ArticlePDF Available

Inertial-Force Feedback for the Treadport Locomotion Interface

Authors:

Abstract and Figures

The inertial force due to the acceleration of a locomotion interface is identified as a difference between virtual and real-world locomotion. To counter the inertial force, inertial-force feedback was implemented for the Treadport, a locomotion interface. A force controller was designed for a mechanical tether to apply the feedback force to the user. For the case of the user accelerating forward from rest, psychophysical experiments showed that subjects preferred inertial-force feedback to a spring-feedback force proportional to position or to position control, where the force feedback maintained a force of zero on the subject. 1
Content may be subject to copyright.
Robert R. Christensen
John M. Hollerbach
Yangming Xu
Sanford G. Meek
Mechanical Engineering Department
Computer Science Department
University of Utah
Salt Lake City, UT 84112
jmh@cs.utah.edu
www.cs.utah.edu/jmh/
locomotion.html
Inertial-Force Feedback for the
Treadport Locomotion Interface
Abstract
The inertial force due to the acceleration of a locomotion interface is identified as a
difference between virtual and real-world locomotion. To counter the inertial force,
inertial-force feedback was implemented for the Treadport, a locomotion interface. A
force controller was designed for a mechanical tether to apply the feedback force to
the user. For the case of the user accelerating forward from rest, psychophysical ex-
periments showed that subjects preferred inertial-force feedback to a spring-feedback
force proportional to position or to position control, where the force feedback main-
tained a force of zero on the subject.
1 Introduction
Conventional virtual reality solutions to user movement through virtual
space generally involve position tracking with a variety of sensors that include
mechanical, optical, acoustic, inertial, and magnetic sensors (Durlach & Mavor,
1994). In the simplest scenario, a joystick controls the movement through ter-
rain or large structures of either a user’s avatar or of some spacecraft, airplane,
or vehicle. Some of the realism is lost in this artificial approach as movement is
subjugated to manipulation or graphics realism (Witmer & Kline, 1998).
More-complex scenarios involve a user walking in a confined space and the
tracking of the user’s position. The Walkthrough Project tracks user position by
cameras on a head-mounted visual display viewing light-emitting diodes em-
bedded in the ceiling (Ward et al., 1992). The CAVE employs surround screens
and magnetic sensors for user tracking (Cruz-Neira et al., 1993). Slater et al.
(1995) investigated the use of walking in place for navigation in virtual envi-
ronments. Peterson et al. (1998) employ a shallow disc that senses the direction
of a user stepping from the center.
Another approach is whole-body motion platforms, in which a user is pas-
sively seated and moved through small excursions. Examples are flight simula-
tors and vehicle simulators (Cremer & Kearney, 1995; Durlach & Mavor,
1994; Salcudean et al., 1994) that offer kinesthetic cues appropriate for these
simulators. Yet, because the user is seated passively, the user does not expend
energy in moving through virtual worlds.
Locomotion interfaces are energy-extractive interfaces to virtual environ-
ments and fill needs that are not met by conventional position-tracking ap-
proaches or whole-body motion platforms:
They allow wide ranging in the virtual world in a limited space in the real
world.
Presence, Vol. 9, No. 1, February 2000, 1–14
r
2000 by the Massachusetts Institute of Technology
Christensen et al. 1
They increase the user’s involvement in the virtual
world.
They affect the decision-making process through
fatigue and reflection of the virtual world’s surface.
They free the user to work with other VR appli-
ances, such as position trackers, head-mounted dis-
plays, and so forth.
Possible applications for locomotion interfaces include:
Virtual design. A designer can walk through archi-
tectural models, virtual assembly lines, and ship inte-
riors.
Education. Students can maneuver through histori-
cal sites, the moon surface, and insect colonies.
Training. Individual training for the military, law
enforcement, and fire fighting with virtual reality
systems are more realistic when users are forced to
expend energy to achieve goals.
Adversarial gaming. Individuals can compete in
organized arcades of locomotion interfaces or
against computer-generated forces.
Health rehabilitation. The systems can provide
customized exercise and monitoring environments
for those afflicted with mobility problems.
Exercise and recreation. City-bound hikers may
safely enjoy back-country rigors without leaving the
confines of their living rooms.
Psychophysical research. Sensorimotor integration
and navigation issues can be studied through ma-
nipulation of the virtual environment, such as the
gain between optical flow and locomotion (Rieser et
al., 1995).
Several different types of locomotion interfaces have
been previously developed.
1. Pedaling devices. The OSIRIS (Lorenzo et al.,
1995) is a simulator for night-vision battle and
employs a stair-stepper device. The Uniport (Sar-
cos Research, Inc.) is a unicycle with electrically
actuated pedals that provide resistance based upon
terrain slope and locomotion speed. Steering is
achieved by twisting at the waist. Distler and
Bulthoff (1996) employ a reclining bicycle with
handlegrips for steering in a CAVE environment.
Ensor and Carraro (1997) employ a bicycle with
programmable pedal resistance. Brogan et al.
(1997) employ an instrumented bicycle on a plat-
form that can pitch 12 deg. to simulate the effect
of going up or down hills.
2. Programmable foot platforms. Uneven terrain
can be presented by active three-degree-of-free-
dom (DOF) platforms for each foot. The Sarcos
Biport employs three-axis hydraulic serial-drive
platforms for each foot. (Force sensing and control
are required to unburden the foot during the
swing phase.) Roston and Peurach (1997) report
on an electrically driven three-DOF, two-foot plat-
form system.
3. Sliding surfaces. Iwata has created several versions
of devices that involve a stationary human sliding
his or her feet over the ‘‘ground’’ during walking.
In one version, a user wears rollerskates and is rig-
idly attached to a frame by a belt around the waist
(Iwata, 1992). Foot forces for climbing or de-
scending are created by a single string belt for each
foot that is routed through pulleys to the bottom
of each foot and actuated by a DC motor. More
recently, Iwata and Fujii (1996) replace the roller-
skates by shoes with low-friction films on the soles.
A brake pad on the toe allows completion of the
swing phase. The user is loosely constrained by a
hoop frame.
4. Treadmills. Brooks et al. (1992) employed a pas-
sive or nonmotorized treadmill and instrumented
bicycle handlebars for steering. Witmer and Kline
(1998) also employed a passive treadmill. The AT-
LAS system (Noma & Miyasato, 1998) employs an
active treadmill on a spherical joint; body motion is
measured with a string potentiometer and by video
tracking markers on the feet. The Sarcos Treadport
employs a large, tilting, active treadmill and a hori-
zontal mechanical tether to measure body move-
ment and apply forces to the user for various pur-
poses. Darken and Cockayne (1997) report on the
use of the Omni-Directional Treadmill, a two-di-
mensional active treadmill that employs two or-
thogonal roller belts. A mechanical tracking arm
on an overhead boom measures body position and
2 PRESENCE: VOLUME 9, NUMBER 1
applies bias forces to center a user. Iwata (1999)
drives a torus-shaped surface consisting of ten belts
to generate an infinite surface.
