Human hip joint center analysis for biomechanical design of a hip joint exoskeleton

Abstract

We propose a new method for the customized design of hip exoskeletons based on the optimization of the humanmachine physical interface to improve user comfort. The approach is based on mechanisms designed to follow the natural trajectories of the human hip as the flexion angle varies during motion. The motions of the hip joint center with variation of the flexion angle were measured and the resulting trajectory was modeled. An exoskeleton mechanism capable to follow the hip center’s movement was designed to cover the full motion ranges of flexion and abduction angles, and was adopted in a lower extremity assistive exoskeleton. The resulting design can reduce human-machine interaction forces by 24.1% and 76.0% during hip flexion and abduction, respectively, leading to a more ergonomic and comfortable-to-wear exoskeleton system. The human-exoskeleton model was analyzed to further validate the decrease of the hip joint internal force during hip joint flexion or abduction by applying the resulting design. © 2016, Journal of Zhejiang University Science Editorial Office and Springer-Verlag Berlin Heidelberg.

Full text

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
792
Human hip joint center analysis for biomechanical
design of a hip joint exoskeleton*
Wei YANG, Can-jun YANG†‡, Ting XU
(State Key Laboratory of Fluid Power & Mechatronic Systems, Zhejiang University, Hangzhou 310027, China)
†E-mail: ycj@zju.edu.cn
Received Sept. 4, 2015; Revision accepted Jan. 13, 2016; Crosschecked July 11, 2016
Abstract: We propose a new method for the customized design of hip exoskeletons based on the optimization of the human-
machine physical interface to improve user comfort. The approach is based on mechanisms designed to follow the natural tra-
jectories of the human hip as the flexion angle varies during motion. The motions of the hip joint center with variation of the
flexion angle were measured and the resulting trajectory was modeled. An exoskeleton mechanism capable to follow the hip
center’s movement was designed to cover the full motion ranges of flexion and abduction angles, and was adopted in a lower
extremity assistive exoskeleton. The resulting design can reduce human-machine interaction forces by 24.1% and 76.0% during
hip flexion and abduction, respectively, leading to a more ergonomic and comfortable-to-wear exoskeleton system. The human-
exoskeleton model was analyzed to further validate the decrease of the hip joint internal force during hip joint flexion or abduction
by applying the resulting design.
Key words: Hip joint exoskeleton, Hip joint center, Compatible joint, Human-machine interaction force
http://dx.doi.org/10.1631/FITEE.1500286 CLC number: TP242.6
1 Introduction
With rapid progress in mechatronics and robotics,
anthropomorphic exoskeletons have been widely
studied for rehabilitation applications and for general
walking assistance. Key contributions in these areas
include lower extremity exoskeletons for post-stroke
patient rehabilitation on treadmills (Lopes (Veneman
et al., 2006) and Lokomat (Hidler et al., 2009)),
wearable exoskeletons for paraplegic daily walking
(HAL (Suzuki et al., 2007), Indego (Farris et al.,
2011), and Rewalk (Esquenazi et al., 2012)), and
upper arm exoskeletons for upper body rehabilitation
(Armin-III (Nef et al., 2013) and IntelliArm (Ren et
al., 2013)). Although such exoskeletons can assist or
guide the motions of humans, especially patients,
there is potential for discomfort and injury if the de-
signs are not compatible with human biomechanics
(Wang et al., 2014). Without the ability to fully sense
discomfort, paraplegic or post-stroke patients may
even suffer from serious injuries during repeated
rehabilitation, where comfort is far from ideal when
wearing traditional exoskeletons. To address such
problems, we focus on the lower body and present a
human-biomechanics-based exoskeleton for provid-
ing support to the hip joint in a natural way. Based on
the human anatomical experimental data, the de-
signed mechanical hip joint center (HJC) can follow
naturally occurring motions as the flexion angle varies.
Traditional exoskeleton designs are often based
on assumptions that bionic joints are simplified to ‘pin
and socket’ or ‘ball and socket’ jointed engineered
designs to reduce kinematic complexity. It is usually
these simplifications that cause the incompatibility of
the exoskeleton’s motions with human movements.
Schiele and van der Helm (2006) improved
Frontiers of Information Technology & Electronic Engineering
www.zju.edu.cn/jzus; engineering.cae.cn; www.springerlink.com
ISSN 2095-9184 (print); ISSN 2095-9230 (online)
E-mail: jzus@zju.edu.cn
‡ Corresponding author
* Project supported by the National Natural Science Foundation of
China (No. 51221004)
ORCID: Can-jun YANG, http://orcid.org/0000-0002-3712-0538
© Zhejiang University and Springer-Verlag Berlin Heidelberg 2016

