Content uploaded by Matthew P Reed
Author content
All content in this area was uploaded by Matthew P Reed on Oct 13, 2014
Content may be subject to copyright.
2004-01-2176
Torso Kinematics in Seated Reaches
Matthew P. Reed, Matthew B. Parkinson and David W. Wagner
University of Michigan
ABSTRACT
Simulations of humans performing seated reaches
require accurate descriptions of the movements of the
body segments that make up the torso. Data to
generate such simulations were obtained in a laboratory
study using industrial, auto, and truck seats. Twelve
men and women reached to push-button targets located
throughout their right-hand reach envelopes as their
movements were recorded using an electromagnetic
tracking system. The data illustrate complex patterns of
motion that depend on target location and shoulder
range of motion. Pelvis motion contributes substantially
to seated reach capability. On padded seats, the
effective center of rotation of the pelvis is often within the
seat cushion below the pelvis rather than at the hips.
Lumbar spine motions differ markedly depending on the
location of the target. A categorization of reach targets
into four zones differentiated by torso kinematics is
proposed.
INTRODUCTION
Driver reach simulations are among the more common
applications of digital human models in the design of
vehicle interiors. The analysis is typically concerned
with whether controls or other targets can be reached by
a sufficient percentage of drivers. Prior to the
widespread use of human figure models in CAD, reach
analyses were primarily conducted using the reach
surfaces in SAE J287 (SAE 2003). The conditions under
which the data underlying J287 were gathered
(Hammond and Roe 1972) differ substantially from the
conditions in current vehicles, particularly in the restraint
system design.
J287 provides surfaces within which 95 percent of
drivers are expected to be able to reach with different
levels of torso restraint. This approach provides the
vehicle designer with important information for locating
controls, but is not useful for simulating reaches with
individual figure models. That is, J287 does not indicate
how or how far a person described by a particular set of
body dimensions would reach. Current SAE practice
also provides no guidance on the perceived difficulty of
submaximal reaches or reach kinematics. Work is
currently underway to replace the surfaces in J287 with
new, more flexible models based on data obtained in
conditions representative of current vehicles (Reed et al.
2003a).
Robotics approaches have been used to predict the
maximum reach envelope (Abdel-Malek et al. 2002),
although actual reach capability is not well predicted by
strictly kinematic approaches (Reed et al. 2003b).
Regardless, the maximum reach capability is of less
practical interest than the kinematics and subjective
difficulty with which
submaximal
reaches are performed,
because any plausible design for a control will place it
within the maximum reach capability of most drivers.
In the absence of standardized approaches to modeling
seated reach in vehi cles, research teams have
developed methods to predict either terminal postures or
motions using regression equations (Ryan 1970; Snyder
et al. 1972), optimization-based inverse kinematics
(Wang and Verriest 1998), analytical inverse kinematics
(Jung et al. 1995), optimization-based differential inverse
kinematics (Zhang and Chaffin 2000), and functional
regression on stretch-pivot parameters (Faraway 2003).
Additionally, most human figure models used for
ergonomic analysis (e.g., Jack, Safework, and RAMSIS)
provide for prediction of reach postures using inverse
kinematics. Reach motions are predicted by interpolating
between starting and ending postures. A variety of
heuristic and optimization-based approaches are used to
address the redundancy of the linkage. Interpolation-
based inverse-kinematics methods tend to produce
distinctly artificial movement patterns because the
interpolation methods are not based on human behavior
data.
The diversity of approaches to seated reach prediction in
commercial human models indicates that none of the
previously developed models of seated reach have
achieved widespread acceptance. One explanation for
the lack of consensus in motion prediction is that no
models have been published in a form that can be
Copyright © 2004 SAE International
readily implemented, and hence there is no opportunity
for independent validation of the reach prediction in
commercial tools. Several detailed posture prediction
models for seated reaches have been published (e.g.,
Synder et al . 1972), but those results are not
generalizable to motion. Other models lack sufficient
complexity in the torso. For example, the torso is
commonly represented by a single link between the hip
and shoulder (Jung et al. 1995) or between L5/S1 and
the sternoclavicular joint (Zhang and Chaffin 2000).
None of the published models predict pelvis motions,
which have been found to be important contributors to
torso mobility in seated reaches (Reed et al. 2003b).