1.1 Treadmill Locomotion
This paper is concerned with treadmill-style loco-
motion interfaces—and the Sarcos Treadport in particu-
lar. Treadmill-style locomotion interfaces support many
human mobility activities:
walking, running, shuffling, sprinting, and skipping
moving backwards or forwards
turning left or right
sidling
kneeling and crawling
A key issue is whether the treadmill is motorized. In
nonmotorized or passive treadmills, because there is no
motor, the belt motion is due to the forces applied to
the belt by the user. The sensed belt speed is then used
to update the user’s position in the virtual environment.
Unrestrained human mobility that involves walking or
running results in highly variable energy utilization.
Nonmotorized treadmills provide a fairly crude approxi-
mation to natural walking and running, however, with
distortions from unnatural gait. A major problem with
passive treadmills is that the belt’s velocity is constantly
changing. As the user’s foot makes contact with the belt,
it first slows the belt and then speeds it up as the foot
pushes off. The belt then starts to slow down again due
to friction, until the user’s other foot strikes and the
cycle repeats.
In active treadmills, measurements of user position are
accomplished by optical tracking (the ATLAS system) or
by a mechanical linkage or tether (Sarcos Treadport or
Omni-Directional Treadmill). These measurements are
employed to control treadmill speed and the movement
of a user’s avatar through the virtual environment. Turn-
ing in the ATLAS system is achieved by swiveling the
treadmill about a vertical axis in the direction that the
user is stepping. A two-dimensional treadmill allows a
natural change of direction, although stability issues in
walking on the roller surface of the Omni-Directional
Treadmill have been reported (Darken & Cockayne,
1997). On the Sarcos Treadport, turning in the virtual
world is accomplished by rate control based on the
amount of sideways excursion when moving or on the
amount of left-right body turning when stationary.
Aside from the naturalness of turning, another chal-
lenge for treadmill-style locomotion interfaces is to accu-
rately simulate straight-ahead motion on the ground.
Interestingly, ground locomotion and treadmill locomo-
tion differ in a critical aspect. When the treadmill belt
moves at a constant speed, running, walking, and crawl-
ing on the treadmill are the same as on the ground.
However, when the treadmill belt accelerates, it applies a
force to the user that would not be felt during natural
locomotion on the ground. This force is an inertial force
due to the acceleration of the belt.
The effect of the inertial force can be quite large as
seen from the results of Frishberg (1983). Frishberg
filmed five sprinters during a 91.44 m (100 yd.) sprint
and recorded their velocity profiles. He then pro-
grammed a treadmill to match these profiles, and each
subject repeated the sprint on the treadmill. The oxygen
debt for treadmill sprinting was 36% less than sprinting
on the ground. The large difference in energy expendi-
tures in Frishberg’s results was more than could be at-
tributed to the lack of air resistance when running on
the treadmill. The inertial force due to the accelerating
treadmill allowed the subjects to use less energy.
Moghaddam and Buehler (1993) proposed to coun-
teract this inertial force by varying the speed of the belt.
Their application is a treadmill for research on legged
robot locomotion. Because it is not possible to achieve
exact compensation for the inertial force by varying the
belt speed alone, they proposed an optimal scheme to
control the belt’s velocity that trades off differences in
the contact forces and the velocity profiles between loco-
motion on the treadmill and on the ground. Simplified
models for the user and the system were employed: the
human was modeled as a wheeled cart with a velocity-
feedback controller. Step inputs to the cart were studied
to simulate acceleration to a maximum velocity.
Flight simulators employ another method to simulate
the forces felt during acceleration. The entire simulator
tilts to change the angle of the force felt by the pilot, so
that a fraction of the gravitational force is used to simu-
Christensen et al. 3
late limited acceleration. The advantage of this method
is that the force is applied to the pilot’s whole body.
While this method works well for flight simulators, when
applied to a locomotion interface it has the disadvantage
of confounding slope and acceleration. A tilting surface
would provide kinesthetic cues to the user from the
change in the angle of the ankle. There is also a practical
implementation issue of changing the slope of the mo-
tion platform quickly enough because of its mass. A
practical method that maintains the independence of
slope and inertial force would be ideal (Tristano et al.,
1999).
In this paper, we use the active mechanical tether on
the Sarcos Treadport to achieve exact compensation of
the inertial force, and we show through user studies the
naturalness of this inertial-force feedback.
1.2 The Sarcos Treadport
The Sarcos Treadport (Figure 1) comprises a large
commercial treadmill (4 8 ft), a six-DOF mechanical
tether, and a CAVE-like visual display. The mechanical
tether (Figure 2) is a key feature of this locomotion in-
terface. Kinematically, the mechanical tether comprises a
Hookean joint at the base, a prismatic joint (which is the
primary joint adjusting for user travel), and a spherical
joint attached to a climber’s harness (Figure 3) worn by
the user. The prismatic joint is actuated through a tim-
ing belt drive, which allows it to exert an axial horizontal
force on the user. The harness worn by the user consists
of a solid piece of material along the user’s back, adjust-
able leg and shoulder loops, an adjustable belt, an ad-
justable chest strap, and adjustable height straps between
the chest strap and belt. The tether is attached to a small
plate in the back of the harness located in the small of
Figure 1. The Treadport.
Figure 2. The Treadport’s mechanical tether.
Figure 3. The Treadport’s harness.
4 PRESENCE: VOLUME 9, NUMBER 1
the user’s back. Since its location is between the belt and
the chest strap, it isn’t possible to get the tether attach-
ment coupled tightly to the body.
The harness is attached to the ceiling by a mechanism
similar to an automobile seat belt. The safety line pre-
vents sudden vertical accelerations. If the user were to
trip and fall, the safety line would catch him. As long as
the user doesn’t move too quickly, the safety line would
allow theuser to bend down,squat, or crawlon his hands
and knees.The tether is mountedbehind the userand has a
breakaway pin in case theuser falls off the back of thetread-
mill. Inthis case, the treadmill would automatically stopthe
belt andturn off the tether. The treadmill also automatically
stops ifthe user moves tooclose to thefront of thetreadmill.
The mechanical tether serves two general purposes.
First, it acts as a position tracker to control the treadmill
according to the user’s motion and to change the
graphical display. More uniquely, it exerts force on the
user which can be employed for several purposes:
1. providing a centering force as a kinesthetic cue to
help the user stay centered on the treadmill;
2. enforcing unilateral constraints by imposing a pen-
alty force when a user attempts to walk through a
virtual wall; and
3. pulling or pushing on the user to simulate the ef-
fects of gravity (Tristano et al., 1999) or uneven
terrain.
Section 2 identifies the inertial force due to the accelera-
tion of the treadmill belt as a difference between virtual
and real world locomotion. Section 3 details the design,
simulation, and implementation of the tether controller
that provide force feedback to cancel the inertial force.