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802 793
ergonomics in human-machine interaction through
the kinematic design of an upper-arm exoskeleton.
Stienen et al. (2009) decoupled joint rotations and
translations to make self-alignment exoskeleton axes,
and the decoupling approach was applied to the up-
per-limb exoskeleton. Jarrasse and Morel (2012) de-
signed kinematics of fixations between an exoskele-
ton and a human to improve the physical connections.
Cempini et al. (2013) presented a complete analytical
treatment of the problem of misalignment in a robotic
chain for human limb torque assistance. These
aforementioned approaches can effectively improve
the compatibility of an upper-arm exoskeleton.
However, for a lower-limb exoskeleton, a disad-
vantage is mainly the vertical orientation of the seg-
ments. As each exoskeleton segment is connected to
each leg segment, without strong translational cou-
plings to other exoskeleton segments, individual cuffs
may slip due to gravity and cyclical inertial forces that
may irritate participants (Stienen et al., 2009).
Several approaches have been used to realize the
alignment of the hip joint motions of a human wear-
ing an exoskeleton. The most common approach ap-
plied is adding some form of size adjustment mecha-
nism, e.g., HAL-3 (Kawamoto and Sankai, 2005),
Lokomat (Hidler et al., 2009), and ALEX (Banala et
al., 2009), which can help with the alignment of the
hip flexion axis. However, the exoskeleton hip ab-
duction axis cannot be regulated in this way because it
leads to deviation of the hip joint between the motions
of the human and the exoskeleton. Valiente (2005)
designed a quasi-passive parallel leg with a cam and
cam roller mechanism at the upper leg to realize hip
abduction joint alignment. Because of the passive
joint design, the friction caused by the mechanism
results in additional energy consumption for the
wearer. To address this problem, Zoss et al. (2006)
developed the Berkeley Lower Extremity Exoskele-
ton (BLEEX), with its flexion and abduction rotation
axes intersecting at the human HJC, which was fixed
during flexion and abduction. Although these ap-
proaches have contributed to realizing better hip joint
alignment, dynamic motions of HJC based on human
biomechanics have not yet been accommodated for.
We present a design method without additional
passive joints to improve the compatibility of the
exoskeleton hip joint. The alignment of the exoskel-
eton hip joint to the human HJC dynamic motions is
the key point of this method. To achieve this goal, the
three hip joint orthogonal axes, the flexion/extension
axis, the abduction/adduction axis, and the internal/
external rotation axis are split and followed up with
translations of the flexion/extension axis and the
abduction/adduction axis. The aforementioned trans-
lations help create dynamic exoskeleton HJC motions
during thigh movements to provide a full coverage of
human HJC motions. Because neither additional
joints nor power units are used, this method leads to a
simpler exoskeleton mechanism. The method we
present provides a convincing alternative for exo-
skeleton mechanical design aiming at joint alignment.
The only challenge is that the human HJC motions
need to be understood well.
Thus, an understanding and quantification of
anatomical joint center motions is necessary for de-
signing exoskeleton joints. The HJC is focused upon
in this study. In a pelvic anatomical coordinate system,
the motions of the HJC have been estimated previ-
ously using a functional method applied by calculat-
ing the center of the best sphere described by the
trajectory of markers placed on the thigh during sev-
eral trials of hip rotation (Leardini et al., 1999).
However, the accuracy of the functional method is
affected by the hip motion range, and research shows
that the shape of the hip deviates from being spherical
and becomes conchoidal (Greenwald and O’Connor,
1971; Afoke et al., 1984; Menschik, 1997) or as-
pherical (Kang, 2004). Meanwhile, results of tracking
the translations of human hip joints show that the
motions of the femoral head (Zakani et al., 2012)
indicate that the HJC is not a fixed point during thigh
movements, which should be kept in mind when de-
signing exoskeletons to support human motion.
Hence, we design an experimental task including
static and dynamic sections. The static section uses a
functional method to calculate the static HJC and the
distance between the static HJC and markers pasted
on the thigh surface during a limited range of hip
motions. During the dynamic section, the thigh moves
freely in the reachable space and specific optimiza-
tion methods based on the results from the static tests
are used to calculate the dynamic motion of the HJC.
The result of the dynamic section is then used to
guide the design of a more biomechanically compat-
ible exoskeleton hip joint based on the derived me-
chanical HJC model. The validity of the compatible