The current study was conducted to provide the basis for
a new model of seated reach motions that would include
detailed torso kinematics and would be suitable for most
driver reach assessments. The primary objective of the
data collection was to record detailed torso kinematics
for seated reaches, including six degrees of freedom for
both the pelvis and thorax, for submaximal and maximal
reaches in a wide range of directions. Previous studies
of seated reaches performed in the Human Motion
Simulation (HUMOSIM) laboratory at the University of
Michigan used fixed targets distributed throughout, for
example, the simulated interior of a vehicle, with each
participant reaching to the same targets (Zhang and
Chaffin 2000). The current study used a computerized
target positioning apparatus to allow customization of the
target locations for each participant (Reed et al. 2003a),
ensuring that each participant would be presented near-
maximal targets in a wide range of reach directions.
This paper presents preliminary observations of some
salient characteristics of torso motion during seated
reaches. Any reasonable simulation method must
produce these characteristics, and hence they form a
foundation for selecting a simulation approach that has
the right balance between complexity and efficiency, an
important consideration with any model intended for real-
time use (Zhang 2003).
METHODS
Facility and Test Conditions
Testing was conducted in the HUMOSIM laboratory in
passenger car, heavy truck, and industrial seats. The
test seat is mounted on a motorized, rotating platform. A
push-button target is located on a motorized apparatus
that can move vertically and horizontally. The angle of
the button-mounting box can also be rotated around a
horizontal axis. By rotating the seat platform and
adjusting the horizontal and vertical target position, the
target can be placed anywhere within the participant’s
reach envelope. The entire system is under computer
control, so that a specified target location in a seat-
ce ntered c oordinat e system ca n b e obt ain ed
automatically. Figure 1 shows a participant in the test
facility.
Each participant was tested in each seat using
approximately 100 target locations distributed throughout
the right-hand reach envelope. After receiving a visual
signal, the participant performed a right-handed reach to
the target, pressed the button for two seconds with their
index finger, and returned to the home position.
A target location matrix was constructed with target
locations on six radial planes and five vector directions
with respect to horizontal. Figure 2 shows the sampling
planes with respect to the seat H-point and centerline.
The target locations were scaled using initial
measurements of each participant’s maximum vertical,
lateral, and forward reach. The scaling was designed to
place about 5 percent of the reach target locations
beyond the participant’s maximum. Target locations
were concentrated in the outer regions of the reach
envelope where the reach difficulty was expected to
change more rapidly with increasing distance from the
H-point. Because the steering wheel interfered with
forward reaches, the origin for the sampling vectors on
the -30, 0, and 30-degree planes (see top view in
Figure!6) was at shoulder height, rather than at H-point
height.
Figure 1. Participant in the test facility, showing rotating seat,
computer-controlled target-positioning apparatus, and motion
capture hardware.
Figure 2. Target location vectors. Angles in degrees.
Motion Capture
Motions were recorded using the electromagnetic Flock
of Birds system (Ascension Technologies). Each sensor
reports both position and orientation, so all six degrees
of freedom for a body segment can theoretically be
monitored with a single sensor. However, relative
movement between the sensor and body segment can
compromise the accuracy of the data. In this study,
redundant sensors were used on the thorax and the
pelvis to provide better tracking of the skeleton. Sensors
were placed on the sacrum and over the left and right
anterior-superior iliac spines (ASIS) of the pelvis. The
ASIS sensors were intended and used for position data
only, since the orientation of the sensors with respect to
the pelvis was not maintained during reaching. Sensors
were mounted on the sternum and over the T8 spinous
process. While the sternum sensor could be mounted
securely, the T8 sensor orientation was more affected by
movement of the skin and soft tissue, so the T8 sensor
was used only to establish, with the sternum sensor, the
orientation of the midsagittal plane. Additional sensors
were placed on the forehead, superior to the right
acromion process of the scapula, and on the lateral arm
immediately proximal to the elbow.
Immediately prior to testing, an FOB sensor attached to
a probe was used to record the locations of landmarks
on the participant’s head, thorax, pelvis, and right arm.
