Spatialized vibrotactile feedback contributes to
goal-directed movements in cluttered virtual
environments
Cephise Louison, Fabien R Ferlay, Daniel R. Mestre
To cite this version:
Cephise Louison, Fabien R Ferlay, Daniel R. Mestre. Spatialized vibrotactile feedback contributes
to goal-directed movements in cluttered virtual environments. 3DUI - IEEE Symposium on 3D User
Interfaces, Mar 2017, Los Angeles, United States. pp.99-102, 10.1109/3DUI.2017.7893324. hal01691483
HAL Id: hal-01691483
https://hal.science/hal-01691483
Submitted on 22 Dec 2022
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Spatialized vibrotactile feedback contributes to
goal-directed movements in cluttered virtual
environments
Cephise Louison, Fabien Ferlay, Daniel Mestre
To cite this version:
Cephise Louison, Fabien Ferlay, Daniel Mestre. Spatialized vibrotactile feedback contributes to goaldirected movements in cluttered virtual environments. 2017 IEEE Symposium on 3D User Interfaces (3DUI), Mar 2017, Los Angeles, United States. pp.99-102, 10.1109/3DUI.2017.7893324. hal03886902
HAL Id: hal-03886902
https://hal.archives-ouvertes.fr/hal-03886902
Submitted on 21 Dec 2022
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Distributed under a Creative Commons Attribution - NonCommercial| 4.0 International
License
Spatialized Vibrotactile Feedback Contributes to Goal-Directed
Movements in Cluttered Virtual Environments
Céphise LOUISON*
Fabien FERLAY*
CEA/IRFM, France
Aix-Marseille University/ISM, France
CEA/IRFM, France
In this context, the lack of kinesthetic (force) feedback for
example due to technical constraints, results in incomplete
sensorial feedback as soon as the subject interacts with virtual
objects. For example, collisions of the body with virtual objects
do not typically result in proprioceptive and tactile feedback;
nothing is actually there to stop the operator’s movement.
To address these limitations, vibrotactile (cutaneous)
stimulation appears as a good candidate to substitute for force
feedback [1], and make operators aware of impending or actual
collisions in highly constrained spaces. We thus hypothesized that
vibrotactile feedback, positioned on the operator’s skin, will
enhance spatial awareness of elements in the VE.
In this paper, we present preliminary results from an
experimental study, exploring the contribution of multi-localized
vibrators to visuo-proprioceptive consistency, during goaldirected movements in a cluttered virtual environment.
ABSTRACT
In virtual reality (VR), spatial awareness is a dominant research
topic. It plays an essential role in the assessment of human
operators’ behavior within virtual environments (VE), notably for
the evaluation of the feasibility of manual maintenance tasks in
cluttered industrial settings. In such contexts, it is decisive to
evaluate the spatial and temporal correspondence between the
operator’s movement kinematics and that of his/her virtual avatar
in the virtual environment. Often, in a cluttered VE, direct
kinesthetic (force) feedback is limited or absent. We tested
whether vibrotactile (cutaneous) feedback would increase visuoproprioceptive consistency, spatial awareness, and thus the
validity of VR studies, by augmenting the perception of the
operator’s contact(s) with virtual objects. We present preliminary
experimental results, obtained using a head-mounted display
(HMD) during a goal-directed task in a cluttered VE. Data suggest
that spatialized vibrotactile feedback contributes to visuoproprioceptive consistency.
2
RELATED WORK AND BACKGROUND
Vibration stimulation has long been used as an informational
medium in different applications. It can be divided into three main
application areas [2]: multimedia and entertainment, abstract
information and physical information. This last area directly
addresses spatial awareness.
In simulated tasks, the use of vibrotactile devices has gained a
lot of attention. For instance, Weber et al. [3] noted firstly that the
cutaneous information channel is less overloaded than the visual
and acoustic channels in complex work scenarios and secondly
that vibrotactile stimulation is unobtrusive and informs the
operator without disturbing other users. Finally, the stimulation is
localized on the operator’s skin. It triggers an intuitive form of
spatialized feedback. Moreover, vibrotactile interfaces are light
and wearable. Therefore, vibrotactile feedback is a good candidate
to substitute for kinesthetic stimulation sense and to make users
aware of impending or actual collisions in highly constrained
spaces.
Recent studies, have successfully investigated the effect of
vibrotactile stimulation while interacting with virtual objects.
Results suggest that cutaneous stimulation is effective for
improving the operator’s perception of impending or effective
collisions between the operator’s body and environmental
elements within a VE [1, 4–7].