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
794
exoskeleton hip joint is examined by studying
human-machine interaction and hip joint internal
forces, and the conclusions are presented.
2 Hip joint center experimental task
The experimental task of measuring HJC was
designed to obtain the anatomical motions of HJC
during normal walking. The OptiTrack motion cap-
ture system (Krupicka et al., 2014) was used in
measuring walking activity. As shown in Fig. 1, seven
reflective markers were pasted on the right leg. One
marker was placed on the anterior superior iliac spine
(ASIS) to record any shaking of the pelvis and also
was regarded as the origin of the body coordinate
system. Two markers were placed on the lateral
femoral epicondyle (LFE) and medial femoral epi-
condyle (MFE) to calculate the femoral orientation
based on the International Society of Biomechanics
(ISB) recommendations (Wu et al., 2002). Another
four markers were located on the thigh surface,
grouped as a block to minimize the influences of
human soft tissues (Gao et al., 2007). Six infrared
cameras (V100: R2 (OptiTrack Inc., USA)) were
placed in a semicircle pattern to record the motions of
the seven markers. Three volunteers have participated
in this experimental task and their details are shown in
Table 1.
2.1 Static section
During static section tests, the participants were
asked to lift their right leg slightly in the sagittal plane,
and the flexion angle was limited within 10°. Because
of the minute movements of the femoral head in the
acetabulum, the HJC was assumed to stay fixed dur-
ing the static section tests. Therefore, if the marker
group block stays in the same position and the influ-
ence of human soft tissue is ignored, the distances
between HJC and the markers in the block are invar-
iant. This marker group block method has been shown
to be a reasonable way to minimize errors introduced
by human soft tissues (Gao et al., 2007). Thus, the
distances between the static HJC and the markers can
be calculated using a functional method (Leardini et
al., 1999) that is well known for obtaining the optimal
center of rotation position in human ball-and-socket
joints. Different objective functions of the functional
method were compared and validated by Camomilla
et al. (2006). We used the Spheric-4 (S4) algorithm
for its high precision and repeatability (Gamage and
Lasenby, 2002).
Fig. 2 shows the implementation of the S4 algo-
rithm with one marker on the surface of the thigh, the
global coordinate system (CS), and the body CS. The
global CS O-XYZ is defined by the motion capture
system, and the body CS o-xyz is established
Fig. 2 Human hip coordinate systems
Fig. 1 Reflective marker locations
Table 1 Information of subjects for the experimental task
Subject Gender Age (year) Height (cm)
1 Male 27 169
2 Male 24 170
3 Male 23 173

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802 795
referring to the ISB recommendations (Wu et al.,
2002) with the origin o determined by a marker on the
ASIS. Then the objective function of the S4 algorithm
can be denoted by (Gamage and Lasenby, 2002)
2
22
11
(, ) || || || || ,
MN
mmm
n
mn
f





mr p m r (1)
where rm and m are vectors from the HJC to the
markers and the origin o to the HJC respectively, M
and N are the marker number and sample number
during thigh flexion movements respectively (we
choose M=4 and N=700 in this study), and m
n
p can be
calculated by
.
mom
nnij
ppp (2)
Here om
n
p denotes the position of the four markers of
the block and pij denotes the position of the ASIS
marker, which points from the globe CS origin O to
the body CS origin o.
By minimizing Eq. (1), the HJC position can be
calculated. During the static section tests, only the
motions of the ASIS marker and the four markers of
the block were used.
Differentiating Eq. (1) with respect to rm and m,
we obtain
,A