All of the FOB sensors were sampled simultaneously
with the probe sensor so that the locations of the
landmarks with respect to the coordinate systems of the
associated FOB sensors could be determined. These
landmarks were used to reference the FOB sensor
locations to anatomically based coordinate systems for
each body segment using relationships described in
Reed et al. (1999). For example, the locations of the hip
and L5/S1 joints were calculated using data from the
probe measurements of the left and right posterior
superior iliac spine landmarks and the FOB sensor
positions of the FOB sensors at the left and right ASIS,
and the position and orientation of the sacrum FOB
sensor. The pelvis orientation calculated using the
locations of the sacrum and ASIS sensors was
compared to the orientation obtained from the sacrum
FOB sensor to verify that the pelvis sensors did not shift
appreciably with respect to the participant during testing.
Data were sampled from each sensor at 25 Hz during
the motion. Joint locations were calculated from the
motion data and measured landmark locations, and
transformation matrices were calculated for each
segment of a linkage system consisting of pelvis,
abdomen, thorax, neck, head, right clavicle, and right
arm. Note that because data describing six degrees of
freedom are available for the pelvis, thorax, and head,
the length of the abdomen and neck segments is not
fixed in the data, which allows the motion of the lumbar
and cervical portions of the spine to be described in
ways that are more complex than are provided by one-
or two-joint lumbar or neck linkages.
In this paper, torso kinematics are illustrated using a
model of the skeleton. The measured scale, positions,
and orientations of the pelvis, thorax, clavicle, head, and
arm are used to display the associated skeletal
segments. The lumbar and cervical spines are
interpolated between the adjacent segments to provide
visualization of the associated changes in spine contour.
The lumbar and cervical spine visualizations should be
assessed qualitatively since no data were actually
gathered on spine contours in these regions. Similarly,
no data were gathered on scapula motion independent
of the clavicle, so the scapula in the visualizations
moves with the clavicle.
Data were obtained from six men and six women
stratified on stature to span the range from 154 cm to
194 cm. All participants were young adults ranging in
age from 22 to 28 years. Participants with low body
mass index (median 21.2 kg/m2, maximum 25 kg/m2)
were selected to facilitate placement of the sensors and
tr acking of the underly ing skeleta l st ructure s.
Consequently, the sample is not suitable for estimating
the range of movements that would be observed in a
larger sample more representative of the driving
population but may be adequate for quantifying the
features of typical seated reach motions.
RESULTS
Pelvis Kinematics
Pelvis mobility has been shown to be a key determinant
of reach capability for people with the minimal torso
constraint produced by safety belts equipped with
emergency locking retractors (Reed et al. 2003). When
people reach up, forward, or to the side to targets that
require torso motion, the pelvis rolls to facilitate the
reach. Pelvis motions for several types of reaches are
illustrated here. In all cases, we are concerned with
targets that require a substantial engagement of the
torso.
One important observation from this study is that the
pelvis does not generally roll around either the hips or
the ischial tuberosities. Relative to a seat surface,
seated pelvis motion is rarely centered on the hips, even
though changes in torso orientation in human figure
models are often based on hip rotation. Rotating around
the hips would require that the base of the pelvis shift
against the seat, a motion that is not feasible when a
significant fraction of the sitter’s body weight is borne by
the buttocks. On a rigid seat, the center of rotation of
the pelvis might be in the area of the ischial tuberosities,
since the skin under the buttocks typically remains
stationary with respect to the seat while the pelvis rolls
inside the skin.
In the current study, participants sat on three padded
seats, two of which had relatively thick foam cushions
(the truck and car seats). With padding under the
buttocks, changes in torso and pelvis orientation cause
changes in cushion penetration. Reaching to forward
targets that require torso involvement causes the pelvis
to roll forward, concentrating pressure under the
buttocks as the fraction of the torso weight that is
offloaded from the backrest is now over the buttocks.
The additional load carried by the pelvis drives the pelvis
lower in the seat as it is rolling forward, resulting in a
center of rotation that is below and slightly behind the
pelvis within the seat cushion.
Figure 3 shows torso kinematics for a high forward reach
and Figure 4 shows a low forward/lateral reach. Both
reaches were performed in the truck seat. In these
reaches to relatively distant targets, the pelvis rolls as
the sitter reaches. The deformation of the cushion is
seen as the pelvis drops relative to the seat H-point. In
both cases the average center of rotation for the pelvis is
slightly below the pelvis in the seat cushion.
Lumbar Spine Motion
The data show that torso motion occurs when either (1)
the target distance from the starting location of the
sternoclavicular joint is larger than the sum of the
lengths of the hand, forearm, arm, and clavicle
segments, or (2) reach to the target without torso motion
would exceed the ranges of motion of one or more upper
extremity joints.
