Short communication

Robot-based methodology for a kinematic and kinetic analysis of
unconstrained, but reproducible upper extremity movement

Nikica Popovic

a,



, Sybele Williams

b

, Thomas Schmitz-Rode

a

, Gu¨nter Rau

a

,

Cantherine Disselhorst-Klug

a

a

Department of Applied Medical Engineering, Helmholtz Institute, RWTH Aachen University, Germany

b

Department of Physics, The University of the West Indies, St. Augustine, Trinidad and Tobago

a r t i c l e

i n f o

Article history:
Accepted 27 March 2009

Keywords:
Upper extremity
Movement analysis
Reproducibility
Kinematics
Kinetics

a b s t r a c t

Although arm movements play an important role in everyday life, there is still a lack of procedures for
the analysis of upper extremity movement. The main problems for standardizing the procedure are the
variety of arm movements and the difficult assessment of external hand forces. The first problem
requires the predefinition of motions, and the second one is the prerequisite for calculation of net joint
forces and torques arising during motion. A new methodology for measuring external forces during
prespecified, reproducible upper extremity movement has been introduced and validated. A robot-arm
has been used to define the motion and 6 degrees of freedom (DoF) force sensor has been attached to it
for acquiring the external loads acting on the arm. Additionally, force feedback has been used to help
keeping external loads constant. Intra-individual reproducibility of joint angles was estimated by using
correlation coefficients to compare a goal-directed movement with robot-guided task. Inter-individual
reproducibility has been evaluated by using the mean standard deviation of joint angles for both types
of movement. The results showed that both inter- and intra-individual reproducibility have significantly
improved by using the robot. Also, the effectiveness of using force feedback for keeping a constant
external load has been shown. This makes it possible to estimate net joint forces and torques which are
important biomechanical information in motion analysis.

&

2009 Elsevier Ltd. All rights reserved.

1. Introduction

Today, the standardised measurement of both three-dimen-

sional kinematics and kinetics together with muscle activity using
surface EMG (SEMG) is the usual procedure in clinical gait
analysis (

Chambers and Sutherland, 2002

). Motion analysis

systems in combination with underlying biomechanical rigid
segment models (

Kadaba et al., 1990

;

Davis et al., 1991

) have been

used to calculate joint angles. From these, other kinematic data
such as joint velocity and acceleration of lower extremity
movements can be determined. For the kinetic description of
motion it is necessary to measure the forces acting on the body
during movement. In gait analysis, those external forces are
commonly acquired using force plates which detect the ground-
reaction forces. The kinematic and kinetic data can then be used
as inputs for a kinetic model (

Bresler and Franke, 1950

;

Cavagna

and Magaria, 1966

), which calculates net joint moments and net

joint forces.

However, there is a lack of methods for the assessment of

arbitrary upper extremity movements, which are not restricted or
repeatable, as compared to the movement’s characteristic of gait
(

Rau et al., 2000

). Many robot-assisted methods which can be

end-effector-based (

Hogan et al., 1995

;

Krebs et al., 1998

;

Burgar

et al., 2000

) or in form of an exoskeleton (

Sanchez et al., 2006

;

Nef

et al., 2007

) have been used in rehabilitation for arm therapy.

However, there are no reports on using robots in the motion
analysis of upper extremities. The reason that disqualifies them
from being used as a standard procedure in movement analysis is
at least one of the following limitations: the investigated move-
ment cannot be arbitrary, the movement is 2D, range of motion is
limited, the method cannot be applied for activities of daily living,
movement in one joint is disabled or the arm joint chain is not
free.

Additionally, in contrast to gait, the external forces that are

compensated by the neuromuscular system are less defined and
have lower magnitudes. As a consequence, information about the
forces and torques acting on the joints during upper extremity
movements is often unavailable. Furthermore, the interpretation
of the muscular-coordination pattern depicted by SEMG becomes
complex and sometimes impossible. Human arm dynamics have
been less investigated than the kinematics and the procedures

ARTICLE IN PRESS

Contents lists available at

ScienceDirect

journal homepage:

www.elsevier.com/locate/jbiomech

www.JBiomech.com

Journal of Biomechanics

0021-9290/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:

10.1016/j.jbiomech.2009.03.042



Corresponding author. Tel.: [phone]; fax: [phone].