mB
(3)
where

T
T
11 1 1
111
2,
MN N N
mm m m
nn n n
mn n n
ANNN
  








  
pp p p 
 
32
11 1 1
111
.
MN N N
mmm
nnn
mn n n
NNN
  







  
Bppp
Because the optimal HJC position m is obtained,
the distance between HJC and the four markers of the
block can be obtained by
1
1(),
N
mm
n
n
N


rpm m=1, 2, 3, 4. (4)
Table 2 presents the results of six repeated ex-
periments conducted on the first participant. The
mean distance (AVE) and standard deviation (STD) of
the four markers are also listed. The results show
good data consistency and agree well with research
results provided by Leardini et al. (1999).
2.2 Dynamic section
The participants were asked to start the dynamic
section trials once they had finished the static section
tests. During the dynamic section tests, the hip joints’
arc movements, consisting of flexion and abduction
motions, were selected because normal walking also
comprises hip joint flexion and abduction motions.
The participants performed the arc movements 10
times repeatedly at their self-selected speeds, and
were asked to make the flexion and abduction ranges
as wide as possible. Because rm had been obtained and
was considered to be constant during this section due
to that the marker block was located at the same po-
sition during both sections, the optimal HJC motions
of the dynamic section can be calculated by mini-
mizing the following equation (Yan et al., 2014):
22
1
() || || ( ).
Mmm
n
m
f







mpmr
(5)
During the hip joint arc movements, the flexion
angle could be calculated by the two markers located
on the LFE and MFE according to Eqs. (6) and (7):
FE FE FE LFE MFE
(, ,)( )/2 ,xyz

vv m
(6)


22
flex FE FE FE
arccos ,yxy


(7)
where vLFE and vMFE are positions of markers on the
LFE and MFE under body CS respectively, and θflex is
Table 2 Distances between markers and HJC
Experiment |r1| (cm) |r2| (cm) |r3| (cm) |r4| (cm)
1 27.38 30.30 39.95 39.86
2 27.74 30.41 40.21 40.11
3 26.78 29.50 39.18 39.14
4 27.90 30.50 40.26 40.13
5 27.28 29.80 39.57 39.42
6 28.36 30.85 40.79 40.56
AVE 27.57 30.23 39.99 39.87
STD 0.55 0.49 0.56 0.52
AVE: mean distance; STD: standard deviation. |r1|–|r4| represent
the distances between markers and the HJC

