1 Introduction
STUDIES ON wheelchair propulsion have mainly been
focused on physical stress or on technical improvements.
Some studies have tried to combine physiological and
more biomechanical techniques to find criteria for opti-
misation of wheelchair design. In this analysis of the
wheelchair/user interface, physiological and biomechanical
measurements under known, manipulatable conditions
which simulate real-life wheelchair propulsion, are needed.
Measurements should not only include exact information
on the applied workload, but also information on the
forces applied by the user to overcome this workload and
on the metabolic cost of doing so.
The validity of any parameters to he measured is
directly related to the verisimilitude of the simulation. An
ergometer must not only allow accurate measurement of
power output and applied forces, but must also simulate
wheelchair driving, at least mechanically, as realistically as
possible. Moreover, because comparison of different wheel-
chair dimensions is required, the ergometer should allow
easy adaptation of the wheelchair dimensions. A stationary
ergometer is preferable, because it would facilitate the
measurement of physiological and kinesiological par-
ameters.
First received 5th April and in final form 31st August 1989
~) IFMBE: 1990
The demands with which such an ergometer must
comply can be summarised as follows:
(a) adequate simulation of frictional losses due to air and
rolling resistance, velocity and slope, together leading
to continuous control of workload calculated in a
power balance
(b) realistic simulation of the linear inertia of the wheel-
chair user system
(c) measurement of torques and forces exerted on rims as
well as on the seat and backrest, during wheelchair
propulsion
(d) possibility of analysing the effect of different wheelchair
configurations
(e) possibility of measuring physiological and kinesio-
logical parameters (movement analysis, electromyo-
graphy).
At present various types of wheelchair ergometer design
are used. These designs can be classified as follows:
(i) measurements on a track (GORDON, 1952; PEIZER
et
al.,
1964; LEHMAN
et al.,
1974; GLASER
et al.,
1980;
SANDERSON and SOMMER, 1985), which are however
hard to standardise, whereupon power output is diffi-
cult to quantitfy. Because the subject is not fixed to
one place physiology can only be studied with port-
able devices and/or telemetry
Medical & Biological Engineering & Computing July 1990 329
(ii) wheelchairs used on a treadmill (VOIGHT and BAHN,
1969; ENGEL and
HILDEBRANDT, 1973; DILLMANN and
NIETERT, 1980; LESSER, 1986; VAN DER WOUDE
et al.,
1986; 1989c). This experimental set-up allows proper
analysis of physiology, kinematics and electro-
myography. Wheelchair propulsion is simulated in a
most realistic way (inertia; steering required), although
air resistance is absent. However, the latter is only
important at relatively high velocities. Power output
can be determined empirically in a separate drag test
(VAN DER WOUDE
et al.,
1986; 1989c). Forces and
torques applied to the propulsion mechanism and/or
seat and backrest can not, however, be studied in dif-
ferent wheelchair configurations
(iii) wheelchairs used on rollers
(BROUHA and KROBATH,
1967; STOBOY
et al.,
1971; BRAUER, 1972; MOTLOCH
and BREARLEY, 1983)
(iv) wheelchair assemblies connected to a bicycle ergome-
ter
(BRATTGAARD et
al.,
1970; WICKS
et al.,
1977;
GLASER
et al.,
1980; FORCHHEIMER and LUNDBERG,
1986). (iii) and (iv) have the advantages of a stationary
set-up, However, there is again the difficulty in mea-
suring torques and forces. As in the treadmill and
track analysis, different wheelchair configurations and
anthropometry-based variations of the v~heelchair are
hard to study. Moreover, wheelchair propulsion is
simulated with a simple mechanical or electric brake.
Simulation of inertia (for instance with rotating disks)
generally cannot be adapted to individual character-
istics. Determination of power output can be inaccu-
rate. This is especially important in short duration
maximum effort tests (20-30 s). Isometric or isokinetic
studies generally cannot be conducted on these
devices
(v) seats with separate, instrumented rims
(BRUBAKER et
al.,
1981; JARVIS and ROLFE, 1982; LESSER, 1986;
BURKETT
et al.,
1987). These set ups do meet our
demands to a certain but again still limited extent.