E-mail address:

[e-mail] (N. Popovic).

Journal of Biomechanics 42 (2009) 1570–1573

were either task specific where the upper extremity kinetics has
been analysed during crutch-assisted gait (

Requejo et al., 2005

) or

during wheelchair propulsion (

Ensminger et al., 1995

); or without

measured external loads (

Riener and Straube, 1997

).

For these reasons, there is a need for a methodology that not

only improves the reproducibility of upper extremity movements
but also defines and measures the external forces during any
freely definable upper extremity movements.

2. Method

To enhance the reproducibility of upper extremity movement, 6 degrees of

freedom (DoF) KUKA robot-arm (

Fig. 1

) was used to predefine the motion. For the

measurement of the external forces on the robot’s end-effector a 6 DoF force
sensor with a ball-shaped handle had been attached. The subject held this handle
during the movement test. Additionally, a force feedback about the current
external load provided by a display connected to the sensor has been used to
maintain a predefined force vector.

The display acts as a tool, which allows the definition of a target force in all

degrees of freedom as well as a visualisation of the difference between the target
and applied force vector. This target vector can be either constant or variable
during a movement test. The insert in

Fig. 1

. shows two cases which may be

displayed.

On the left-side, the applied force should be corrected since the target force

vector is not achieved. The vector resulting from the difference between the two
force vectors is presented on the screen as a black star. The position of the star on
the screen depends on the manipulation of the handle by the subject and
simultaneously indicates the direction in which the applied force vector should be
corrected in order to move it into the target circle. On the right-side of the insert,

the target force vector has been attained. As such the star has become green and is
positioned in the central circle.

The experiments were performed on the dominant arm of eight subjects

(5 male and 2 female) who participated in the study. They were all healthy, ages
22–32, and gave informed consent prior to the experiments.

3. Validation

3.1. Reproducibility of joint angles

For validation of the reproducibility of joint angles, a goal-

directed movement was compared with the same motion guided
by the robot. For this purpose, a relatively complex, three-
dimensional daily activity referred to as ‘Removing a parking
token’ (

Williams et al., 2006

) has been chosen. The subject was

asked to perform three times the sequence of movements
required to remove a parking token from a dispenser at the car-
park, from a seated position in a car. The robot-guided movement
was performed using the preprogrammed 3D motion path, also
with three motion cycles. Both trials were repeated at least a day
after the first measurement. For this movement, all three shoulder
axes and flexion/extension axis in elbow joint are well defined,
while the two hand axes and elbow pronation/supination axis are
left free for subject to choose whether to use them or not. The
joint angles were calculated (

Schmidt et al., 1999

;

Williams et al.,

2006

) for the shoulder joint and flexion/extension axis in elbow

joint for each trial.

The intra-individual reproducibility of the movement was

evaluated using the Pearson product–moment correlation coeffi-
cients between the two independent trials for each rotational axis
of shoulder and flexion/extension of elbow joint.

Table 1

shows

the mean values and standard deviations of the correlation
coefficients obtained from the trials performed by 7 subjects.

The mean values of the correlation coefficients (

Table 1

)

obtained for the robot-guided movement (0.66–0.87) were
significantly higher (p

o0.001) than those for the goal-directed

movement (0.42–0.56). The ranges of the standard deviations of
the mean correlation coefficient were 0.11–0.27 and 0.37–0.45,
respectively.

In order to test the inter-individual variations in joint angles,

the mean values and standard deviation of the second repetition
in both trials have been calculated. The mean values of the
standard deviations from 8 subjects for each measured joint axis
were determined.

Table 2

shows that they were significantly

smaller (p

o 0.036) for the guided movement (7.28–21.781) than

for the goal-directed movement (9.59–27.51).

3.2. Validation of the force feedback

For validation of the force feedback, 8 subjects performed three

repetitions, with and without force feedback, of a shoulder flexion

ARTICLE IN PRESS

Visual Feedback

Right

Wrong

Target circle

Robot arm

Force sensor

Handle

Green star

Black star

Fig. 1. Measurement system: a robot-arm presents a 3D path, 6 DoF force/torque
sensor attached at the end effector and a handle used as a user interface between a
subject and the force/torque sensor. Force feedback helps in maintaining a
predefined force vector. Insert: on the left-side, the applied force should be
corrected (target force vector is not achieved, the star is outside the target circle
and black); on the right-side, target force vector is achieved (the star is in the
target circle and green).