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
796
the angle of hip joint flexion during arc movements.
Fig. 3 shows the spatial anatomical HJC position of
participant 1 during the dynamic section.
Considering the anatomical structure of the par-
ticipants’ hip joints, the HJC position should be at the
same location during repeated hip joint flexion. In
other words, the HJC trajectories of each trial should
be curves with the same starting- and end-point.
However, because of the hip joint’s internal/external
rotation during arc movements when adapting to
flexion and abduction, it was hard for the participant
to maintain a constant internal/external rotation angle
and HJC trajectory between different movement trials.
The results shown in Fig. 3 also confirm this. At the
beginning of the arc movement, when the flexion
angle is 30°, the deviations of frontal, lateral, and
upward directions are quite small. When the flexion
angle decreases, the deviations of all three directions
increase, because the influence of the hip joint
internal/external rotation angle becomes larger. The
HJC’s position stays almost the same while the flex-
ion angle is within 10° in the three directions and then
rises rapidly in the frontal and upward directions, and
falls sharply in the lateral direction. Unlike the center
of rotation movements of the knee joint (Lee and Guo,
2010) and the shoulder joint (Yan et al., 2014), which
are more than 30 mm, the HJC’s position movement is
less than 10 mm, but it is clear that it does not stay still.
This result is in accordance with the findings provided
by Zakani et al. (2012) using surgical navigation
methods. Because the human-exoskeleton system is a
closed chain mechanism, the misalignments of
human-exoskeleton HJC positions would lead to in-
ternal forces exerted onto the participant during the
closed chain mechanism’s motions. Therefore, the
influence of these small misalignments was analyzed
and compared with the misalignment compensation
design of the exoskeleton by experiments.
3 Exoskeleton hip joint design
Traditional exoskeleton hip joints were designed
as ball-and-socket joints, which means that the me-
chanical HJC stays still during walking. Because the
anatomical HJC position has been measured and
found to move actually, exoskeleton hip joints should
be designed to be compatible with the motions of the
HJC trajectory to match with it. To keep the me-
chanical HJC close to the anatomical one, the sagittal,
frontal, transverse, and rotation (SFTR) system was
adopted, which means the joint angles in the sagittal,
frontal, and transverse planes were measured. As
shown in Fig. 4, a traditional three-degree-of-
freedom (3-DOF) joint consisting of the X-, Y-, and
Z-axis represents the axes of abduction/adduction,
internal/external rotation, and flexion/extension, re-
spectively. The interaction point O stays still when the
3-DOF joint rotates. Based on the SFTR system, a
new 3-DOF joint was constructed by translations of
the flexion/extension and abduction/adduction axes
that were described in the Y-X and Y-Z planes re-
spectively using polar coordinates. Both coordinates
considered the Y-axis as the polar axis. Considering
both the complexity of the mechanical design for
internal/external rotation axis translation and the
minor change of the internal/external rotation angle
during normal gait, translation of the internal/external
rotation axis was not selected. According to Fig. 4, the
positions of the new intersection points, O0 and O1,
can be expressed as follows:
0
0
1
1
i
0
i
1
e,
e,
O
O
p
p









(8)
Fig. 3 HJC position during thigh arc movements
(a) HJC motion; (b) Frontal direction; (c) Lateral direction;
(d) Upward direction
Z(mm)
Y(mm)
X(mm)
Z(mm)

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802 797
where ρ0, ρ1 and α0, α1 are translation distances and
angles respectively, with respect to origin O and polar
axis Y. Fig. 4 shows the translations of axes with
which the HJC position C can be expressed as follows:
Im( ),
Re( ),
Im( ),
x
y
z
CA
CB
CB





(9)
where
111
000 0
ii()
11
ii() i
00
ee,
ee Re()e,
A
BA


 



 


 


and θ0 and θ1 are the angles of abduction and flexion,
respectively.
With Eq. (9), the exoskeleton HJC position tra-
jectory can be easily obtained when ρ0, ρ1, α0, and α1
are determined. The root mean square (RMS) value of
the distance between the anatomical HJC and me-
chanical HJC was used as a criterion, with which the
four design parameters could be obtained. As shown
in Eq. (10), the four design parameters are considered
to be optimal when E achieves its minimum value:
222
1
1()()(),
ii i i ii
N
xx yy zz
i
EOCOCOC
N





(10)
where (,,)
iii
x
yz
OOO and (, , )
iii
x
yz
CCC are the ana-
tomical and mechanical HJC positions respectively,
during the same hip joint arc movements. Applying
the steepest descent method (Fletcher and Powell,
1963), the minimum value of E and the corre-
sponding translation parameters ρ0, ρ1, α0, and α1 can
be obtained. Table 3 shows the minimum values and
translation parameters of these three participants.
Fig. 5 shows the optimal mechanical HJC sphere
and anatomical HJC based on participant 1. The
origin here denotes the human initial HJC. The ana-
tomical HJC matches well with the mechanical HJC
sphere when the flexion angle is within 0°–30°, while
the deviation is enlarged when the flexion angle is
within −20°–0°. However, if the mechanical HJC
Table 3 Optimal parameters for exoskeleton hip joint
alignment
Subject
E
(mm) ρ0 (mm) ρ1 (mm) α0 (°) α1 (°)
1 1.08 11.5 4.1 278.2 243.1
2 1.52 18.6 5.2 281.4 289.3
3 1.24 19.7 4.0 261.8 296.9
ρ0, ρ1, α0, and α1 are axis translation parameters
Fig. 5 Optimal mechanical HJC sphere and anatomical
HJC
−6
−4
−20
0
2
4
6
8
−1.5
−1.0
−0.5
0
X(mm)
Y(mm)
Z (mm)
Mechanical HJC sphere
Anatomical HJC
Fig. 4 Axis translations and the corresponding HJC
position
(a) Translate abduction/adduction axis (X) and flexion
/
extension axis (Z); (b) Turning around the new flexion
/
extension axis (Z'); (c) Turning around the new abduction
/
adduction axis (X'); (d) New rotation center C
O
Y
Z
X
Z'
X'
Flexion/
Extension
Abduction/
Adduction
O
Y
Z
X
A
Im(A)
Re(A)
O
Y
Z
X
Im(A)
Re(A)
B
O
Y
Z
X
Im(A)
Re(A)
B
C
Re(B)
Im(B)
(a) (b)
(c) (d)
O1
ρ1
α1
α0
ρ0
O0
O1
Z'
X'
O0
θ1
O1
Z'
X'
O0
θ0
Z'
X'
O0
O1
Re(B)
Im(B)