Section 4 presents psychophysical experiments that were
conducted to determine whether subjects preferred iner-
tial-force feedback to other types of control. Conclu-
sions are drawn in Section 5. Portions of this work have
been previously reported (Christensen et al., 1998).
2 Virtual versus Real-World Locomotion
When an automobile accelerates forward, the pas-
sengers ‘‘feel’’ a force pushing them back into their
seats. This imaginary force is known as an inertial force
and is due to the acceleration of the vehicle, which is the
user’s frame of reference. To an observer on the ground,
it is apparent that the passenger is accelerating forward
and the car seat is applying a force to cause this accelera-
tion. The inertial force creates a difference between vir-
tual and real-world locomotion. Consider a very large
treadmill moving at a constant velocity with no external
visual cues. A person on the treadmill would have no
way to know the treadmill’s speed or direction with re-
spect to the ground—or even if it is moving at all. No
matter what direction or speed the person would move
relative to the treadmill, it would feel the same as on the
ground. As long as the treadmill moves at a constant
velocity, it is an inertial reference frame, and Newton’s
laws of motion apply.
If the treadmill belt were to suddenly accelerate, the
user would feel as though he were pushed in the oppo-
site direction of the acceleration. This inertial force is
due to the acceleration of the reference frame. The
treadmill belt is now a non-inertial reference frame be-
cause it is accelerating. For a non-inertial reference
frame, if the inertial force is considered as an additional
external force acting on the system, then it is possible to
apply all the results and principles derivable from New-
ton’s laws of motion relative to an inertial frame (Green-
wood, 1988). The inertial force is in the opposite direc-
tion of the acceleration of the belt, and its magnitude is
equal to the mass of the user multiplied by the belt’s ac-
celeration.
This large treadmill represents virtual locomotion. In
order to analyze virtual locomotion in the same manner
as real-world locomotion, the inertial force due to the
acceleration of the belt must be considered. To illustrate
this concept, consider simplifying the dynamics of the
human by replacing them with those of a motorized,
wheeled cart (Moghaddam & Buehler, 1993). The cart
on the ground, or the real world, has the following dy-
namics:
f
g
f
f
ma
g
, (1)
where f
g
is the contact force between the cart and the
ground, f
f
represents any aerodynamic friction forces, m
is the mass of the cart, and a
g
is the acceleration of the
cart with respect to the ground.
Christensen et al. 5
The cart’s motion is analyzed with respect to a coordi-
nate system attached to the ground. The cart on the
treadmill, or the virtual world, has these dynamics:
f
c
f
f
ma
c/b
ma
b/g
, (2)
where f
c
is the contact force between the cart and the
belt, a
c/b
is the acceleration of the cart with respect to the
belt, a
b/g
is the acceleration of the belt with respect to
the ground, and ma
b/g
represents the inertial force due
to the acceleration of the belt.
The cart’s motion is analyzed with respect to a coordi-
nate system attached to the treadmill belt. By consider-
ing the inertial force as an additional external force, the
equation can be rewritten as
f
c
f
f
ma
b/g
ma
c/b
(3)
For virtual and real-world locomotion to be the same, a
given input should produce the same output in either
system. That is, if the cart produces the same force pro-
file on the ground and on the treadmill, then its accel-
eration and velocity profiles should be the same in the
respective coordinate systems. This is analogous to a hu-
man sprinting from a standstill with maximum effort.
Assuming that the difference in friction forces due to air
resistance and in the different types of surfaces is negli-
gible, the sprinter should have the same acceleration and
velocity profiles in the virtual and real worlds. By com-
paring Equation (1) and (3), it can be seen that—if the
contact force, the frictional force, and the acceleration
with respect to the coordinate system are all the same—
the equations are not equal unless ma
b/g
is zero. This
term is due to the acceleration of the treadmill belt and
is not zero because, if the treadmill belt does not acceler-
ate, the user will run off the front. The inertial force di-
rection is the same as the acceleration direction of the
runner, so it is providing a force that helps the runner
accelerate. This inertial force helps explain Frishberg’s
results.
Moghaddam and Buehler (1993) developed a con-
troller for the treadmill belt’s velocity to minimize the
effect of the belt’s acceleration. The effect cannot be
eliminated through the control of the treadmill belt, and
it is more pronounced with smaller treadmills and larger
accelerations by the user.
The Treadport provides a solution to this problem
with force feedback from the mechanical tether. Equa-
tion (4) shows f
applied
as the feedback force.
f
c
f
f
ma
b/g
f
applied
ma
c/b
. (4)
If the magnitude of f
applied
is equal to the magnitude of
the inertial force but in the opposite direction, then
those two forces cancel, leaving
f
c
f
f
ma
c/b
, (5)
which is identical to the dynamics of the cart on the
ground. The desired force to be applied by the tether in
order to cancel the inertial force due to the acceleration
of the belt is
f
applied
ma
b/g
. (6)
The application of this force is called inertial-force feed-
back. At this point, frictional forces are dropped from
the analysis because this study is concerned with simulat-
ing the inertial force and not air resistance. The differ-
ences between treadmill and overground running re-
corded by Frishberg (1983) were much larger than
could be attributed to air resistance. Figure 4 shows a
runner accelerating on the Treadport. The external
forces acting on the runner are the interaction force f
c
,
the inertial force f
b/g
⫽⫺ma
b/g
, and the applied tether
force f
applied
. Inertial-force feedback will use the tether
force to cancel the inertial force, leaving only the inter-
action force which is the same as it would be on the
ground.
Figure 4. External forces during forward acceleration.
6 PRESENCE: VOLUME 9, NUMBER 1
3 Control of the Treadport
The Treadport controller, as shown in Figure 5,
consists of three parts: the velocity controller, the tether
controller, and the model of the human subject. Con-
trollers were designed for the force-feedback tether and
the velocity of the treadmill belt. The purpose of the
tether force controller is to cancel the inertial force due
to the belt’s acceleration. At the same time, the belt’s
velocity controller is needed to keep the user safe and to
decrease the inertial force. As part of the process of de-
signing the controllers, a model of human running was
required. The three parts of the Treadport control
scheme are discussed in the following sections.
3.1 Running Model
Hill (1927) derived an equation that describes
rather accurately the sprint velocity curve of a sprinter
(Henry & Trafton, 1951). This velocity equation is
v v
m
(1 e
k
v
t
), (7)
where v is the velocity of the runner, v
m
is the maximum
velocity the runner is approaching, and k
v
is a reaction
velocity constant. The interaction force was assumed to
be constant instead of cyclical with each step. This
model agrees with the model of a wheeled cart with a
feedback velocity controller assumed by Moghaddam
and Buehler (1993). The human running model is
shown in the upper left block of Figure 5.