Table 1
Mean values and standard deviations of the correlation coefficients of joint angles
between two trials for the goal-directed and robot-guided task.

Movement

Goal directed

Robot guided

Correlation coefficients (mean value with standard deviation)
Shoulder

Flex/ext

0.56

70.39

0.81

70.22

Abd/add

0.55

70.37

0.87

70.11

Inn/out

0.42

70.45

0.66

70.27

Elbow

Flex/ext

0.52

70.39

0.79

70.24

The flexion/extension (flex/ext), abduction/adduction (abd/add) and inner/outer
rotation (inn/out) axes of the shoulder joint and the flexion/extension (flex/ext)
axis of the elbow joint were considered.

N. Popovic et al. / Journal of Biomechanics 42 (2009) 1570–1573

1571

movement with a target force of 15 N upwards. For the measure-
ment without feedback, the subjects were instructed to try to
generate the same force as in previous measurement with
feedback without information about the currently generated
loads.

Fig. 2

shows the external force measured in the z-axis for

the 8 subjects, the mean values and standard deviations obtained
with (a) and without force feedback (b).

Fig. 2

shows that when using force feedback, the mean values

of the external force fluctuate less and the standard deviation
decreases.

4. Discussion

The results show that both intra- and inter-individual differ-

ences in joint angles decreased using predefined robot paths. In
contrast to goal-directed tasks, the procedure developed allows,
both the preprogramming of the desired test path, which allows
guidance during the complete movement, and also the control of
velocity in each part of the movement. The ability to calculate the
joint angles for the complete joint chain of the arm facilitates not
only monitoring of the functionality of the joint investigated, but
also the analysis of individual movement strategies. It can thus
provide an answer to a number of questions e.g. ‘How is an injury
of one joint compensated for in other joints?’.

Introducing the force sensor with force feedback, a measure-

ment system has been created, which allows the functional
testing of upper extremity movement performance. This includes
the assessment of movement kinematics as well as the measure-
ment of external loads. These data can be further used in
biomechanical models to calculate kinetic data such as net joint
forces and net joint moments.

By utilizing this procedure, it will be possible to more fully

compare reproducible, unconstrained movements of upper ex-
tremities. Therefore, normal and/or patient collectives can be
formed and compared. Additionally, comparisons can be made
between a patient and a normal collective. This information can
be used to establish the movement patterns and compare the
ranges of motion characteristic of different patient groups. In
combination with SEMG, this procedure can be used to illustrate
the muscular-coordination patterns at different contraction levels.

It could be used for stroke patients, patients with plexus lesion

or patients with other upper extremity disorders and injuries. The
main principles would remain the same, but some changes in the
procedure such as tracking task or holding the robot’s handle,
choosing the appropriate robot path and velocity or the appropriate
force level, have to be made to adapt to the patient’s group or age.

Through movement standardization, the ability to compare

data that will be used for evidence-based decision-making or the
evaluation of rehabilitation programs is greatly improved. As such,
this methodology has a direct impact on clinical applications for
patients suffering from upper extremity disorders.

Conflict of interest statement

The authors would like to disclose any financial and personal

relationships with other people or organizations that could
inappropriately influence their work.

Acknowledgment

The authors gratefully acknowledge the financial support

provided by the German Research Council (Deutsche Forschungs-
gemeinschaft DFG) (DI 596/4-1; DI 596/4-2).

References

Bresler, B., Franke, J.P., 1950. The forces and moments of the leg during level

walking. American Society of Mechanical Engineers Transactions 72, 27–36.

Burgar, C.G., Lum, P.S., Shor, P.C., Van der Loos, H.F.M., 2000. Development of robots

for rehabilitation therapy: the Palo Alto VA/Stanford experience. Journal of
Rehabilitation Research and Development 37 (6), 663–674.