A singular aspect of vibrotactile stimulation is the existence of
several perceptual illusions arising from the spatial configuration
of vibrators on the operator’s skin. One of these is called the
“funneling illusion” [8, 9]. It corresponds to “phantom” localized
sensations between two physical stimulations. Moreover, the
position of the phantom sensation can be controlled by
modulating the amplitude of the two stimulations/vibrations (the
distance between the two actuators/stimuli has to be in a range
between 40 and 80 mm) [10]. This modulation can be realized
using a linear, a logarithmic model [8, 11] or the “energy model”
Keywords: Visuo-proprioceptive consistency, Vibrotactile, Goaldirected movements, Cluttered environments.
Index Terms: I.3.7 [Computer Graphics]:Three-Dimensional
Graphics and Realism—Virtual reality; H.5.2 [User Interfaces]:
Haptic I/O
1
Daniel R. MESTRE†
Aix-Marseille University/ISM, CNRS,
France
INTRODUCTION
In virtual reality (VR), spatial awareness is an important concept.
It plays an essential role in the human operator’s behavioral
adaptation to the virtual environment (VE). The operator’s
behavior in future facilities has to be studied, to guarantee their
viability, in terms of accessibility and maintenance. The operator
must often work in a confined environment and has to pay
attention to the position of his/her whole body relative to the
position of various elements of the VE. In this context, visuoproprioceptive consistency is a key aspect for the validation of
behavioral data obtained in VE.
We use the term visuo-proprioceptive consistency to refer to the
spatio-temporal correspondence between the operator’s and
avatar’s kinematics. Spatio-temporal correspondence describes a
bi-directional link of interaction. The avatar must adopt the same
posture as the operator, being co-localized with the operator. In
return, the user must integrate the posture limitations imposed on
the avatar by the virtual environment (VE).
* {cephise.louison|fabien.ferlay}@cea.fr
† daniel.mestre@univ-amu.fr
1
between subjects. Different adjustments and configuration steps
were made at the beginning of each session. At the end of this
setup, the avatar was optimally co-localized with the subject’s
own body, reproducing the subject’s movements in real-time. The
subject could perceive the avatar’s body as his own body, from a
first-person subjective view (Figure 1, bottom right).
Subjects were instructed to reach, with their right hand, targets
within the VE, then to maintain the reached position for one
second, to validate it. They were asked to do it with the most fluid
and natural movement. No specific information was given
concerning the behavior to be adopted in case of contact with
VE’s objects. During an initial training phase, they could get
familiar with the task, with the virtual body and, depending on the
condition, with the potential vibrotactile feedback.
One trial consisted of the following sequence: the subject was
standing with arms along the body, and on the experimenter input,
a new target appeared. Following a beep, the subject had to reach
for the target. When a trial was validated, another beep was
played and the target disappeared. The subject had to take the
initial position before another trial could start. Participants were
allowed to have a short rest before starting the next trial, to avoid
physical exhaustion.
The experiment was interrupted by a short break between each
session. Each session consisted of twenty trials (five repetitions of
four targets). Within a given session, the targets’ succession order
was randomized across subjects. The subjects were split into four
groups. Subjects were randomly assigned to one group, each
group having a different succession order of the four experimental
conditions, to control for potential order or time effects (fatigue,
learning, etc.).
proposed by Israr and Poupyrev [12], which allows more precise
control of the intensity and the localization of the phantom
sensation.
3
MATERIAL AND METHODS
The following experimentation was designed to test the
hypothesis that vibrotactile feedback enhances spatial awareness
and visuo-proprioceptive consistency. Furthermore, using multilocalized vibrators should improve the precision of the operator’s
perception of impending or effective collisions with
environmental elements in highly constrained spaces.
3.1 Experimental Conditions
The independent variable was the type of vibrotactile feedback,
with 4 conditions. The information was either only visual or
augmented with vibrotactile stimulation. The vibrotactile
feedback was used to represent any contact between the right arm
and the environmental elements. There were three different
conditions of vibrotactile feedback: nearest, funneling or single
(See “Vibration Modes”). These three modes offer different levels
of collisions rendering accuracy. No visual metaphor (e.g. contact
arrows) was added to indicate collisions.
3.2 Procedure
Twenty-eight subjects (8 female) voluntarily took part in this
experiment. They were chosen from university students and staff,
with ages between 20 and 34 years (mean = 23.6; SD = 3.4).
Three volunteers were left-handed, but no differences were found,
concerning handedness. All participants were naïve as to the
purpose of the experiment.
