatial and dynamic touch cues.

FiDTouch is a wearable 3-DoF cutaneous haptic display for the finger pad that uses a tiny inverted Delta robot to deliver localized and spatially controllable touch cues to the distal phalanx, including contact, pressure, encounter, skin stretch, and vibrotactile feedback (Trinitatova et al., 10 Jul 2025). It is positioned as a compact, low-cost device for virtual reality, teleoperation, medical training, and rehabilitation. Its central premise is that the finger pad is the primary site for perceiving contact location, pressure, tangential slip, and surface interaction during manipulation and probing, so direct stimulation of that surface can make virtual or remote interaction more realistic and can improve dexterity and situational awareness.

1. Scope and nomenclature

FiDTouch denotes the system introduced in “FiDTouch: A 3D Wearable Haptic Display for the Finger Pad” (Trinitatova et al., 10 Jul 2025). It should not be conflated with the unrelated wireless-device identification system “FID: Function Modeling-based Data-Independent and Channel-Robust Physical-Layer Identification,” which concerns RF fingerprinting for IoT transmitters and does not mention “FiDTouch,” touch-based interaction, or any touch-oriented implementation variant (Zheng et al., 2019).

Within haptics and human-computer interaction, FiDTouch belongs to the class of wearable mechanical fingertip or finger-pad displays. The paper’s stated novelty is the use of a tiny inverted Delta robot, abbreviated iDelta, as a parallel mechanism that positions a point contact in three dimensions over a useful area of the finger pad while maintaining a wearable form factor. The device is worn on the distal phalanx so that the end effector moves over an almost flat portion of the finger pad rather than the most curved fingertip apex. This design choice is presented as important for more controlled spatial stimulation and skin-stretch rendering.

The paper explicitly frames FiDTouch as a cutaneous display rather than a kinesthetic exoskeleton. Its output modalities are localized and surface-oriented: contact, pressure, encounter, skin stretch, shear, and vibration. This suggests that its intended operating regime is the rendering of contact-rich fingertip sensations rather than whole-finger joint torques or large workspace arm-level force feedback.

2. Mechanism, structure, and electronics

Mechanically, FiDTouch is built around a tiny inverted Delta parallel robot driven by three DM-S0020 micro servo motors. The base is an equilateral triangle with a central circular hole through which the mechanism accesses the finger pad: diameter $16\,\mathrm{mm}$ for the index finger and $18\,\mathrm{mm}$ for the thumb. The device is attached to the finger using elastic tapes via two fasteners on the base. The whole mechanism is enclosed in a box-like volume of $50 \times 50 \times 32\,\mathrm{mm}$, and the total device weight is $20\,\mathrm{g}$ (Trinitatova et al., 10 Jul 2025).

Component	Specification	Function
End effector	Metal ball, diameter $3\,\mathrm{mm}$	Localized point contact
Vibrotactile unit	Coin vibration motor, diameter $8\,\mathrm{mm}$	Vibrotactile feedback
Actuators	Three DM-S0020 micro servo motors	3 translational DoF
Electronics	ESP32 + PCA9685 PWM driver	Servo and vibration control

The structure uses 3D-printed PLA parts for the base and motor holders, stainless-steel rods for the links, and metal spherical joints between the links and moving platform. Each DM-S0020 motor weighs $2.1\,\mathrm{g}$, has dimensions $23 \times 8 \times 17\,\mathrm{mm}$, and maximum torque $0.2\,\mathrm{kg\cdot cm}$. The control electronics consist of an ESP32 microcontroller and a PCA9685 I$^2$C-controlled PWM driver, which drives the three micro servos and the vibration motor. The vibration motor is driven by PWM control to generate different vibrotactile cues.

The active mechanism has 3 translational degrees of freedom. Although the iDelta mechanism itself can achieve a larger reachable space, the authors intentionally constrain operation to a circular area of diameter $18\,\mathrm{mm}$0 within the base plate, with vertical displacement of $18\,\mathrm{mm}$1. They also report a finger-pad work area of $18\,\mathrm{mm}$2–$18\,\mathrm{mm}$3. The non-contact approach position used in the user studies was $18\,\mathrm{mm}$4 above the finger pad before moving down into contact, consistent with the available vertical range.

The device’s maximum normal force is reported as $18\,\mathrm{mm}$5. At the same time, the paper does not report shear force limits, output bandwidth, update rate, closed-loop latency, positional accuracy, repeatability, maximum end-effector speed, acceleration, or spatial resolution in a formal metrological sense. It also does not provide full derivations of the inverse or forward kinematics, Jacobians, force models, or control laws; no explicit mathematical equations are included in the paper text. The practical sensing implied is servo actuation and commanded positioning rather than measured contact force sensing.

3. Rendering principles and tactile modalities

FiDTouch is explicitly designed to render contact, pressure, encounter, skin stretch, and vibrotactile feedback, and more generally normal and shear forces with arbitrary pressure at the contact point (Trinitatova et al., 10 Jul 2025). The device achieves this by moving a $18\,\mathrm{mm}$6 spherical tactor laterally over the finger pad and vertically into or out of contact.