To determine the validity of this model, the sprint
starts of five subjects were filmed. A film digitizer was
used to determine the position and time data of retrore-
flective markers attached to the subjects. Equation (7)
was integrated to give the position of the runner:
x v
m
1
t
e
kt
k
1
k
2
. (8)
Equation (8) was fit to the data to determine the accel-
eration parameters to use in the simulations. Figure 6
shows the data and fitted equation for a typical sprint
start.
3.2 Tether Control
The tether is chain-driven by a DC motor. A load
cell is at the end of the tether where it attaches to the
Figure 5. Treadport controller.
Christensen et al. 7
harness. The tether is capable of producing a sustained
force of nearly 190 N. The open-loop frequency re-
sponse of the tether was found using Siglab (DSP Tech-
nology, Inc.). The tether was analyzed by swept-sine
analysis of the input voltage to the force sensor output
while its position was constrained. Data were collected
for several different input amplitudes. A second-order
model was fit to the data and used in the simulations.
A force controller similar to that used by An et al.
(1988) was used for the tether. A simplified block dia-
gram of the system is shown in Figure 5, where v
bd
is the
desired treadmill belt velocity, v
b
is the actual velocity of
the belt, a
b
is the estimated acceleration of the belt from
the Kalman filter, F
d
is the desired tether force, and F
m
is
the measured tether force. The control algorithm was
implemented digitally at 400 Hz, and the servo rate was
verified with a digital oscilloscope.
The force servo loop was tested while its position was
constrained. Figure 7 shows the force step response of
the tether while its position was constrained. Steps are a
qualitative test for the stability of the tether, but a more
realistic performance test is sine-wave command follow-
ing (An et al., 1988). The next test was to follow a sinu-
soidal input while the position was constrained. The re-
sponses to the sinusoidal inputs are compared in Figure
8. The tether tracks the sine wave up to 10 Hz, although
there is considerable lag. The bandwidth measured in
this manner is at least 10 Hz.
When the force controller was implemented on sub-
jects, the harness attachment was loose and caused back-
Figure 6. Position of an accelerating runner on the ground.
Figure 7. Experimental force controller step response.
Figure 8. Experimental force controller sine response.
8 PRESENCE: VOLUME 9, NUMBER 1
lash, the position was no longer constrained, and the
user created disturbance forces. All of these factors im-
paired the performance of the tether force control. The
actual inertial-force feedback had much more tracking
error than the simulations, but its overall behavior was
still acceptable.
3.3 Treadmill Control
The commercial treadmill includes a built-in veloc-
ity controller. Its frequency response was characterized
with a Hewlett Packard 3562A Dynamic Signal Ana-
lyzer. Swept-sine analysis was done at several different
input amplitudes. More tests of the system were com-
pleted using Siglab. Burst random noise was the stimulus
to the system for different input amplitudes and band-
widths. It was found that a fourth-order model fit the
data, and this model was subsequently used in the simu-
lations and in the acceleration filter.
Figure 5 shows the velocity control scheme for the
Treadport. The desired velocity command for the tread-
mill belt is derived from a proportional-integral (PI)
controller based on the user’s distance from an equilib-
rium point, as shown in Equation (9):
v
bd
k
p
x k
i
xdt (9)
the user’s distance from the equilibrium point is x, k
p
is
the proportional constant, and k
i
is the integral constant.
The proportional term increases the speed of the tread-
mill as the user moves farther away from the equilibrium
point. The integral term increases the speed of the tread-
mill the longer the user is away from the equilibrium
point.
The integral control is important for the safety of us-
ers because it recenters users whenever they are moving
at constant velocity. Without the integral control, a us-
er’s position would just be proportional to the velocity.
Consider a subject who accelerates to 0.5* v
m
, where v
m
is a subject’s maximum speed. Suppose that, after run-
ning at this speed for a while, the subject suddenly accel-
erates to v
m
. If the velocity control did not have an inte-
gral term, the subject would already be close to the front
of the treadmill. If there were enough delay in the tread-
mill’s response, the subject could step off the front of
the treadmill and be injured. Integral control would
have recentered the subject after the initial acceleration,
and the Treadport would have had time to respond to
the second acceleration.
If the treadmill were controlled so that the user was
perfectly centered at all times, a fairly high inertial force
would have to be applied by the tether. For example, a
typical peak acceleration for a runner is approximately 10
m/s
2
. For a person weighing 75 kg, the tether would
have to apply a force of close to 750 N. If, on the other
hand, a user were allowed to move forward on the tread-
mill, the user would experience some actual acceleration
that would reduce the force requirements on the tether.
If the treadmill were very long, the belt could accelerate
very slowly and the tether would need to apply less
force. For a realistic treadmill, a user could be allowed to
accelerate somewhat close to the front edge to reduce
the tether force requirements.
Simulations were run to determine the integral and
proportional gains for the velocity control of the belt so
that the user nearly reaches the front of the treadmill
during maximum acceleration. The simulations em-
ployed the runner’s model and the treadmill model dis-
cussed earlier. Figure 9 shows the simulation results for
the subject’s position during maximum acceleration on
the treadmill using the acceleration parameters from
Equation (8). To minimize acceleration, it was decided
to start subjects approximately 0.4 m behind the center
of the treadmill. This reduced the required inertial force
for forward acceleration and still allowed subjects to
slow down and stop in a normal fashion. Figure 9 also
compares the velocity profiles (on the ground and on
the treadmill) and the interaction forces. The simula-
tions show that the velocities and contact forces are al-
most exactly the same on the ground as they are on the
Treadport. Figure 10 shows the commanded tether
force and belt velocity on the Treadport.
The acceleration of the treadmill belt was needed to
calculate the proper inertial force. The built-in velocity
controller of the treadmill provides an output with the
velocity of the treadmill. The data from this output were
very noisy, and calculating acceleration from this data
required filtering, which introduced lag that was too
great to use to implement the inertial force. To solve this
Christensen et al. 9
problem, a Kalman filter was used to estimate the accel-
eration. The inputs into the filter were the desired veloc-
ity and the velocity from the sensor. The fourth-order
model of the treadmill was used in the filter.
3.4 Effect of Saturation
The simulations showed that the inertial force
feedback would make the velocity profiles and interac-
tion forces exactly the same on the ground as on the
treadmill. However, Figure 11 shows the same simula-
tion with the maximum force limit of the tether (190 N)
included. Also, the maximum velocity was limited to 5
m/s for safety reasons. Due to these saturations, the
simulated results on the Treadport do not match exactly
with those on the ground. Because the tether force is
limited, it does not slow the user enough, so the velocity
and contact force on the Treadport both overshoot the
ideal curves for on the ground. Figure 12 shows the
commanded inputs with saturation.
4 Psychophysical Experiments
The dynamics behind the new control algorithm
were developed using a model of a cart with a velocity
feedback. To test if users preferred this new dynamics-
based algorithm, several comparisons were made
through psychophysical experiments. Three tether force
strategies were compared:
Figure 9. Simulated position, velocity, and interaction force during
acceleration on the treadmill.