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
798
stays still at the origin, the deviation will be much
larger. Considering the simplicity of the biocompati-
ble joint, these results show that adopting such an
approach in the design of exoskeletons could be quite
beneficial in enhancing the comfort of wearers.
To realize a compatible hip joint mechanism, the
optimal results were applied to translate both the
abduction/adduction axis and the flexion/extension
axis (Figs. 6a and 6b). The traditional exoskeleton
HJC is the intersection of the abduction/adduction
axis and the flexion/extension axis, which means that
the HJC stays still during hip joint motions. Transla-
tion of both axes made them lie in different surfaces.
Fig. 6d shows the compatible hip joint model and the
traditional hip joint model as well for comparison.
Applying axis translation vectors V1, V2, V3, and V4,
the axis translation mechanism is acquired, which
helps make both the abduction/adduction axis and the
flexion/extension axis lie in the desired surfaces. In
Fig. 6f, the 3D printed mechanism for axis transla-
tions makes the HJC move along the optimized me-
chanical HJC sphere during hip joint motions. The
translation vectors V1, V2, V3, and V4 are also shown
with yellow arrows for better understanding of axis
translations.
The translation vectors V1, V2, V3, and V4 can be
expressed as follows:
V=ρα, (11)
where V=[V1, V2, V3, V4]T, ρ=diag(ρ0, ρ0, ρ1, ρ1), and
α=[sinα0, cosα0, sinα1, cosα1]T.
Considering the diversity of the various partici-
pants’ skeletal parameters, this method uses experi-
mental data from one participant to translate the ex-
oskeleton hip abduction/adduction axis and flexion/
extension axis. This indicates that the resultant
mechanism is individually suitable for the participant
who provides the experimental data. However, the
axis translation parameters of each participant can be
acquired and calculated, and the human-exoskeleton
HJC alignment can then be realized by adjustment of
the aforementioned 3D printed mechanism according
to the translation parameters. These three participants’
HJC motions were acquired through the static and
dynamic sections. Each experimental result leads to
independent exoskeleton HJC axis translation pa-
rameters (Table 3).
For a correct alignment of the exoskeleton joints
to human joints, the human ASIS point was selected
as a reference point. Because the vector from ASIS to
the human initial HJC point m had been calculated,
the relative position between ASIS and exoskeleton
initial HJC point was made explicit. Hence, the exo-
skeleton joint could be aligned to the human joint
based on this relative position.
4 Experimental results and discussion
An anthropomorphic lower extremity exoskele-
ton with biocompatible hip joints was designed and
manufactured by implementing the optimal transla-
tion parameters. The exoskeleton hip and knee
flexion/extension joints were driven by flat motors
Fig. 6 Traditional and compatible hip joint mechanism
design
(a) Sketch of traditional exoskeleton HJC; (b) Sketch of
compatible exoskeleton HJC; (c) Traditional hip joint
model; (d) Compatible hip joint model; (e) Traditional hip
joint mechanism; (f) Compatible hip joint mechanism.
References to color refer to the online version of this figure

