In this work we investigate how the tactile detection thresholds of air-driven actuators are influenced by design parameters as membrane thickness and diameter, when the finger is in contact with the deformable circular elastic membranes. We report lower thresholds with decreasing membrane thickness and increasing diameters. These results can help in the design of soft actuators for pin-arrays used as graphical tactile displays for visually impaired people.
In general, the tactile display market was static in the recent
years and no breakthrough is being observed. However, this
topic attracts a lot of attention and many research centers
and companies are dealing with large area tactile displays to
make the digital world more accessible to the visually
impaired community or to provide eyes-free
computerhuman interfaces. Information is generally organized as
small pin-shaped actuators, named taxels as they can be
considered the tactile equivalent of the pixel [
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
TacTT2014, November 16th, 2014, Dresden, Germany.
Copyright is held by the owner/author(s).
Novel approaches [
Most of the MEMS-based soft actuation mechanisms have
been applied to Braille geometries, mainly because Braille
dots are small, but very few of the considered architectures
are purely designed for tactile graphics, without necessarily
providing text [
When dealing with perception of tactile graphics, as
compared to Braille reading, the objective shifts from
understanding information coded as single, separated small
dots to picturing a sequence of symbols forming
geometrical primitives. The specifications of Braille can
therefore be relaxed: on the one hand, graphical raised line
drawings can be perceived as low as 200 µm [
Therefore, the question is if one can build controllable actuators able to deliver sufficient force, enough stroke, without increasing too much the dot diameter, which would limit the resolution and as a consequence the number and type of displayed information.
Pneumatically-driven actuators [
Amplifier
NI card
PC+MATLAB
We also considered different thicknesses of the contact membrane, namely 200, 600, 1000 and 2000 µm. Our goal was to derive minimal physical requirements, in terms of force per area unit, so that a single dot is statistically distinguished from an otherwise smooth, non-rigid surface. We investigated possible dependencies from taxel diameter and membrane thickness.
They can also very well mimic soft actuation systems, such as dielectric elastomers and shape memory alloys, for which the pressure below the actuator is constant, i.e. does not increase when being pressed by the finger. Constant pressure also allows to mimic latching properties i.e. the capability of the system to withstand the finger force with minimal energy consumption.
In addition, the kind and thickness of the membrane in contact with the finger is generally taken as a fixed parameter. Therefore it is generally unknown if varying thickness also varies perceived information.
In this preliminary study we investigate static tactile
thresholds of single air-driven taxels, when the dot diameter
is larger than Braille specifications, namely 4mm and
10mm. The first value corresponds to the inter-dot distance
known to be optimal for roughness perception [
TacTT’14, November 16–19, 2014, Dresden, Germany.
Copyright is held by the owner/author(s).
finger-like
The first experimental setup is aimed at measuring physical characteristics of taxels having variable diameter and contact membrane thickness. The second setup was aimed at estimating tactile detection thresholds when using the taxels of the first setup.
Ecoflex®50, Smooth-On, a Polydimethylsiloxane (PDMS) composite, was used as the membrane in contact with the finger. Elastosil 43, WACKER, an adhesive, was used to stick the PDMS membrane on top of the support, made of rapid prototyping material Verowhite 835. Rapid prototyping technology was used to fabricate taxels shaped as cuboids of variable size. Pneumatic energy was output from a piezolelectric micro-blower from Murata, able to reach pressures higher than 2kPa.
Ecoflex®50 is two-component based PDMS, composed of A and B; the ratio of mixing (A:B) is 1:1. The mixture was spin-cast by 300 rpm for 30 s to create homogeneous surface on TEFLON petri-dish. The spin-cast mixture was treated in vacuum to avoid air bubble when it solidifies before curing overnight at 80 ºC. When membrane was ready, it was adhered on the Elastosil coated substrate, which was printed in two configuration: a first setup involved small cubes 19 x 21 x 12 mm , a second setup involved bigger cubes 30 x 33 x 32 mm (see Figure 1b and 1c). Then it was cured overnight at 50 ºC to obtain fully cured membrane and stable properties. The 3D-printed component, with the membrane on top, was wedged on the nozzle of the piezoelectric micro-blower. In the second setup (big cubes), a further 3D-printed component surrounded the micro-blower to minimize accidental motion of the nozzle. deformation behaved as an inflating sphere of radius r given by:
The force-displacement setup is shown in Figure 1a. The tactile unit was inflated by the micro-pump. To produce the desired pressure, the micro-blower was voltage-driven at resonance by a sinusoidal wave, with fixed frequency 43 kHz and amplitude proportional to the desired pressure. We calibrated the voltage-pressure relationship to compensate for different micro-pumps, which exhibited repeteable but slighlty different characteristic functions. The micro-pump was supplied via DC power supply (TTi, Ex354 Tv, Triple Power Supply 300 W), and controlled by a MATLAB script (R2012b, 32 bit) through a National Instrument card (NI USB 6211, 16 inputs, 16 bit, 250 kS/s, Multifunction I/O), and a custom-built amplifier.