The lumbar spine shows complex kinematics during
seated reaches that require torso motion. For many
reaches with torso involvement, the sternoclavicular joint
is at or near the boundary of its range of motion (ROM),
so that clavicle and thorax begin to rotate together as a
unit. For example, when reaching forward, the thorax
rotates contralaterally to allow the clavicle to point
toward the front of the body to a greater extent than
permitted by the sternoclavicular joint (of course, the
ROM at the sternoclavic ul ar joint is effectively
determined by the kinematic limitations of the entire
shoulder complex and not only locally).
Figure 3. Kinematics for a forward reach in a truck seat showing rotation of the pelvis. Crossed lines indicate the seat H-point.
Figure 4. Kinematics for a forward/lateral reach in a truck seat showing rotation of the pelvis. Crossed lines indicate the seat H-point.
There are four target zones within which torso
movements are distinctly different. These zones are
illustrated in Figure 5 with representative terminal
postures. In zone A, the target is sufficiently close to the
sitter that the shoulder ROM prevents the target from
being reached without a contralateral movement of the
torso. This motion is accomplished by lateral bending
and twisting in the lumbar spine, usually without
significant pelvis motion. The thorax is moved away
from the target and rotated to bring the target closer to
the front of the thorax. Since the outer margin of zone A
is determined primarily by shoulder range of motion,
zone A is located near and particularly behind the
shoulder. The external rotation limit of the humerus is
commonly the limit that necessitates torso motion in
zone-A reaches. In vehicles, seatbelts are often stowed
in zone A, necessitating a contralateral motion of the
torso to grasp the belt (Ebert and Reed 2002; Monnier et
al. 2003) and the console between the front vehicle
seats can lie in zone A, particularly for people who
choose more-forward seat positions.
In zone B, the target can be reached without torso
motion. Within zone B, the humerus does not reach the
boundary of the glenohumeral ROM and the target
distance from the initial position of the glenohumeral joint
does not exceed the combined lengths of the arm,
forearm, and hand (with a small amount of additional
distance achieved for forward and vertical reaches by
clavicle motion without thorax motion). However, the
outer limit of zone B, defined by the maximum reach
distance without torso motion, may be less than what
would be calculated from the total lengths of the upper
extremity segments (Delleman et al. 2003). Vehicle
designers usually attempt to locate reach targets within
zone B for most drivers by using the reach curves in
SAE J287 obtained from drivers wearing fixed-length
torso restraints. Reaches within zone B receive the
lowest difficulty ratings (Reed et al. 2003b).
In zone C, the lumbar spine flexes, bends, and rotates to
move the glenohumeral joint in the direction of the
target. By definition, there is little or no pelvis motion in
zone C, because thorax motion alone is sufficient. The
pelvis motion that is observed is due to cushion
deflection produced by movements in torso center of
mass. The outer margin of zone C is determined by the
ROM of the lumbar spine. Zone C reaches are rated as
more difficult than zone B reaches.
In zone D, the pelvis is rotated in the direction of the
reach, and lumbar spine flexion/bending/rotation is often
less than in zone C. That is, the lumbar spine
straightens out to increase the distance between the
pelvis and the glenohumeral joint.
The changes in lumbar spine motion across the zones
can be visualized by examining the terminal postures for
forward reaches shown in Figure 5. Zone A is
essentially non-existent for forward reaches at chest
level, because the shoulder ROM is well oriented for
targets directly in front of the body. Zone B spans the
range from directly in front of the sitter to a point
approximately at the upper (and forward) rim of the
steering wheel. In fact, one heuristic criterion used to
position steering wheels is that sitters can reach the top
of the rim while in zone B, that is, without moving the
thorax. As the target is moved further forward, into
zone C, lumbar flexion
increases
and lateral bending
and twisting of the spine is observed. However, as the
target moves into zone D, lumbar spine flexion
decreases
as the pelvis rolls forward. Lateral bending
and twisting of the spine remain the same as at the zone
C/D margin or increase as the pelvis rotation permits
greater movement (the bending and twisting ranges of
motion for the lumbar spine are greatest when near the
middle of the flexion/extension range of motion). At the
maximum reach envelope, the distance from the bottom
of the pelvis to the finger tip has been maximally
lengthened by extension, lateral bending, and rotation in
the lumbar spine, as shown in Figure 6 for a near-
maximal overhead reach.