Participants were first informed about the general procedure, and
were subsequently equipped with seventeen tracking markers, the
vibrotactile device, and the head-mounted display (HMD)
(Figure 1, left).
3.3 Virtual Environment
The experimental environment represents a part of an industrial
facility. The VE was composed of a platform to access to a
complex machine made of a support beam, pipes, etc. Four small
targets (blue spheres 7 cm in diameter) were positioned in this
environment, in the middle of the pipe forest. These targets were
positioned within arm’s reach (Figure 1, top right), distributed on
the top, bottom, right and left sides.
3.4 Apparatus
The experiment was conducted in a square area (3 x 3 meters) at
CRVM (www.crvm.eu). Subjects were equipped with an Oculus
Rift DK2 device. The tracking system (ArtTrack®), using optical
trackers with eight cameras, was used to monitor the subject’s
full-body posture and movements (Figure 1, left). The ARTHuman® software was used to reconstruct the subject’s skeleton
from the tracked markers’ positions. This body reconstruction was
used to calculate a real-time and co-localized virtual
representation of the subject’s body in the Unity application
(Figure 1, top right). The real-time VR system operated at 60 Hz.
Figure 1: Left. A subject in the experimental setup. Top right. Global
view of the experimental VE: avatar in the initial position and
targets. Bottom right. First person view inside the HMD.
A calibration step was required to build a biomechanical model
of the subject, using the ART-Human® software (Advanced
Realtime Tracking GmbH). A configuration file with the bone
lengths was generated and used by the XDE engine (See
“Apparatus”) to create a morphological avatar of the subject in the
VE. The virtual avatar was made of simple primitives (cylinders)
(Figure 1). It has been shown that abstract avatars don’t reduce the
illusion of virtual body ownership [13] and they are easier to use
for the physical engine computation.
The experiment was divided into four sessions, one for each
experimental condition, the succession order was counterbalanced
Figure 2: Vibrotactile suit with the vibrotactile controller and position
of the ten actuators along the arm .
2
individual trial. These positions (markers and avatar) had the same
spatial frame of reference and were measured at the center of
segments/bones. This integrated value gave us a direct indication
of the subject’s behavior with respect to physical objects of the
environment.
We next computed the mean contact time with environmental
objects during a trial. We measured contact time between the
moment where a contact point was detected along the upper limb
and the moment when no more contact points could be detected.
This indicator was meant to analyze how subjects dealt with the
occurrence of contacts, as a function of experimental conditions.
The analyses were conducted using repeated measures analysis
of variance (ANOVA). We then used post-hoc analyses with
Bonferroni adjustment and Planned Comparison tests.
Moreover, visual feedback (virtual environment and self-avatar)
could be augmented by vibrotactile feedback. The vibrotactile
device was based on an Arduino-like microcontroller
communicating via Bluetooth with the PC on which the
simulation was running. The controller addressed ten vibrators,
positioned on the subject’s right upper limb (Figure 2); with
variable amplitude using Pulse-width modulation (PWM). The
controller activated vibrators (DC motor with an eccentric mass,
from Parallax Inc. ref. 28822). Due to Bluetooth wireless
connection and the vibrator activation, there was a delay between
the instruction decision and when the vibration actually occurred.
However it was never noticed by the subjects.
The VR loop (motion capture to sensorial rendering) was
controlled by a PC. The experimental software was developed
with Unity3D. It allowed experimental control, data recording and
all scenario actions. The XDE physical engine developed by CEA
LIST (http://www-list.cea.fr/), integrated into Unity3D, allowed
realistic and real-time physical simulation and human control
from ART-Human data.
4
RESULTS
We first computed the co-localization Root Mean Square Error
(RMSE). The error of co-localization refers to the distance
between the subject’s positions and the corresponding avatar
positions. We computed RMS for the right (reaching) upper limb.
The ANOVA with the experimental conditions as the withinsubjects factor revealed a significant effect of conditions (F (3,
81) = 7.5127; p =.00017). Subsequent post-hoc analyses with
Bonferroni adjustment showed that RMS values in the three
vibrotactile conditions were significantly lower than in the “visual
only” condition (p <.005), and that these three conditions were not
statistically different from one another (see Figure 3).
This result suggests that visuo-proprioceptive consistency was
improved by the vibration feedback. Spatialized feedback
(funneling or nearest conditions) was not shown to have a
significant effect compared to a single vibrator (located on the
hand, whatever the contact position).