Figure 10. Simulated commanded tether force and belt velocity.
Figure 11. Simulated position, velocity, and interaction force during
acceleration on the treadmill, with saturation.
10 PRESENCE: VOLUME 9, NUMBER 1
1. Zero force. The tether controller attempts to
maintain zero tether force on the user, and the us-
er’s position controls the velocity of the Treadport.
2. Spring force. The Treadport has been used suc-
cessfully for many demonstrations without inertial-
force feedback. The control scheme used thus far
applies a spring force proportional to the user’s
position; its purpose is to provide a kinesthetic cue
as to the middle of the treadmill. This spring force
has been tested with many users and has been
tuned to satisfy them. The force is not dependent
upon the user’s mass.
3. Inertial force. The tether force’s magnitude is
proportional to the product of the user’s mass and
the belt’s acceleration. The force is in the same di-
rection as the belt’s acceleration.
Eleven subjects volunteered to participate in the psycho-
physical experiments. There were three female and eight
male subjects ranging from twenty to fifty years old and
48 to 90 kg. Each subject was given instructions and
then allowed to warm up and become accustomed to the
Treadport. The subjects were told they would be com-
paring two different controllers at a time and would
have to pick the one that felt more realistic. They were
instructed to start each trial from rest and then acceler-
ate as quickly as they felt comfortable. Once they
reached a final velocity, they were asked to maintain it
for several seconds and then to gradually slow down un-
til they were stopped. At least two accelerations were
done for each controller, and the subjects were allowed
to change between the two controllers until they could
make a decision.
In preliminary testing, some subjects preferred a frac-
tion of the inertial force instead of the whole amount.
This could be due to the differences between the cart
model and the actual human dynamics. Another factor
could be that inertial force is distributed over the entire
body, while the equivalent tether force is concentrated at
one point. To address these initial observations, com-
parisons were made of acceleration with different per-
centages of the inertial force feedback: 60%, 80%, and
100% of the inertial force. The two subjects who chose
60% were also allowed to experience 40% of the inertial
force. Neither of them preferred this lower force, and it
was not tested on any of the other subjects. Table 1
shows that, on average, subjects preferred 80% of the
inertial force. The inertial force percentage that a subject
chose was then employed in comparison to the zero-
force and spring-force tether controllers.
During the psychophysical experiments, subjects were
instructed to accelerate as quickly as they felt comfort-
able. One subject accelerated to approximately 4.5 m/s,
which is near the current safety limit of the treadmill.
Figure 12. Simulated commanded force and velocity, with saturation.
Table I. Preferences in Psychophysical Experiments
Subject
Mass
(kg)
Preferred
control
algorithm
Preferred
percentage
of force
1 75 Inertial force 60
2 90 Inertial force 80
3 84 Inertial force 80
4 73 Inertial force 100
5 73 Inertial force 100
6 59 Inertial force 60
7 77 Inertial force 80
8 86 Inertial force 100
9 72 Inertial force 80
10 52 Inertial force 100
11 48 Inertial force 80
Median 73 Inertial force 80
Christensen et al. 11
Simulations had shown that subjects should be started
0.4 m behind the center of the treadmill. In one of his
initial starts, the subject was accidentally started in the
middle of the treadmill. He accelerated quickly enough
to come too close to the front and trip an emergency
switch. In this situation, an emergency shutoff automati-
cally stops the belt, and the tether’s power is turned off.
The subject was safely stopped before he ran off the
front of the treadmill and into the rear-projection display
screen. After this, he was started in the proper position.
The remainder of his trials went smoothly. This shows
that the simulations based on the human motion param-
eters agreed well with experimental data.
Table 1 also shows the preferences of the subjects
from the psychophysical testing. All subjects preferred
inertial-force feedback, then spring force, and then no
force. Subjects preferred 80% of the inertial force instead
of the full force. It was observed that the no-force strat-
egy made the treadmill locomotion practically unstable,
and subjects moved very tentatively.
Figure 13 shows real data during forward acceleration
of a subject on the Treadport with inertial-force feed-
back. The solid line is the force measured by the load
cell, and the dashed line is the commanded force, which
is 80% of the product of the user’s mass and the accelera-
tion of the treadmill. The data have the opposite sign of
the simulations due to the data acquisition. Although
the tether motor cannot produce a steady force of over
190 N, the force sensor measured larger forces. This
occurs because the tether motor cannot move the tether
fast enough, so the subject is also pulling the tether.
Note the high-frequency content of the measured force
due to the backlash in the tether connection.
5 Discussion
This paper identified a difference between locomo-
tion on the ground and locomotion on a treadmill to be
the inertial force caused by the acceleration of the tread-
mill. A controller was proposed for the Treadport to
cancel the inertial force with a force applied by a pow-
ered tether. Simulations using simplified models showed
that this controller should cancel the inertial force.
However, the actual implementation did not cancel the
inertial force as well as the simulations, mostly due to the
backlash of the tether-to-harness connection. Psycho-
physical experiments were conducted to determine
whether users preferred this theoretical control. The
inertial force was chosen unanimously over spring force
and zero force.
Subjects preferred 80% of the inertial force instead of
the entire force. A possible reason for this preference
could be due to the fact that the inertial force is distrib-
uted over the whole body and the force feedback is con-
centrated at the harness contact points.
The results indicate that inertial-force feedback causes
a definite improvement, a finding that is supported by
two cases. First, the psychophysical experiments show
that users prefer inertial-force feedback to a spring force
and no force. Second, the results of Frishberg (1983)
show that, without inertial force, there is significantly
less energy extraction during acceleration.
Future work should pursue several ideas in this area.
In this paper, the preferred inertial force was found for
each user and then compared to a spring force. Al-
though this spring force had been chosen by trial and
error with different subjects when the Treadport was
first developed, it was not presented to the subjects in
this test as often as inertial control was. Nor was it indi-
Figure 13. Actual force tracking of the tether during acceleration.
12 PRESENCE: VOLUME 9, NUMBER 1
vidually tuned to each subject. This could have caused a
bias towards inertial control. An improved experiment
would be to find the average inertial force and spring
force preferred by users and then test these two control
schemes against each other. Future research is also
planned in other tasks besides acceleration from rest.
A new Treadport being built at the University of Utah
will include a larger treadmill and a more powerful
tether. Improvements in actuation and sensing should
allow for even better results.
Acknowledgments
This material is based upon work supported under a National
Science Foundation Graduate Research Fellowship, by NSERC
NCE IRIS Project AMD-2, and by ONR Grant N00014-97-
1-0355.
References
An, C. H., Atkeson, C. G., & Hollerbach, J. M. (1988).
Model-Based Control of a Robot Manipulator. Cambridge,
MA: MIT Press.
Brogan, D. C., Metoyer, R. A., & Hodgins, J. K. (1997). Dy-
namically simulated characters in virtual environments. In
SIGGRAPH’97 Visual Proc. (p. 216).