Localized contact is generated by positioning the ball at a selected lateral $18\,\mathrm{mm}$7 location and lowering it from the non-contact position into the skin. Because the mechanism can position the ball over different parts of the finger pad, contact can be perceived at different spatial points. Pressure is rendered by the same vertical-contact mechanism with controllable indentation or contact force. The paper does not give a force-control law, but it states that the 3-DoF inverted Delta structure allows free movement of the end effector inside the workspace and generation of a 3D force vector at the contact point.

Encounter feedback refers to the sensation of touching an object as the finger approaches it. FiDTouch creates encounter cues by moving the end effector quickly through free space to a desired position and then bringing it into contact at the right time and location, while avoiding unnecessary skin contact during transit. The paper specifically emphasizes that the parallel robot can move rapidly within the workspace while avoiding unintended contact, enabling dynamic interactions with a virtual environment.

Skin stretch is generated by first contacting the finger pad and then translating the end effector tangentially across the skin in a desired direction. Because the end effector is a small ball, lateral motion produces localized shear deformation. The study tested eight directions separated by $18\,\mathrm{mm}$8: $18\,\mathrm{mm}$9. Shear-force feedback more generally is produced by laterally moving the contact ball while it remains in contact with the skin, so the resultant contact force can have both normal and tangential components.

Vibrotactile feedback is generated by the coin-type vibration motor attached to the moving platform. Because the motor is carried with the end effector assembly, vibration can be delivered at the active contact location. The authors describe this as suitable for multimodal tactile rendering, including the combination of texture-like vibration cues with force or contact rendering. However, the paper does not report vibration frequency range, amplitude range, or spectral performance. This suggests that the present contribution is primarily a mechanism-and-perception paper rather than a full dynamic characterization of tactile output.

4. Experimental evaluation and perceptual performance

The empirical evaluation is a two-stage user study focused on discrimination of static spatial contact stimuli and directional skin-stretch stimuli on the finger pad (Trinitatova et al., 10 Jul 2025). Sixteen participants took part in both experiments; five were female, the age range was 19 to 30 years, the mean was $50 \times 50 \times 32\,\mathrm{mm}$0 years, the standard deviation was $50 \times 50 \times 32\,\mathrm{mm}$1, and two participants were left-handed. All participants gave informed consent, the study had local IRB approval, and participants had no prior experience with the device. In both studies, participants sat at a desk, wore FiDTouch on the index finger of the dominant hand, and the device was hidden from view. Participants were provided visual guides showing the tactile patterns to be identified.

Study	Stimuli	Main result
Static spatial contact	$50 \times 50 \times 32\,\mathrm{mm}$2	Mean recognition rate $50 \times 50 \times 32\,\mathrm{mm}$3
Directional skin stretch	$50 \times 50 \times 32\,\mathrm{mm}$4	Mean recognition rate $50 \times 50 \times 32\,\mathrm{mm}$5

In the first experiment, nine contact-point patterns were defined over the finger-pad workspace: $50 \times 50 \times 32\,\mathrm{mm}$6. The contact force for each stimulus was calibrated so the participant perceived it as comfortable. In each trial, the end effector first moved to the desired lateral location while held in a non-contact position $50 \times 50 \times 32\,\mathrm{mm}$7 above the finger pad, then moved downward to deliver a contact stimulus for $50 \times 50 \times 32\,\mathrm{mm}$8, and then returned to the non-contact position. Each of the nine patterns was presented five times in random order, for 45 stimuli per participant.

Results of the first study showed that the average recognition rate across patterns ranged from $50 \times 50 \times 32\,\mathrm{mm}$9 to $20\,\mathrm{g}$0, with mean correct rate $20\,\mathrm{g}$1. Average confidence was $20\,\mathrm{g}$2 on a five-point scale. The most recognizable points were $20\,\mathrm{g}$3, $20\,\mathrm{g}$4, and $20\,\mathrm{g}$5, while the least recognizable was $20\,\mathrm{g}$6. The paper notes a general asymmetry: points on the right side of the finger pad were recognized better than those on the left. The confusion matrix showed four pairs with error rates of $20\,\mathrm{g}$7 or more: UR confused with R, UL confused with U, DR confused with R, and DL confused with L. Using a Kruskal–Wallis test at significance level $20\,\mathrm{g}$8, the authors report $20\,\mathrm{g}$9, showing a statistically significant difference in recognition rates across contact positions.

In the second experiment, eight skin-stretch patterns were used, corresponding to the eight movement directions $3\,\mathrm{mm}$0. In each trial, the end effector moved to the start position $3\,\mathrm{mm}$1 above the finger pad, then moved down into contact, translated along the designated direction to stretch the skin, and then returned to the center non-contact position. Each pattern was presented five times in random order, for 40 stimuli per participant.