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802 799
(Maxon Inc., Sachseln, Switzerland) with harmonic
gearboxes (CTKM Inc., Beijing, China). Fig. 7a
shows the exoskeleton structure worn by a patient.
The participant’s ASIS was used as the reference
point for the exoskeleton hip joint to guarantee an
accurate HJC alignment. The comfort of the human-
exoskeleton physical interface is mostly evaluated by
the interaction forces between the human and the
exoskeleton (Lenzi et al., 2011). This misalignment
between the human and the exoskeleton HJC gives
rise to an interaction force, which presses onto the
human soft tissues and reduces the wearing comfort.
Furthermore, the internal forces of the hip joint
caused by additional human-machine forces can
cause injury to the femoral head and the acetabulum.
Therefore, to assess the comfort quality of the re-
sulting exoskeleton, these physical interaction forces
at the hip joint were compared with the traditional
design. The interaction forces were measured by
packaged force sensors consisting of two one-
dimensional force sensors (Tecsis Inc., Offenbach,
Germany) as shown in Fig. 7b.
4.1 Human-exoskeleton interaction tests
The test volunteers were asked to participate in
the interaction force experiments, which were ap-
proved by the Institutional Review Board of Zhejiang
University. Informed consent was obtained from each
participant. The hip flexion and abduction move-
ments were repeated by the participant with the exo-
skeleton five times. The constraint conditions for the
experiment were: (1) the exoskeleton hip joint flexion
speed was set at 15°/s and the abduction speed was set
at 10°/s; (2) the exoskeleton flexion/extension range
was −20°–30° and the abduction/adduction range was
0°–30°. Fig. 8 shows the mean interaction force
between participant 1 and the exoskeleton during
flexion and abduction movements driven by the ex-
oskeleton. Feθ compatible and Fer compatible mean
the normal and tangential interaction forces with
respect to the connecting surface when wearing the
exoskeleton with the compatible hip joint respectively,
while Feθ traditional and Fer traditional refer to the
normal and tangential interaction forces with respect
to the contact surface when wearing the traditional hip
jointed exoskeleton respectively. Both normal and
tangential forces with compatible joints decrease
during flexion movements. However, only the tan-
gential force with compatible joint decreases during
abduction movements. Table 4 shows the averaged
interaction force reduction during flexion and
Fig. 7 Exoskeleton system
(a) Exoskeleton structure; (b) Interaction force test-bed
Abduction/
Adduction
Flexion/
Extension
HJC
Force
sensor
package
(a) (b)
Table 4 Interaction force reduction during flexion
and abduction
Subject Flexion (%) Abduction (%)
1 25.5 85.5
2 22.1 63.1
3 24.8 79.4
Fig. 8 Interaction force during flexion (a) and abduc-
tion (b)
Force (N)
Force (N)

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
800
abduction when the compatible hip joint design
method was applied. The results show the advantage
of the biocompatible hip jointed exoskeleton over the
traditional one. However, the normal forces in the
biocompatible jointed exoskeleton are close to the
forces in the traditional joint during abduction
movements. A reasonable explanation might be that
the abduction speed is slow and the anatomical HJC
movement in the Z direction is not significant, as
shown in Fig. 3.
4.2 Effect of exoskeleton on internal hip joint
force
To study further the influence of the biocom-
patible jointed exoskeleton on the participant, the
internal hip joint force was calculated by applying
kinematic and kinetic analysis of the human-
exoskeleton model (Fig. 9). OH and OE are the HJC
positions of the human and exoskeleton, respectively,
which would move along the trajectories given in
Fig. 9a. E is the connection point where the packaged
force sensors were located.
According to the human-exoskeleton modeling
analysis, the kinematic and kinetic equations can be
obtained, as listed in Eqs. (12) and (13):
0
0
0
sin sin ,
cos cos ,
,
lzh
lyh


 






(12)
2
()sin sincos,
()cos cossin,
hh ere
hhrere
mh h mg F F F
mh h mg F F F



 


  

  