None of these designs fully comply with the demands in
our first list. A complete simulation of the frictional losses
in wheelchair driving is not possible in any of them.
Wheelchairs used on a treadmill give good validity, but do
not easily allow for either the exact calculation of work-
load or the measurement of forces on the hand-rims. Instru-
mentation to meet these requirements would be very
costly, if different wheelchair designs are to be compared.
Wheelchairs on rollers or combined with a bicycle ergome-
ter need proper compensation for the inertial properties of
the wheelchair/user system. Some ergometers have a provi-
sion for this (JARVIS and ROLFE, 1982, FORCHHEIMER and
Fig. 1
LUNDBERG, 1986; BURKETT
et al.,
1987). Again, the instru-
mentation of the chair is expensive.
In a joint project with the Free University Amsterdam,
the Erasmus University, Rotterdam has developed a sta-
tionary wheelchair ergometer (Fig. 1) which is designed to
overcome the above problems. It comprises:
(a) a mechanical design which allows for the study of indi-
vidual characteristics of the wheelchair/user interface.
This design is highly adjustable for a wide range of
different positions of handrims, seat and backrest posi-
tions and angles, without interfering with the instru-
mentation
(b) an electronic control system simulating frictional losses
on the basis of feedback and also making possible iso-
kinetic and isometric measurements. The rear wheels
can be controlled separately, so that the simulation of
wheelchair propulsion on a side slope is made possible
(c) a force measuring system allowing for the measurement
of tangential, radial and medio-lateral forces of the
hand on the handrims and for the measurement of
reaction forces of the seat and backrest and the lower
body and trunk
(d) this ergometer design will serve the detailed analysis of
the wheelchair/user interface. Such a fundamental
approach should ultimately lead to guidelines for
wheelchair design and to a fitting model of the wheel-
chair to the user.
The wheelchair ergometer in use in the biomechanical
laboratory
2 Controlling forces in hand-rim wheelchair
propulsion
The most important of the controlling forces is the one
the wheelchair user exerts on the hand-rim (Fig. 2). If Fh is
the resulting tangential force on the rim there will be a
propelling force on the axis of the wheelchair
F~ = F h
Rh
R--~ (1)
in which Rh and R w are the radii of the hand-rim and the
wheel. As soon as F~ overcomes the stationary rolling
resistance the wheelchair will start to roll and from this
moment on the following factors will have their influence
on the acceleration and speed of the wheelchair:
(a) inertia of the wheelchair/user system
(b) rolling resistance
(c) air resistance
(d) slope.
Rw
Rh
Fs=F h Rh
Rw
Fig.
2
Force on hand-rim results in propelling force F~
330 Medical & Biological Engineering & Computing July 1990
To express these factors in a formula of the form F =
ma
we will use the following definitions:
(i) rolling resistance (Fig. 3) is a torque calculated from
the total weight of the wheelchair user system W and
the imaginary distance between the weight Vector and
the normal reaction vector N. The force on the wheel-
chair as a result of rolling resistance is
We
F, - (2)
R~
In this formula W and
R w
are constants for a wheel-
chair user system; e varies with tyre and road charac-
teristics.
(ii) Airresistance is a force proportional to the square of
the relative airspeed (V w - 111) and the frontal area of
the wheelchair user system O. It can be stated as
F a = 89 w -- v1)ZOCa
(3)
in which
p = density of the air (1.23kgm -3 at atmospheric
conditions)
V~ = wheelchair velocity m s-1
V1 = windspeed m s- 1
Ca
= resistance factor
The sign of F, is determined by the sign of (Vw - 1"1).
(iii) When the wheelchair is running on a slope there will
be a force on the wheelchair
F~ = W sin ct (4)
where ct is the angle of the slope9 An upward slope will
9 . "~ , .
be indicated as positive.