3.5 Vibration Modes
The XDE physical engine allowed the computation of contact
points between rigid bodies. The position of the contact, the
normal and the force applied was computed for each contact
point. The physical engine did not allow interpenetration between
rigid bodies (for example, between the avatar and the VE): If a
subject tried to walk through a wall in the VE, the avatar was
stopped by the (virtual) wall. In that case, the distance between
the avatar and the subject’s body was no longer null. It is a case of
visuo-proprioceptive inconsistency. The contact information
(position and force) was used to compute which vibrator should
be activated to represent the contact on the subject’s arm and the
amplitude of the vibration.
In this experiment, three different conditions (modes) of
vibration were used. Each mode had a specific algorithm to select
the amplitude of vibrators during a contact.
Single: In this mode, only one actuator (actuator ‘0’ on
Figure 2, right) was used to represent the existence of a contact
between the avatar and the VE, regardless of the position of the
contact on the avatar’s arm. The vibration amplitude was mapped
according to the contact force.
Nearest: In this mode we selected the nearest vibrator to the
contact position on the avatar’s arm. The amplitude was related to
the force of contact. Several contacts could be represented at the
same time by different actuators.
Funneling: We decided to use this effect to create the sensation
of a contact position that would be closest to the “actual” contact
point. If possible, a pair of vibrators such that the contact point
was between the positions of the two vibrators was chosen. We
applied the energy model [12] to calculate the amplitude of both
vibrators, according to the contact position and force. If no pair
could be found, we applied the “Nearest” algorithm.
Figure 3: Average values (with standard error) of RMS error, as a
function of experimental conditions.
The mean contact time during a single trial was analyzed. A
significant effect of experimental conditions was found (F (3, 81)
= 7.2249; p < .001). We observed a trend of decreasing contact
time with increasingly localized vibration conditions (Figure 4).
Post-hoc analysis (Bonferroni) shows that the “single” condition
is not statistically different from the “visual” condition (p=.49).
However, both “funneling” and “nearest” conditions gave
significantly lower values of mean contact time as compared to
the “visual” condition (p <.005). The “funneling” and “nearest”
conditions do not differ statistically (p>.50). Finally, Planned
Comparisons tests reveals that “Single” is statistically different
from the group {“nearest”, “funneling”} (p <.05).
This pattern of results suggests that the “single” non-localized
vibrator informs the subjects about the occurrence of a contact.
However, subjects still have to “search” for the contact location,
as in the “visual” condition. As a consequence, they need a little
more time to react. In that case, localized vibration (funneling or
nearest) directly gives an indication about the contact location,
thus reducing contact time.
3.6 Data Analysis
During the experiment, several behavioral indicators and events
were recorded and analyzed. To reduce noise in the data, we
discarded initial failed attempts to reach the target and we kept
only the final (and successful) reaching movement for each
individual trial.
The main behavioral metric was related to the concept of visuoproprioceptive consistency. To analyze that, we computed (for the
right upper limb segments) the Root Mean Square (RMS) of the
distance between the real subject’s 3D positions (obtained from
the tracking system) and the avatar’s positions throughout each
3
“funneling illusion”. Nevertheless, preliminary results from the
experimental study failed to show that this phenomenon had a
significant effect. Future analysis and work will pursue that line of
investigation.
ACKNOWLEDGMENTS
The authors wish to thank the CRVM team for decisive assistance
in the experimental process. We also thank the CEA
DEN/MAR/DTEC/SDTC/LSTD and CEA DRT/LIST/DIASI/LSI
teams for their involvement in the project. Céphise Louison is
supported by a doctoral grant form CEA.
Figure 4: Average values (with standard error) of the mean contact
time in a single trial, as a function of experimental conditions.
5
REFERENCES
[1]
DISCUSSION
These preliminary results show evidence that individuals make
fewer co-localization errors and react faster to the occurrence of a
collision when presented with localized vibrotactile feedback.
First of all, subjects showed a lower co-localization RMSE in
the presence of vibrotactile stimulation. This positive effect can be
explained by the fact that vibrotactile feedback increases spatial
awareness and visuo-proprioceptive consistency, thus helping the
subjects to maintain the spatio-temporal co-localization between
their real arm and the virtual arm.
This result also suggests that, globally, there is no major
statistical difference between the different conditions of
vibrotactile feedback. This suggests that, before localized
information, vibrators deliver a “warning” signal. However, we
also found evidence that, with the multi-localized vibrotactile
feedback, subjects made shorter contact times, as compared to a
single actuator or without vibrotactile feedback. It appears that the
reaction to a contact is faster with localized information. It can be
explained by the fact that the multi-localized vibrotactile feedback
allows faster localization of a contact point without visual search
and a faster reaction to it. This is supported by subjective data
(post-hoc open interview). Subjects generally reported that the
extra information given by the multi-localized vibrotactile device
was relevant to find the position of contact points.