Zone A Zone B Zone C Zone D
Torso moves away from
target to accommodate
shoulder range of motion
Minimal torso motion; target
can be reached using only
clavicle and upper extremity
Lumbar spine
flexion/bending/twisting in
direction of target but little
pelvis motion
Pelvis rolls and lumbar spine
moves toward extension to
lengthen torso
Figure 5. Illustration of typical terminal postures for reaches to targets in four zones.
Figure 6. Example of zone-D reach to an overhead target in a
car seat (starting posture on left, ending posture on right)
illustrating movement of lumbar spine toward extension. Seat
H-point location is shown.
DISCUSSION
These data demonstrate that seated reach motion
studies require measurement of six degrees of freedom
on the pelvis and at least three degrees of freedom on
the thorax. Clavicle mobility is also important and must
be quantified to obtain an accurate representation of the
motion. This suggests that seated reach prediction
requires modeling a kinematic chain with a fairly large
number of degrees of freedom: 6 (pelvis) + 3 (thorax) + 2
(clavicle) + 3 (shoulder) + 1 (elbow) + 3 (wrist) = 18. For
push-button motions, this could be reduced to 15, a
number that is still larger than the number of degrees of
freedom addressed by most of the previously reported
kinematic models of seated reach. Simpler linkages will
not be able to reproduce, for example, the transition from
zone C to zone D.
Pelvis m ob ility was found to be a n impo rt ant
characteristic of seated reaches, particularly reaches to
targets sufficiently distant to be of interest for ergonomic
analysis. On padded seats, the pelvis pivots around a
moving axis generally located below the pelvis, within
the seat cushion. This effective center of rotation is
produced by forward/lateral pelvis rolling and increased
penetration into the cushion as the weight of the torso is
offloaded from the seat back.
Examination of the lumbar spine motion revealed four
distinct motion patterns, differentiated by target zone,
that must be reproduced in a motion prediction model
applicable to seated reaches. The differences in
kinematics among the four zones indicate several
features of a good motion prediction model:
1. The model must take into account joint ranges of
motion, particularly at the shoulder, because the
shoulder ROM determines whether a near target will
produce zone-A or zone-B behavior in the torso.
2. The model must be capable of identifying the
transition between zone C, in which lumbar spine
mobility but not pelvis mobility is important, and zone
D, in which the pelvis rolls and the lumbar spine
flexion/extension can be opposite of that in zone C.
3. The model must be able to produce realistic pelvis
kinematics, which has been shown to depend on
penetration into the seat and hence seat cushion
stiffness.
Reach targets have been divided into zones based on
the characteristics of terminal postures in previous work
(e.g., Ryan 1970, Jung et al. 1995). The idea that the
torso becomes involved only when movements of the
distal segments is insufficient is not new (see Delleman
et al. 2003 for a review). The current work, however,
identifies the importance of the zone A/B and zone C/D
transitions, in which the to rs o motion in cl ud es
components in a direction opposite from or
perpendicular to the reach. Delleman et al. (2003),
analyzing terminal postures, showed that the transition
from zone B to zone C for lateral reaches occurred prior
to reaching the end of the range of motion of the upper
extremity. This finding highlights the need for detailed
information on torso kinematics for zone C reaches.
Previous work has also shown that lateral and near-
lateral maximum seated reaches are balance limited,
rather than joint range-of-motion limited (Reed et al.
2003b). Hence, accurate prediction of lateral reach
capability and motions requires consideration of balance
and the ability of the sitter to generate counterbalancing
forces with the contralateral hand (by gripping the
steering wheel, for example).
The generality of the observations in this study is limited
by the relatively homogenous participant pool. The
participants were all young people with low body fat.
Older sitters are known to have different maximum reach
capability due to balance maintenance limitations
(Parkinson et al. 2002) and also may have different
reach kinematics (Chaffin et al. 2000). For example, the
size of zone A, which is determined by shoulder ROM
motion, might be quite different for older sitters. Sitters
with higher levels of body fat may also have different
reach kinematics. The extensive data gathered in the
current study will allow subsequent work with more
diverse populations to be conducted more efficiently.
Future studies should examine the influence of
environmental obstructions, such as the seat back.
Even more important will be generalizing the current
work to examine the effects of hand orientation and force
application at the termination of the reach.