In this second study, mean confidence was $3\,\mathrm{mm}$2, and mean recognition rate was $3\,\mathrm{mm}$3. The most distinctive patterns were straight directions: $3\,\mathrm{mm}$4, $3\,\mathrm{mm}$5, and $3\,\mathrm{mm}$6. The most confusing pattern was $3\,\mathrm{mm}$7, which was confused with R and D in $3\,\mathrm{mm}$8 of cases each. The paper remarks that diagonal patterns $3\,\mathrm{mm}$9, $8\,\mathrm{mm}$0, and $8\,\mathrm{mm}$1 were hard for some participants, with some individual recognition rates as low as $8\,\mathrm{mm}$2. Using a Kruskal–Wallis test, the authors found no statistically significant difference among recognition rates across the eight skin-stretch directions: $8\,\mathrm{mm}$3.

Taken together, the results indicate that spatially separated static point contacts on the finger pad can be distinguished with fairly high accuracy, but not uniformly across all locations, while directional skin stretch is highly discriminable overall and performed better on average than static point localization in this setup. The paper specifically interprets this as supporting the utility of FiDTouch for directional cues, guidance, slip simulation, and dynamic contact rendering.

5. Research position and relation to adjacent touch systems

Within the paper’s own prior-art framing, FiDTouch sits among wearable mechanical fingertip displays but tries to combine 3D cutaneous rendering, small localized contact, dynamic responsiveness, vibrotactile integration, and compactness (Trinitatova et al., 10 Jul 2025). The prior-art discussion includes devices using parallel mechanisms, origami-inspired structures, electro-tactile arrays, pneumatic or hydraulic systems, and finger-pad-free designs. The authors state that, compared with many fingertip devices that stimulate mainly one point or one force direction, FiDTouch offers 3-DoF translational control of a point contact over the finger pad. Compared with hRing and other finger-pad-free systems, it directly stimulates the finger pad itself. Compared with pure vibrotactile or electro-tactile systems, it renders mechanical skin deformation and localized contact. Compared with pneumatic or hydraulic designs, it offers rigid-body positioning and point-contact localization rather than distributed pressure via inflation.

The paper also explicitly claims several advantages over Schorr et al.’s Delta-based 3-DoF fingertip device: a larger working area while keeping comparable size and output forces, and the use of low-cost, easily controlled servo motors instead of brushless DC motors. The integrated vibration motor on the moving platform further distinguishes the system as a multimodal display rather than a purely mechanical indenter.

In a broader arXiv context, FiDTouch occupies a different part of the touch-research landscape from resources such as OpenTouch, which is a dataset-and-benchmark paper for egocentric full-hand tactile sensing rather than a wearable display (Song et al., 18 Dec 2025). A plausible implication is that these lines of work are complementary: FiDTouch concerns tactile output to the finger pad, whereas OpenTouch concerns measured tactile input from the hand during natural interaction. FiDTouch is likewise distinct from touchscreen interaction methods that expand input expressiveness through finger identity or finger pose, such as “Glass+Skin” and BiFingerPose, which focus on touch input rather than tactile rendering (Roy et al., 2019). This suggests that FiDTouch is best understood as a fingertip output device within a wider ecosystem of touch sensing, touch interaction, and tactile-data modeling.

6. Limitations, open issues, and future directions

The paper identifies several clear limitations of the present FiDTouch configuration (Trinitatova et al., 10 Jul 2025). The device remains somewhat bulky relative to everyday wearable form factors at $8\,\mathrm{mm}$4, even though it is light at $8\,\mathrm{mm}$5. The active contact area is deliberately limited to a $8\,\mathrm{mm}$6 diameter region and does not cover the most curved fingertip apex. There is no integrated force sensing or closed-loop force control described, which may limit precision in rendering exact contact mechanics. The paper also does not provide detailed performance characterization of dynamics, bandwidth, latency, or accuracy, so claims of fast dynamic rendering are not quantitatively substantiated.

The user studies were restricted to perceptual discrimination of predefined patterns rather than functional task performance in virtual reality, teleoperation, medical simulation, or rehabilitation. The sample size is modest, no long-term comfort or fatigue analysis is reported, and diagonal stimuli were less robustly perceived than central, vertical, or cardinal-direction stimuli. These findings delimit the present evidence: FiDTouch has been validated as a spatial and directional cue display on the finger pad, but not yet as a task-level haptic interface in realistic interactive scenarios.

The authors propose a more thorough user study on multimodal feedback discrimination and the exploration of tactile cues in telemanipulation tasks. The examples named are rendering interactions with rigid and compliant objects and simulating an object slipping out of a gripper, with the goal of improving the naturalness of human-robot interaction. This suggests that the next stage of the line of work is likely to shift from stimulus discriminability toward contact-rich task execution, where the practical value of contact, pressure, encounter, skin stretch, and vibration can be assessed under more application-specific conditions.