Rw
I
( .- ,
W ! =
Fr W E
/ Rw
/I// sl II//
Fig. 3
Definition of rolling resistance F,
Now the resulting force F t on the wheelchair can be calcu-
lated with the equation
F t = F~ - F, F a F~
(5)
and the equation of motion is
W
F~ = -- a (6)
g
where g is gravitation and a is the acceleration of the
wheelchair.
This results in
Ftg
a - (7)
W
Now let us suppose we have a stationary wheelchair--a
Medical & Biological Engineering & Comput;ng
simulation of a real wheelchair--with flee-turning wheels
in which we continuously measure the torque
Fh,
also
when the wheels are turning.
And let us suppose we have found values for/7,, F a and
F~; then we can calculate at any moment the acceleration
a of the simulated wheelchair. Integration of a according
to
v = fa dt
(8)
will give us the momentary speed of the chair.
To calculate F, we have to determine two variables, W
and ~. W can be calculated from the weight of the user and
an assumed weight for the wheelchair9 During the test this
is a fixed value, e can be varied during the test according
to the road conditions to be simulated9
/7, can be calculated with an estimated value for O, a
given value for the windspeed, which may vary during the
test, and a value for the drag Ca. This factor is approx-
imately 1.4 (CoE, 1979)9 Further, we need a value for the
wheelchair speed Vw, but this is exactly what we want to
calculate. However, as these computations can be carried
out by computer at a rate of approximately 50 per second,
the latest calculated value for Vw can be used here without
causing a relevant error.
After a value for a has been set, F~ can easily be calcu-
lated. The slope angle ~ can also be varied during the test.
Thus the acquired 'speed' of the simulated wheelchair is
known at any moment9 If we now have the wheels propel-
led proportionally to this calculated speed by means of
speed-controlled servomotors we have created a realistic
simulation of wheelchair propulsion under varying condi-
tions.
3 Internal forces in the wheelchair/user system
Applying a force to the hand-rim will result in a reaction
force on the user of the wheelchair. This force will be
equalised by forces in the seat and the backrest of the
chair. The study of these forces requires force transducers
in the seat and backrest in frontal and vertical directions9
During actual propulsion the user exerts forces on the
hand-rim which can be resolved into three directions:
tangential, radial and axial9 Only the tangential force will
contribute to the propulsion of the wheelchair, the other
two can be seen as losses9 To study these forces three-
directional force transducers will be required in the frame
of the wheel suspension9
4 Measuring modes
Biomechanical research on human power output
requires an analysis of three different ways of force gener-
ation
(a) realistic wheelchair simulation (isoinertial)
(b) isokinetic
(c) static.
Isoinertial measurements will be obtained by means of a
realistic simulation of wheelchair propulsion. As described
in the previous sections, the system reacts to the forces
applied to the hand-rim. Isokinetic measurements are the
study of manual forces at a constant speed of the wheels,
which is not affected by forces exerted on the hand-rims.
Static measurements in hand-rim wheelchair propulsion
means the study of all forces while the wheels are totally
blocked.
These three measuring modes complete the list of
demands with which the ergometer must comply9
July 1990 331
5 Stationary wheelchair design
5.1
Mechanics
Fig. 4
gives
an impression of the design of the stationary
wheelchair ergometer. Seat and backrest are mounted on a
console (11) with a hydraulic foot-operated height control.
The console itself can be moved in a forward or backward
direction. In this way the chair can be positioned accu-
rately with regard to the independently mounted wheels.
The seat (9) and backrest (4) consist of two stiff frames,
connected by three two-directional force transducers (10, 5)
for frontal and vertical force measurement. Cushions of
industrial manufactured wheelchairs can be mounted on
the seat and backrest. Both seat and backrest can be tilted
independently over 45 ~ Arm rests and legrests are adjust-
able and optional in use.