Brooks, Jr., F. P., Airey, J., Alspaugh, J., Bell, A., Brown, R.,
Hill, C., Nimscheck, U., Rheingans, P., Rohlf, J., Smith, D.,
Turner, D., Varshney, A., Wang, Y., Webber, H., & Yuan, X.
(1992). Walkthrough Project: Final Technical Report to Na-
tional Science Foundation Computer and Information Sci-
ence and Engineering (Technical Report TR92-026). Univ.
North Carolina-Chapel Hill, Dept. Computer Science.
Christensen, R., Hollerbach, J. M., Xu, Y., & Meek, S. (1998).
Inertial force feedback for a locomotion interface. In Proc.
ASME Dynamic Systems and Control Division, DSC-Vol. 64
(pp. 119–126).
Cremer, J., & Kearney, J. (1995). Improvisation and oppor-
tunism in scenario control for driving simulation. First Work-
shop on Simulation and Interaction in Virtual Environments
(pp. 114–119).
Cruz-Neira, C., Sandin, D., & DeFanti, T. A. (1993). Sur-
round-screen projection-based virtual reality: The design
and implementation of the CAVE. In ACM SIGGRAPH’93
Proceedings (pp. 135–142).
Darken, R. P., & Cockayne, W. R. (1997). The omni-direc-
tional treadmill: A locomotion device for virtual worlds. In
Proc. UIST (pp. 213–221).
Distler, H., & Bulthoff, H. H. (1996). Psychophysical experi-
ments in virtual environments. Virtual Reality World ’96.
http://www.mpik-tueb.mpg.de/projects/bicycle/
bicycle.html.
Durlach, N. I., & Mavor, A. S. (Eds.). (1994). Virtual Reality:
Scientific and Technological Challenges. Washington, D.C.:
National Academy Press.
Ensor, J. R., & Carraro, G. U. (1997). Pelton: A VRML-based
bicycling simulator. In SIGGRAPH’97 Visual Proc. (p. 198).
Frishberg, B. A. (1983). An analysis of overground and tread-
mill sprinting. Medicine and Science in Sports and Exercise,
15, 478–485.
Greenwood, D. T. (1988). Principles of Dynamics. Englewood
Cliffs, NJ: Prentice-Hall.
Henry, F. M., & Trafton, I. R. (1951). The velocity of sprint
running with some observations on the muscle viscosity fac-
tor. Research Quarterly, 22, 409–422.
Hill, A. V. (1927). Muscular Movement in Man. New York:
McGraw-Hill.
Iwata, H. (1992). Force displays for walkthrough simulation.
In Proc. 2nd Intl. Symposium on Measurement and Control
in Robotics (pp. 481–486).
———. (1999). Walking about virtual environments on an
infinite floor. IEEE Virtual Reality ’99 Conference.
Iwata, H., & Fujii, T. (1996). Virtual perambulator: A novel
interface device for locomotion in virtual environment. In
Proc. IEEE 1996 Virtual Reality Annual Intl. Symp. (pp.
60–65).
Lorenzo, M., Poole, D., Denso, R., Lu, Y., Kekesi, A., Cha, J.,
Slayton, D., Williams, M., Moultan Jr., R., Kast, V., Mac-
Krell, W., Paddison, R., Riecks, J., Roth, M., & Wodoslaw-
sky, F. (1995). OSIRIS. In SIGGRAPH’95 Visual Proceed-
ings (p. 129).
Moghaddam, M., & Buehler, M. (1993). Control of virtual
motion systems. In Proc. IEEE/RSJ Intl. Conf. on Intelligent
Robots and Systems (pp. 63–67).
Noma, H., & Miyasato, T. (1998). Design for locomotion in-
terface in a large scale virtual environment. ATLAS: ATR
Locomotion INterface for Active Self Motion. In Proc.
ASME Dynamic Systems and Control Division, DSC-Vol. 64
(pp. 111–118).
Peterson, B., Wells, M., Furness III, T. A., & Hunt, E. (1998).
Christensen et al. 13
The effects of the interface on navigation in virtual environ-
ment. In Proc. Human Factors and Ergonomics Society.
Rieser, J. J., Pick Jr., H. L., Ashmead, D. H., & Garing, A. E.
(1995). The calibration of human locomotion and models
of perceptual-motor organization. J. Experimental Psychol-
ogy: Human Perception and Performance, 21, 480–497.
Roston, G. P., & Peurach, T. (1997). A whole body kinesthetic
display device for virtual reality applications. In Proc. IEEE
Intl. Conf. Robotics and Automation (pp. 3006–3011).
Salcudean, S. E., Drexel, P. A., Ben-Dov, D., Taylor, A. J., &
Lawrence, P. D. (1994). A six degree-of-freedom, hydraulic,
one person motion simulator. In Proc. IEEE Intl. Conf. Ro-
botics and Automation (pp. 2437–2443).
Slater, M., Steed, A., & Usoh, M. (1995). The virtual tread-
mill: A naturalistic metaphor for navigation in immersive
virtual environments. In M. Gobel (Ed.), Virtual Environ-
ments’95 (pp. 135–148). New York: Springer-Verlag.
Tristano, D., Hollerbach, J. M., & Christensen, R. (1999).
Slope display on a locomotion interface. Preprints 6
th
Inter-
national Symposium on Experimental Robotics (pp.
159–167).
Ward, M., Azuma, R., Bennett, R., Gottschalk, S., & Fuchs,
H. (1992). A demonstrated optical tracker with scalable
work area for head-mounted display system. In Proc. Symp.
Interactive 3-D Graphics (pp. 43–52).
Witmer, B. G., & Kline, P. B. (1998). Judging perceived and
traversed distance in virtual environments. Presence: Teleop-
erators and Virtual Environments, 7, 144–167.
14 PRESENCE: VOLUME 9, NUMBER 1
... The result is the Multi-Sensory TreadPort Active Wind Tunnel (MS.TPAWT, pronounced Ms. Teapot), a modified version of the TreadPort Active Wind Tunnel (TPAWT). The TreadPort was originally a VR environment that consisted of a large treadmill locomotion interface with a CAVE display [6], which lets users walk around and explore a VR world. A wind display was added via a large wind tunnel built around the TreadPort to create steerable wind that appeared to come from the graphical displays [7]. ...
... The result is the Multi-Sensory TreadPort Active Wind Tunnel (MS.TPAWT, pronounced Ms. Teapot), a modified version of the Tread-Port Active Wind Tunnel (TPAWT). The TreadPort was originally a VR environment that consisted of a large treadmill locomotion interface with a CAVE display [6], which lets users walk around and explore a VR world. A wind display was added via a large wind tunnel built around the TreadPort to create steerable wind that appeared to come from the graphical displays [7]. ...