Cavagna, G.A., Magaria, R., 1966. Mechanics of walking. Journal of Applied

Physiology 21, 271–278.

Chambers, H.G., Sutherland, D.H., 2002. A practical guide to gait analysis. Journal of

the American Academy of Orthopaedic Surgeons 10 (3), 222–231.

Davis, R.B., O

˜ unpuu, S., Tyburski, D., Gage, J.R., 1991. A gait analysis data collection

and reduction technique. Human Movement Science 10 (5), 575–587.

ARTICLE IN PRESS

Table 2
The mean values of the standard deviations of the second repetition for both tasks
from 8 subjects for each measured joint axis.

Movement

Goal directed

Robot guided

Standard deviation (mean value)
Shoulder

Flex/ext

717.21

714.691

Abd/add

79.591

77.281

Inn/out

721.451

714.161

Elbow

Flex/ext

727.51

721.781

The flexion/extension (flex/ext), abduction/adduction (abd/add) and inner/ outer
rotation (inn/out) axes of the shoulder joint and the flexion/ extension (flex/ext)
axis of the elbow joint were considered.

Fig. 2. Forces in the z-axis for 8 subjects, in addition to the mean values and
standard deviations (a) without and (b) with force feedback.

N. Popovic et al. / Journal of Biomechanics 42 (2009) 1570–1573

1572

Ensminger, G.J., Robertson, R.N., Cooper, R.A., 1995. A model for determining 3-D

upper extremity net joint forces and moments during wheelchair propulsion.
Engineering in Medicine and Biology Society, IEEE 17th Annual Conference,
September 20–23, Montreal, Canada, vol. 2, pp. 1179–1180.

Hogan, N., Krebs, H.I., Sharon, A., Charnnarong, J., 1995. Interactive robotic

therapist. US patent, 5466213.

Kadaba, M.P., Ramakrishna, H.K., Wotten, M.E., 1990. Measurement of lower

extremity kinematics during level walking. Journal of Orthopaedic Research 8
(3), 383–392.

Krebs, H.I., Hogan, N., Aisen, M.L., Volpe, B.T., 1998. Robot-aided neurorehabilita-

tion. IEEE Transactions on Rehabilitation Engineering 6, 75–87.

Nef, T., Mihelj, M., Colombo, G., Kiefer, G., Perndl, C., Muller, R., Riener, R., 2007.

ARMin—Exoskeleton for Arm Therapy in Stroke Patients. In: Proceedings of the
2007 IEEE 10th International Conference on Rehabilitation Robotics, June
12–15, Noordwijk, The Netherlands.

Rau, G., Disselhorst-Klug, C., Schmidt, R., 2000. Movement biomechanics goes

upwards: from the leg to the arm. Journal of Biomechanics 33 (10), 1207–1216.

Requejo, P.S., Wahl, D.P., Bontrager, E.L., Newsam, C.J., Gronley, J.K., Mulroy, S.J.,

Perry, J., 2005. Upper extremity kinetics during Lofstrand crutch-assisted gait.
Medical Engineering & Physics 27, 19–29.

Riener, R., Straube, A., 1997. Inverse dynamics as a tool for motion analysis: arm

tracking movements in cerebellar patients. Journal of Neuroscience Methods
72, 87–96.

Sanchez, R.J., Liu, J., Rao, S., Shah, P., Smith, R., Rahman, T., Cramer, S.C., Bobrow, J.E.,

Reinkensmeyer, D.J., 2006. Automating arm movement training following
severe stroke: Functional exercises with quantitative feedback in a gravity-
reduced environment. IEEE Transactions on Neural Systems and Rehabilitation
Engineering 14 (3).

Schmidt, R., Disselhorst-Klug, C., Silny, J., Rau, G., 1999. A marker-based

measurement procedure for unconstrained wrist and elbow motions. Journal
of Biomechanics 32 (6), 615–621.

Williams, S., Schmidt, R., Disselhorst-Klug, C., Rau, G., 2006. An upper body model

for the kinematical analysis of the joint chain of the human arm. Journal of
Biomechanics 39 (13), 2419–2429.

ARTICLE IN PRESS

N. Popovic et al. / Journal of Biomechanics 42 (2009) 1570–1573

1573