Next to the chair there are two side frames (1) for wheel
suspension. Each side frame is mounted on three three-
directional force transducers (12) (Fig. 5) for measuring all
forces exerted on the hand-rims during propulsion. The
wheels (3) (together with the driving motors) are mounted
on subframes (2) to make possible a camber of the wheels
of up to 15 ~
The main suspension frames (1) can be moved sideways
to adjust the width of the chair. Rims with different dia-
meters can be mounted on the wheels. Between the hand-
rims and the wheelchair-axis torque transducers (8)
(GOMMERS, 1976; PRONK and NIESING, 1983) are mounted
to measure the real torque that causes the propulsion. The
motors (7) are 1000W printed circuit motors (Mavilor
1000) mounted behind the wheel axis. Transmission is rea-
lised by timing-belts (6).
Tables 1 and 2 show the possible adjustments and the
location and range of the force transducers.
5.2
Electronics
The electronic control unit (ECU) calculates the 'acceler-
ation' and 'speed' of the wheelchair. Parameters for weight,
air resistance, rolling resistance, acquired velocity and
slope are taken from the software system (Fig. 6). The
torque transducers give real values through a specially
:~--~ I
_-- -rl "
',
l
Fi~
4
Schematic front (left) and side view (right) of the ergometer. 1: side frames: 2: subframes : 3: wheels: 4: backrest: 5: two-
directional force transducers; 6: timing belts: 7: motors; 8: torque transducers; 9: seat; 10: two-directional force transducers. 11 :
console: 12: three-directional force transducers
Table 1 Adjustments of wheelchair dimensions
Adjustment Range Reference
Height control scat and backrest -6-24cm wheel axis
Fore/aft alignment and scat and backrest -22.5-10 cm wheel axis
Tilting of scat and backrest 0~5' horizontal
Wheelbase (side frames) 56-80 cm
Camber 0-1Y vertical
Table 2 Force transducers
Location Number Measured force Range
Fig. 5
332
Three-directional force transducer
Scat 3 perpendicular on scat (z) 3000N
Seat 3 in plane of the scat (x) 750 N
Backrest 3 perpendicular on backrest (z) 1150 N
Backrest 3 in plane of the backrest (x) 1150N
Wheels 2 torque handrim-wheel 100 N m
Side frames 2 x 3 vertical force on side frame Iz) 850N
Side frames 2 x 3 horizontal force on side frame (xl 850N
Side frames 2 x 3 horizontal force on side frame (y) 850N
Medical & Biological Engineering & Computing July 1990
designed floating rotating measuring system. With eqns.
5-7 the actual 'speed' and thus the signals for the control
amplifiers of the motors can be computed.
A second part of the electronic system contains the
amplifiers for the 30 strain-gauge bridges for force mea-
surement. Together with the torque transducers and the
tachogenerators there are 34 variables which are sampled
at 50Hz. To handle this amount of information
(20 400 bits s -1) a data-acquisition unit (DAU) has been
developed with a parallel connection to the PC-bus
through a dual port RAM-memory.
Fig. 6
[ pc HEC U $ervo [ motor
Diaoram of wheel speed control
I
t
II
wheel ]
,orqu, I
trQnsducer
t Fhr h
Stepwise multiple regression on all side frame trans-
ducers showed that only transducers measuring in corre-
sponding directions contributed significantly to the applied
calibration forces. The standard error of estimate stayed
below 4.5 N for a gauging range of 400 N in all three direc-
tions. Subsequent tests showed independence of the posi-
tion of forces applied to the rims. Moreover, the accuracy
of simple linear regression equations using summated
values of transducers in the x, y and z directions was
shown to be comparable to the more complex multilple
regression methods. Crosstalk of other transducers stayed
below 3 per cent in all cases.
5.3
Software
The custom-made made software for use with the wheel-
chair ergometer contains four modules:
(i) administration
(ii) data communication
(iii) curve processor
(iv) processing utilities.