... CAVE displays combined with locomotion systems are very popular for locomotion studies and gait therapy, however [4,63]. The user can interact with their physical environment, for example, using the railings on a treadmill or via advanced haptic displays that render terrain features [4], slopes [64][65][66][67], or inertial forces [6,68], while also experiencing their virtual world unencumbered. Physical therapists can more easily interact with the users, helping to guide their therapy. ...
Article
Full-text available
The purpose of the study was to understand how various aspects of virtual reality and extended reality, specifically, environmental displays (e.g., wind, heat, smell, and moisture), audio, and graphics, can be exploited to cause a good startle, or to prevent them. The TreadPort Active Wind Tunnel (TPAWT) was modified to include several haptic environmental displays: heat, wind, olfactory, and mist, resulting in the Multi-Sensory TreadPort Active Wind Tunnel (MS.TPAWT). In total, 120 participants played a VR game that contained three startling situations. Audio and environmental effects were varied in a two-way analysis of variance (ANOVA) study. Muscle activity levels of their orbicularis oculi, sternocleidomastoid, and trapezius were measured using electromyography (EMG). Participants then answered surveys on their perceived levels of startle for each situation. We show that adjusting audio and environmental levels can alter participants physiological and psychological response to the virtual world. Notably, audio is key for eliciting stronger responses and perceptions of the startling experiences, but environmental displays can be used to either amplify those responses or to diminish them. The results also highlight that traditional eye muscle response measurements of startles may not be valid for measuring startle responses to strong environmental displays, suggesting that alternate muscle groups should be used. The study’s implications, in practice, will allow designers to control the participants response by adjusting these settings.
... omnidirectional treadmills (Nagamori, Wakabayashi, and Ito 2005), and virtusphere (Fernandes, Raja, and Eyre 2003), require expensive specialized equipment that allows the user to walk in place. The main drawback of these techniques is that the inertial force feedback provided is not equivalent to natural walking (Christensen et al. 2000), therefore causing vection. ...
... Some methods even consolidate external physical props (Cheng et al. 2015), or manipulate the entire PTS (Suma, S. Clark, et al. 2011;Suma, Lipps, et al. 2012). Although they are plausible solutions to locomotion, they either fail to provide an inertial force-feedback necessary to furnish the sensation of self-motion (Christensen et al. 2000), or have many external dependencies. Software-based approaches: These techniques solely rely on digital manipulations for redirection. ...
Preprint
Full-text available
Modern Redirected Walking (RDW) techniques significantly outperform classical solutions. Nevertheless, they are often limited by their heavy reliance on eye-tracking hardware embedded within the VR headset to reveal redirection opportunities. We propose a novel RDW technique that leverages the temporary blindness induced due to saccades for redirection. However, unlike the state-of-the-art, our approach does not impose additional eye-tracking hardware requirements. Instead, SaccadeNet, a deep neural network, is trained on head rotation data to predict saccades in real-time during an apparent head rotation. Rigid transformations are then applied to the virtual environment for redirection during the onset duration of these saccades. However, SaccadeNet is only effective when combined with moderate cognitive workload that elicits repeated head rotations. We present three user studies. The relationship between head and gaze directions is confirmed in the first user study, followed by the training data collection in our second user study. Then, after some fine-tuning experiments, the performance of our RDW technique is evaluated in a third user study. Finally, we present the results demonstrating the efficacy of our approach. It allowed users to walk up a straight virtual distance of at least 38 meters from within a $3.5 x 3.5m^2$ of the physical tracked space. Moreover, our system unlocks saccadic redirection on widely used consumer-grade hardware without eye-tracking.
... Hardware-based techniques e.g. omni-directional treadmills [45], virtusphere [26], are not only an expensive solution to this problem but also fail to provide inertial force feedback equivalent to natural walking [7]. ...
... These prototypes, although they are possible solutions to the problem of infinite walking in VR, they all lack inertial force feedback. For this reason natural walking is considered to be the most preferred and natural way for locomotion in VR [7]. Moreover, the multimodal nature of natural walking allows free user movements such as jumping or crouching. ...
Article
Full-text available
Redirected walking is a Virtual Reality(VR) locomotion technique which enables users to navigate virtual environments (VEs) that are spatially larger than the available physical tracked space. In this work we present a novel technique for redirected walking in VR based on the psychological phenomenon of inattentional blindness. Based on the user’s visual fixation points we divide the user’s view into zones. Spatially-varying rotations are applied according to the zone’s importance and are rendered using foveated rendering. Our technique is real-time and applicable to small and large physical spaces. Furthermore, the proposed technique does not require the use of stimulated saccades but rather takes advantage of naturally occurring saccades and blinks for a complete refresh of the framebuffer. We performed extensive testing and present the analysis of the results of three user studies conducted for the evaluation.
... Hardware-based techniques e.g. omnidirectional treadmills [45], virtusphere [26], are not only an expensive solution to this problem but also fail to provide inertial force feedback equivalent to natural walking [7]. ...
... These prototypes, although they are possible solutions to the problem of infinite walking in VR, they all lack inertial force feedback. For this reason natural walking is considered to be the most preferred and natural way for locomotion in VR [7]. Moreover, the multimodal nature of natural walking allows free user movements such as jumping or crouching. ...
Preprint
Full-text available
Redirected walking is a Virtual Reality(VR) locomotion technique which enables users to navigate virtual environments (VEs) that are spatially larger than the available physical tracked space. In this work we present a novel technique for redirected walking in VR based on the psychological phenomenon of inattentional blindness. Based on the user's visual fixation points we divide the user's view into zones. Spatially-varying rotations are applied according to the zone's importance and are rendered using foveated rendering. Our technique is real-time and applicable to small and large physical spaces. Furthermore, the proposed technique does not require the use of stimulated saccades but rather takes advantage of naturally occurring saccades and blinks for a complete refresh of the framebuffer. We performed extensive testing and present the analysis of the results of three user studies conducted for the evaluation.
... The position of the person may also be measured with use of infrared sensors or rope sensors [32]. Another, slightly different solution that allows us to estimate the position of the patient is the application of pressure sensors located under the treadmill belt [5]. ...
Article
Training with use of mechatronic devices is an innovative rehabilitation method for patients with various locomotor dysfunction. High efficiency of training is noted in systems that combine a treadmill or orthosis with a body weight support system. Speed control is a limitation of such rehabilitation systems. In commercially available devices, the treadmill speed is constant or set by the therapist. Even better training results should be obtained for devices in which the speed of the treadmill will be automatically adjusted to the patient walking pace. This study presents a mechatronic device for locomotor training that uses an algorithm to adjust the speed of the treadmill. This speed is controlled with use of a sensor that measures the rope inclination. The end of rope is fastened to the orthopaedic harness. Speed control is realized in such a way that ensures the smallest possible swing angle of the rope. A fuzzy controller was applied to adjust the treadmill speed. The drive system of the treadmill is equipped in a servodrive with PMSM motor and energy recovery module, which allows smooth speed control, limiting acceleration and minimizing electricity consumption. The presented solution was implemented in a real object and subjected to experimental tests.