Preliminary measures were obtained with a DMA (Dynamic Mechanical Analyzer) to obtain force-displacement curves when force exceeded 1 mN (the DMA precision). Subsequent displacement measures were obtained with a moveable micro-scale station (THORLABS) and a force meter (Mecmesin, BFG 10 N). Specifically, a zero displacement is equivalent to a PDMS surface with 2 mN (the force-meter precision) of contact force with the force meter probe. The microblower was controlled to provide variable pressure values. For each pressure value, displacement was decreased up to zero with the micro-scale while measuring blocking forces.
Since no control on the actual pressure at the membrane was available (only the output from the micropump nozzle was known) we compensated possible air leakages out of the small/big cubes by normalizing pressure values fed to the psychophysical setup.
The true pressure values were estimated with the model depicted in Figure 2: we assumed the membrane (1) where r = radius of curvature; δ2mN = displacement at 1or 2 mN (for the DMA and the force meter respectively) of preload wrt the reference surface; w = width of the hole (the taxel diameter).
When the finger or force meter exerted a force (red arrow) on the "on" taxel, the elasticity of PDMS allows the probe to flatten the sphere up to a distance δ from the background.
The flattened area is the probe-taxel contact area is given by: ( ) [( (( ) ) ] (2) and the estimated pressure is obtained by dividing the measured force by A(δ).
In a first psychophysical experiment, thought as a pilot study, we used the small cubes, to be held between the thumb and the middle finger. The index finger had to touch the PDMS membrane. In a second experiment (see Figure 3), participants had their dominant hand resting on an Aluminium plate, where the setup with big cubes was mounted and fixed.
Nineteen participants (thirteen women) had a mean age of
33 years (range 26-41). For each participant, 5 min practice
preceded threshold estimation. Then, they underwent at
least 6 tactile detection threshold estimation blocks to
define for each participant their detection threshold (target
level of 75% correct) using the method of constant stimuli.
Each threshold estimation block comprised 30 trials in
which tactile stimuli with different pressure levels were
presented pseudorandomly either in a first or in a second
interval lasting 3.5 seconds. Each pressure level was
reached as described in the force-displacement setup.
First interval was defined by a single-beep sound. Second
interval was signaled by a double-beep sound. Participants
were instructed to touch the tactile unit with the index
finger of their dominant hand each time they heard the
sound and to verbally report whether they perceived a
displacement of the membrane in the first or the second
interval (two-interval forced choice detection task; 2IFC).
Control condition was a zero displacement interval.
To estimate the threshold for 75% correct detection, we
used the psignifit toolbox (psignifit.sourceforge.net) version
3.0 for Matlab, which implemented the
maximumlikelihood method described in [
When the estimated pressure of different setups resulted different, we normalized the thresholds for the relative efficiency. Unless otherwise stated all the reported statistics are one-tailed t-tests. measures ANOVA with membrane thickness (200, 600, 1000 and 2000 µm) as within subjects factor. The ANOVA showed a main effect of thickness [F(3,27) = 19.11; p =
As an example, Figure 4 shows preliminary forcedisplacement relationships, using a 10mm diameter taxel and variable membrane thicknesses. As expected, the blocking force increases as displacement decreases. Thinner membranes seem to transmit more force on equal displacements, or alternatively seem to be deflected more on equal forces. Qualitatively, thicker membranes also look stiffer. This is expected to have consequences at psychophysical level.
Figure 5(A), instead, shows the force-displacement curves obtained with two independent setups: a small and a big cube topped by a 200 µm membrane. The same was done with bigger cubes (not shown). The lowest forces (this time measured with the force meter) are limited to 2 mN. Figure 4 also shows that at the considered pressure (0.32 kPa), the displacement values are above known stroke thresholds (i.e. 200 µm for rigid mechanical couplings). The flattenedsphere model allows to derive Figure 5(B), which shows the force-area relationship.
Since a bias between the setups is apparent, the derivative of the force-area curves is taken between 0 and 10 mN to normalize subsequent psychophysical results. For example, the setup with the small cube, which appears more efficient, exhibits an overestimated pressure of a factor 1.72.