Work is now underway to develop a motion-prediction
model for seated reaches that reproduces the behaviors
observed in the four reach zones described in this paper.
The ultimate goal is to publish a seated reach model in
fully implementable form. Such a model will provide an
independent validation of the reach models available in
commercial digital human modeling tools. Additional
data collection will be necessary to assess the
generalizability of the model to more diverse populations
and to simulate reaches to targets other than push-
button controls.
ACKNOWLEDGMENTS
This research was sponsored by the University of
Michigan Automotive Research Center and by the
partners of the Human Motion Simulation (HUMOSIM)
program at the University of Michigan. HUMOSIM
partners include DaimlerChrysler, Ford, General Motors,
International Truck and Engine, United States Postal
Service, U.S. Army Tank-Automotive and Armaments
Command, and Lockheed Martin.
REFERENCES
Abdel-Malek, K, Yang, J., Brand, R., Tanbour, E.,
(2001). Towards understanding the workspace of the
upper extremities. Technical Paper 2001-01-2095.
Society of Automotive Engineers, Inc., Warrendale, PA.
Chaffin, D. B., J. J. Faraway, X. Zhang, and C. Wooley,
(2000). Stature, Age, and Gender Effects on Reach
Motion Postures,
Human Factors,
42:3, 408-420.
Delleman, N.J., Huysman, M.A., and Kuijt-Evers, L.F.M.
(2003). Postural behavior in static reaching sidewards.
Technical Paper 2003-01-2230. SAE International,
Warrendale, PA.
Ebert, S.M. and Reed, M.P. (2002). Characterization of
Driver Seatbelt Donning Behavior. Technical Paper
2002-01-0783. SAE International, Warrendale, PA.
Faraway, J.J. (2003). Data-based motion prediction.
Technical Paper 2003-01-2229. SAE International,
Warrendale, PA.
Hammond and Roe (1972). SAE controls reach study.
Technical Paper 720199. Society of Automotive
Engineers, Inc., Warrendale, PA.
Monnier, G., Wang, X., Verriest, J.P., and Goujon, S.
(2003). Simulation of complex and specific task-
orientated movements – application to the automotive
seat belt reaching. Technical Paper 2003-01-2225. SAE
International, Warrendale, PA.
Parkinson, M.B., Chaffin, D.B., and Reed, M.P. (2002).
Maintaining Balance in Seated Reaches. Proc. RESNA
Annual Conference. Rehabilitation Engineering and
Assistive Technology Society of North America,
Arlington, VA.
Reed, M.P., Parkinson, M.B., and Klinkenberger, A.L.
(2003a). Assessing the validity of kinematically
generated reach envelopes for simulations of vehicle
operators. Technical Paper 2003-01-2216. SAE
International, Warrendale, PA.
Reed, M.P., Parkinson, M.B., and Chaffin, D.B. (2003b).
A new approach to modeling driver reach. Technical
Paper 2003-01-0587. SAE International, Warrendale,
PA.
Reed, M.P., Manary, M.A., and Schneider, L.W. (1999).
Methods for measuring and representing automobile
occupant posture. Technical Paper 990959. SAE
Transactions: Journal of Passenger Cars, Vol. 108.
Ryan, P. W. (1970). Cockpit geometry evaluation, Joint
Army-Navy Aircraft Instrumentation Research Report
700201. The Boeing Company, Seattle, WA.
SAE International (2003). Automotive Engineering
Handbook. SAE International, Warrendale, PA.
Synder, R.G., Chaffin, D.B., and Schutz, R.K. (1972).
Link system of the human torso. Technical Report
AM RL-TR 071-8 8. Ae orspa ce Medic al Re searc h
Laboratory, Wright-Patterson Air Force Base, Ohio.
Wang, X. and Verriest, J.P. (1998). A geometric
algorithm to predict the arm reach posture for computer-
aided ergonomic evaluation.
Journal of Visualization
and Computer Animation
, 9:33-47.
Zhang, X. (2001). Biomechanical Realism Versus
Algorithmic Effeciency: A Trade-off in Human Motion
Simulation and Modeling,
SAE Transactions,
110:6,
2184-2191.
Zhang, X. and D. Chaffin (2000). A Three-Dimensional
Dynamic Posture Prediction Model for Simulating In-
Vehicle Seated Reaching Movements: Development and
Validation,
Ergonomics,
43:9,1314-1330.