... The Sarcos Treadport uses an active harness to center the user and provide forces to make locomotion more natural (Christensen, Hollerbach, Xu, & Meek, 2000). Unlike our passive harness, it is an active mechanical tether that pulls and pushes the user. ...
Chapter
An immersive virtual environment (VE) portrays a scene that changes in response to a person's actions to give that person the impression of dealing directly with the three-dimensional (3D) simulated world. This new medium allows people to experience and interact within computer-generated environments. It relies on sensors to track a person's actions, processing to model and render the VE, and displays to present feedback. VEs enable the creation of action interfaces that allow the integration of skill development and cognitive decision-making. We consider an action interface to be a high-end interface that allows a person to perform a task in a manner similar to how it is done in the real world. With an action interface, a person performs perceptual/action sequences guided by procedural knowledge. The goal can be to develop proficiency at a task, engage in a purely recreational activity, or both. This chapter describes the design approach and implementation of an action interface for training small unit military tactics, techniques, and procedures (TTPs) for close quarters battle (CQB). Here we use the term CQB for systematic building clearing somewhat like SWAT methods used by police forces. CQB is a complex operation involving a high level of skill, knowledge, and team coordination. Creating an action interface to support CQB training and rehearsal is quite challenging. This process has lead us to develop a new general approach for designing user interfaces for virtual simulation based on properties derived from an analysis of how people perform real world actions to achieve their goals. This approach generalizes to less realistic, more abstract 3D interfaces as well, because they are defined by the same set of properties, although at a lower level of fidelity. The properties form a framework that can also be used for both usability analysis and to formulate structured scientific investigations.
Article
To achieve treadmill speed adaption to the user’s walking speed and reduce treadmill acceleration ripples during walking for enhancing a non-intrusive walking sense, this paper presents a novel strategy to analytically design the acceleration of treadmill by introducing a virtual force and a virtual force moment balance kinetic model. To explicitly handle the treadmill velocity to adapt the user’s walking speed as well as treadmill acceleration ripples, the designed acceleration of treadmill and its derivative are specially selected as the input variables of treadmill controller. Since the user’s relative position and absolute walking speed are all taken into account, the resulting velocity and acceleration ripples of treadmill can be properly updated to improve the sense of reality during the user walking by only tuning the defined design parameters and control parameters. The experiments are conducted and the effective of the proposed strategy is verified by results.
Article
Objective: To evaluate the differences between walking on an advanced robotic locomotion interface called the Treadport and walking overground with healthy subjects. Background: Previous studies have compared treadmill-based and overground walking in terms of gait parameters. The Treadport's unique features including self-selected speed capability, large belt, kinesthetic force feedback, and virtual reality environment distinguish it from other locomotion interfaces and could provide a natural walking experience for the users. Method: Young, healthy subjects (N = 17) walked 10 meters 10 times each for both overground and the Treadport environments. Comparison between walking conditions used spatiotemporal and kinematic parameters. In addition, electromyographic data was collected for five of the 17 subjects to compare muscle activity between the two conditions. Results: Gait on the Treadport was found to have no significant differences (p > .05) with overground walking in terms of hip and knee joint angles, cadence and stride length and stride speed, and muscle activation of the four muscle groups measured. Differences (p < .05) were observed in ankle dorsiflexion which was reduced by 2.47 ± 0.01 degrees on the Treadport. Conclusion: Walking overground and on the Treadport is highly correlated and not significantly different in 13 of 14 parameters. Application: This study suggests that the Treadport creates an environment for natural walking experience, where natural gait of users is almost preserved, with great potential to be useful for other applications, such as gait rehabilitation of individuals with walking impairments.
Article
Full-text available
This paper describes a metaphor that allows people to move around an immersive virtual environment by "walking on the spot". Positional data of participants' head motions are obtained from a tracking sensor on a head-mounted display during a training session, where they alternate between walking on the spot and a range of other activities. The data is used to train a feed-forward neural network that learns to recognise the person's walking on the spot behaviour. This is used in a virtual reality system to allow people to move through the virtual environment by simulating the kinds of kinesthetic actions and sensory perceptions involved in walking. An experiment was carried out to compare this method of navigation with the familiar alternative that involves using a hand-held pointing device, such as 3D mouse. The experiment, still continuing, suggests that the walking on the spot method may enhance the participant's sense of presence, but there is not sufficient evidence to suggest that it is advantageous with respect to the efficiency of navigation.
Article
People coordinate the force and direction of skilled actions with target locations and adjust the calibrations to compensate for changing circumstances. Are the adjustments globally organized (adjusting a particular action to fit a particular circumstance would generalize to all actions in the same circumstance); anatomically specific (every effector is adjusted independently of others) or functional (adjustments would generalize to all actions serving the same goal and generating the same perceptible consequences)? Across 10 experiments, changes in the calibration of walking, throwing, and turning-in-place were induced, and generalization of changes in calibration to functionally related and unrelated actions were tested. The experiments demonstrate that humans rapidly adjust the calibration of their walking, turning, and throwing to changing circumstances, and a functional model of perceptual-motor organization is suggested.
Article
In this paper, we focus on the sense of locomotion, in particular, the sense one feels while walking in a large space. Walking is the most basic way of moving around, and a lot of information can be obtained by doing so, e. g., sensing, hardness, humidity, and so on of the space. The locomotion interface we have developed is named ATLAS (ATR locomotion interface for active self motion). Our design goal was to eliminate the need for a walker to "learn" how to walk and for equipping the walker with obstructive sensors. We employ a treadmill approach for ATLAS. In the initial trial state, ATLAS can estimate the working speed from the walking motion visually, and control the speed of a belt to synthesize the real feeling of walking in a virtual space. First, we describe related works of locomotion interfaces and our design. Next, we show the method utilized for motion analysis and how to control the belt speed. Finally, we conclude with a description of our first trial ATLAS and the experimental results.
Article
This paper presents a virtual reality technology for walking through large scale virtual space such as buildings or urban space. Walkthrough simulation is effective for design and presentation of architecture and urban planning. Current systems for virtual reality use gesture input to move around virtual buildings. In terms of natural interaction, however, haptic feedback for walking motion is essential to exploration of virtual space. In this system, a walker wears omnidirectional sliding devices on the feet, which generate feel of walking while position of the walker is fixed in the physical world. The system provides a feeling of uneven surface of a virtual space, such as a staircase. While the walker goes up or down stairs, the feet are pulled by strings in order to apply reaction force from the surface. The motion of the feet and the head are measured by ultrasonicsensors. Scene of a virtual space is displayed on a head‐mounted display corresponding with the motion of the walker. A 6 degree‐of‐freedom master manipulator is developed to apply reaction force to the hand of the walker. The manipulator enables the walker to feel weight or rigidity of virtual objects.