The administration module is a database containing files
on the different test subjects/patients. The data used to
define the experimental conditions (parameters for the fric-
tional losses) are entered in the file. The data communica-
tion module gives real-time feedback for the subject on a
screen containing information on the actual and required
speed and direction of travel of the wheelchair. This
module handles communications between the PC and the
ECU (downloading of parameters), and between the DAU
and the PC.
The curve processor is a program with the help of which
the measure d signals can be analysed interactively. A curve
can be read in and processed. Processed data will be stored
under a unique name, so that it is possible to repeat ana-
lysing steps. In the utility package curve processor com-
mands are called from the command line. Series of calls
can be collected in a batch file, enabling the user to access
these command series by typing the single batch file name.
Medical & Biological Engineering & Computing
6 Preliminary results and discussions
The instrumentation in the analysis of a complex
dynamic process such as manual wheelchair propulsion is
of the utmost importance. In the study of the wheelchair/
user interaction and the subsequent development of guide-
lines for propulsion technique and the geometry of the
wheelchair/user interface, a combined physiological and
biomechanical analysis is required. This required a testing
device, which simulated wheelchair propulsion and
allowed for the study of similar but simplified motions
(isokinetics) and force application (isometrics). This led to
the design and building of the stationary wheelchair
ergometer presented in the previous part. The application
of section such an experimental device lies within several
areas of scientific interest:
(a) analysis of propulsion technique under both sports-like
and daily-use conditions of wheelchair ambulation.
Effective torque and mediolateral and radial hand
forces give an indication of effectiveness of force appli-
cation
(b) development of an (inverse dynamics based) segment
model of the arm-shoulder-trunk. Forces generated on
the propulsion mechanism in conjunction with kine-
matics and electromyography serve the calculation of
net torques over the joints and the study of muscle
co-ordination. Thus technique may be defined in more
detail and it will serve the understanding of the 'human
engine'. Subsequently, (over-)loading of joints or
muscle groups can be studied. Together with physio-
logical techniques, the variations in cardiorespiratory
parameters between different wheelchair configurations
can be studied in biomechanical detail
(c) the role of the seat and backrest in wheelchair propul-
sion and on the stability of the trunk in lower body
disabled can be studied in detail.
Pilot studies were conducted to attain a first impression of
the overall functioning of the prototype wheelchair ergo-
meter, both under 'normal' daily-use conditions (VAN DER
WOUDE
et aL, 1989a; b)
and under the more extreme con-
ditions of a short-duration wheelchair sprint test (VEEGER
et al., 1989;
VAN DER WOUDE
et al., 1989d).
In the first pilot experiment six male able-bodied sub-
jects conducted three incremental exercise tests. The
required velocity increased 0.28 m s-' every three minutes
(0.56-1.66 m s-i). Each test was conducted at different seat
heights defined by elbow angle when sitting in an upright
position with the hands on the top-dead centre of the rims
(VAN DER WOUDE
et al., 1989b).
The wheelchair/user inter-
face, dominated by the position of the propulsion mecha-
nism with respect to the seat/backrest, was adapted to
individual anthrometric dimensions. These results were in
line with previously results from a study of sitting height
during wheelchair propulsion on a motor-driven treadmill
and indicated an optimum seat height of around 100-200 ~
elbow angle (VAN DER WOUDE
et al., 1989b).
In the second pilot study a number of wheelchair sprint
tests were performed on the wheelchair ergometer at differ-
ent friction loads (VEEGER
et al.,
1989; VAN DER WOUDE
et
al., 1989d).
Different propulsion technique parameters were
determined from the raw torque and velocity data. An
individual curve of the power output (i.e. the product of
momentary torque and angular velocity) at an interme-
diate level of friction is shown in Fig. 7. The sample fre-
quency of 100Hz used iri this study allowed for
within-cycle analysis of data. After recursive digital filter-
ing using one of the curve processor options (f~ = 20 Hz)
each individual push (hand-to-rim contact) or recovery
phase (period in which the hands progress from the end of
July 1990 333