We first compared tactile detection thresholds using small cubes with different membrane thicknesses. 75% correct detection levels (kPa) were entered into a repeated A .000001]. Follow-up one-tailed t-tests showed that detection thresholds significantly increase with membrane thickness (threshold for 200 µm: 0.05 kPa ± 0.01 SE; for 600 µm: 0.07 kPa ± 0.01; for 1000 µm: 0.1 kPa ± 0.01; for 2000 µm: 0.46 kPa ± 0.09; all p values < .03; see Figure 6). Secondly, we analyzed results obtained with big cubes in which we manipulated target diameter and membrane thickness. We first compared tactile detection threshold performance when touching stimuli with different diameters after pooling over for membrane thicknesses. Threshold was significantly higher for 4 mm (5.36 kPa ± 0.85) compared to 10 mm target (0.10 kPa ± 0.01, p < .0000001). Then, we considered separately different membrane thicknesses. Detection thresholds are still significantly higher for 4 mm compared to 10 mm targets (for 200 µm thickness: 4.17 vs. 0.07 kPa; p = .002; for 1000 µm thickess: 6.16 vs. 0.12 kPa; p = .00006; see also Figure 7). On the contrary, detection threshold did not differ with different membrane thicknesses (threshold for 200 µm thickness: 1.89 kPa ± 0.86, threshold for 1000 µm thickness: 2.90 kPa ± 1, Mann-Whitney U = 40.5; p = .23). However, when considering separately 10 mm and 4 mm diameter stimuli a different perceptual profile emerges. While the difference between 200 and 1000 µm thickness for the 4 mm diameter is still not significant (threshold for 200 µm thickness: 4.17 kPa ± 1.16, threshold for 1000 µm thickness: 6.16 kPa ± 1.14; p = .13; see Figure 7b), when considering 10 mm diameter stimuli, detection threshold is significantly lower for 200 µm compared to 1000 µm thickness (threshold for 200 µm: 0.07 kPa ± 0.008, threshold for 1000 µm: 0.12 kPa ± 0.01; p = .047; see Figure 7a).
In our first experiment we checked for possible influences of variable membrane thickness on tactile detection thresholds of a taxel with fixed diameter. Subjects detected more easily thin membranes. This means that mechanical energy is transferred more efficiently from an air chamber to the finger mechanoreceptors with as little transducing material as possible.
Force-displacement relationship indicates that part of the energy is dissipated in the elastic deformation of thicker membranes. Increasing thickness by a factor of 10 increased thresholds by a factor of 5. Thin membranes seem therefore to be preferable when building such actuation systems. In our second experiment we checked for possible influences of taxel diameter on tactile detection thresholds, with varying thicknesses. When small-diameter taxels were "on" they were less distinguishable from their "off" state as opposed to large-diameter taxels. Reducing diameter by a factor of 2.5 increased thresholds by a factor of between 50 (using thicket membranes) and 60 (using thinner membranes). (10mm) (4mm)
A trend suggests that the perception of thicker membranes is more problematic with small diameters as well, even if more experiments are needed to confirm this aspect. One could argue that to better understand what is the main factor underlying perceptual thresholds a constant diameter/thickness ratio should be considered (i.e. the larger taxel is a "zoomed" version of the smaller one). Yet, had we done so, to keep such ratio constant we would have considered thicknesses of 500um and 2000um for the larger taxels, for which, approximately, the perception thresholds of the first might have been three times higher (while with 4mm taxels it increases by 50% only). The higher sensitivity with larger taxels can be possibly explained by a combination of two other factors: a larger area of mechanoreceptors (mainly SA1) and a different finger/taxel mechanical coupling.
In fact, qualitative observations from our subjects report that the hole of the 3d-printed taxel (i.e. the contours of the free-standing membrane) was considered a cue to distinguish "on" from "off" taxels, meaning that the underlying structure and geometry of the rigid support can be very important.
The differences between our setups, which we attempted to
compensate, may be due to air leakage, or to elastic
properties of the Ecoflex membranes [
This study is important for designing perceivable taxels with technologies even beyond air-driven setups, i.e. whenever a taxel is powered without latching mechanisms and whenever pressure under a soft actuator can be assumed constant. Our results can also be useful for the design of static soft taxels, which have very different (generally more stringent) requirements as opposed to vibrating taxels. We emphasize that static taxels better approximate paper-based tactile maps, currently the standard for visually impaired subjects.
Overall, larger taxels with thinner membranes looked perceivable with less energy. However, this implies decreasing the resolution of any pin-array display. Trading off energetic requirements with resolution seems therefore a necessary step in the design of graphical pin-array tactile displays.
This research is partially supported by the EU FP7 STREP project BLINDPAD (Personal Assistive Device for BLIND and visually impaired people), under grant 611621 and partially by the Fondazione Istituto Italiano di Tecnologia. The authors would like to thank Diego Torazza and Claudio Lorini for support about electromechanics of the setup.