Finally, no statistical difference was observed between the two
multi-localized vibrotactile feedback conditions (funneling and
nearest). The fact that there is no difference between these two
algorithms may indicate that an increase in the resolution of the
spatialized feedback display is not necessary. It could also be due
to the limits of the vibrotactile device used in this experiment or to
an insufficient configuration of the funneling algorithm. This
question requires further investigation and additional data
processing of the kinematics of reaching movements.
6
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
CONCLUSIONS AND FUTURE WORK
[10]
Our approach mixed a dynamic and co-localized visual
representation (avatar) of the subject’s body and a multi-localized
vibrotactile feedback as a substitute for force feedback. These
preliminary results suggest that vibrotactile feedback enhances
visuo-proprioceptive consistency. Moreover, a localized
vibrotactile feedback allows faster localization of contact points
by reducing the need of visual search, also favoring visuoproprioceptive consistency.
It seems that using vibrotactile feedback to enhance the
perception of collisions with virtual elements is a very promising
approach to the improvement of spatial awareness of the operator
in a cluttered environment. As such, it might be an important
factor for the ecological validity of virtual experiments.
To finish, a large amount of information can be displayed via
tactile means, making use of perceptual phenomena such as the
[11]
[12]
[13]
4
R. W. Lindeman, R. Page, Y. Yanagida, and J. L. Sibert,
“Towards full-body haptic feedback: the design and
deployment of a spatialized vibrotactile feedback system”,
in Proceedings of the ACM symposium on Virtual reality
software and technology, 2004, pp. 146–149.
S. Choi and K. J. Kuchenbecker, “Vibrotactile Display:
Perception, Technology, and Applications”, Proc. IEEE,
vol. 101, no. 9, pp. 2093–2104, Sep. 2013.
B. Weber, S. Schätzle, T. Hulin, C. Preusche, and B. Deml,
“Evaluation of a vibrotactile feedback device for spatial
guidance”, in World Haptics Conference (WHC), 2011
IEEE, 2011, pp. 349–354.
A. Bloomfield and N. I. Badler, “Virtual training via
vibrotactile arrays”, Presence Teleoperators Virtual
Environ., vol. 17, no. 2, pp. 103–120, 2008.
C. Louison, F. Ferlay, D. Keller, and D. Mestre,
“Vibrotactile Feedback for Collision Awareness”, in
Proceedings of the 2015 British HCI Conference, New
York, NY, USA, 2015, pp. 277–278.
D. R. Mestre, C. Louison, and F. Ferlay, “The Contribution
of a Virtual Self and Vibrotactile Feedback to Walking
Through Virtual Apertures”, in Human-Computer
Interaction. Interaction Platforms and Techniques, M.
Kurosu, Ed. Springer International Publishing, 2016, pp.
222–232.
J. Ryu and G. J. Kim, “Using a vibro-tactile display for
enhanced collision perception and presence”, in
Proceedings of the ACM symposium on Virtual reality
software and technology, 2004, pp. 89–96.
D. S. Alles, “Information Transmission by Phantom
Sensations”, IEEE Trans. Man-Mach. Syst., vol. 11, no. 1,
pp. 85–91, Mar. 1970.
G. v Békésy, “Funneling in the Nervous System and its
Role in Loudness and Sensation Intensity on the Skin”, J.
Acoust. Soc. Am., vol. 30, no. 5, pp. 399–412, May 1958.
J. Cha, L. Rahal, and A. El Saddik, “A pilot study on
simulating continuous sensation with two vibrating
motors”, in Haptic Audio visual Environments and Games,
2008. HAVE 2008. IEEE International Workshop on, 2008,
pp. 143–147.
J. Seo and S. Choi, “Initial study for creating linearly
moving vibrotactile sensation on mobile device”, in
Haptics Symposium, 2010 IEEE, 2010, pp. 67–70.
A. Israr and I. Poupyrev, “Tactile brush: drawing on skin
with a tactile grid display”, in Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems,
2011, pp. 2019–2028.
J. L. Lugrin, J. Latt, and M. E. Latoschik, “Avatar
anthropomorphism and illusion of body ownership in VR”,
in 2015 IEEE Virtual Reality (VR), 2015, pp. 229–230.