  (13)
where γ is the angle deviation between the partici-
pant’s abduction angle φ and the exoskeleton’s ab-
duction angle θ0, l and h are the distances from the
connection point E to OE and OH respectively, Δy and
Δz are the distances between OE and OH in the Y and Z
directions respectively, Fθ is the hip joint’s internal
force perpendicular to the thigh while Fr is parallel to
the thigh, mh is the mass of human, and g is accelera-
tion of gravity. Fig. 10 shows the results of human-
exoskeleton modeling analysis. The internal force Fr
with the biocompatible joint is lower than the Fr ob-
tained using the traditional jointed exoskeleton during
flexion and abduction motions. Although the reduc-
tion of internal forces in the hip joint is not quite
notable, it will relieve the loads on the femoral head
and acetabulum, which would make sense in consid-
ering the repeated movements during rehabilitation
with the exoskeleton. Moreover, the internal force Fθ
with the biocompatible jointed exoskeleton is similar
to the Fθ obtained using the traditional jointed system,
which is a reasonable result considering that the ap-
plied normal interaction forces are almost the same.
Fig. 9 Human-exoskeleton model during abduction
(a) Kinematic parameters; (b) Kinetic parameters
Z
Y
z
y
lh
E
Z
Y
z
y
E
mg
Exoskeleton
HJC motion
Human HJC
motion
Exoskeleton
leg
Human leg
(a) (b)
Fθ
Fr
OH
OE
OH
OE
γ
φ
θ0
Feθ
Fer
Fig. 10 Hip joint internal force during flexion (a) and
abduction (b)

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802 801
5 Conclusions
To realize biocompatible human-exoskeleton
physical interfaces, a new method was presented and
adopted to design a lower extremity exoskeleton with
compatible hip joints. The compatibility of the new
hip joint was validated by human-machine interaction
force experiments, compared with that using an
exoskeleton with traditional joints. The internal
forces on the hip joint were analyzed and calculated
as further evidence for the superiority of the new hip
joint. The design method can also be adopted as a
reference for hip replacement mechanical design and
other exoskeleton compatible joint design.
The key results of this research are summarized
as follows:
1. The dynamic HJC motions were calculated
based on the functional method and specific optimi-
zation. The results provide evidence for that the hip
joint does not constitute a simple ball-and-socket
mechanism, which is in accordance with previous
research reports in the area.
2. The mechanical hip joint was designed with
its HJC best covering the anatomical one by transla-
tion of the flexion/extension and abduction/adduction
axes under the SFTR system. The RMS error of
matching is low compared with the range of ana-
tomical HJC motions, which is about 10 mm.
3. The human-exoskeleton interaction force ex-
periments show that the average force decreases by
24.1% and 76.0% during hip flexion and abduction,
respectively, when applying the new design method.
Meanwhile, the hip joint’s internal force reduction
validates the compatibility of the new hip joint exo-
skeleton. Because neither redundant joints nor com-
plex mechanisms are added, the method presented is
attractive for exoskeleton hip joint design.
In this research, each set of experimental data
was acquired from one participant, because the HJC
varies among different participants owing to diverse
skeleton sizes. Three volunteers participated in the
experiments. Therefore, participants with a wide
range of anthropometric dimensions need to be ex-
amined, and the influence of different sizes on HJC
motion should be studied statistically. Additionally,
the adopted RMS criterion provides a global optimi-
zation that avoids large deviations. The normal-gait-
data-based criteria would be a better determination
for mechanical HJC, because the exoskeleton is used
for walking assistance. Finally, performance evalua-
tion at various walking speeds and various ranges of
lower leg motions need to be studied further. All of
these items will be studied in the next step in our
research.
Acknowledgements
Many thanks to Qian-xiao WEI and Yi-bing ZHAO for
assisting with the experiments, and to Prof. Gurvinder Singh
VIRK and Dr. Jan VENEMAN for aiding in the writing of the
paper.
References
Afoke, N.Y., Byers, P.D., Hutton, W.C., 1984. The incon-
gruous hip joint: a loading study. Ann. Rheum. Dis.,
43(2):295-301. http://dx.doi.org/10.1136/ard.43.2.295
Banala, S.K., Kim, S.H., Agrawal, S.K., et al., 2009. Robot
assisted gait training with active leg exoskeleton (ALEX).
IEEE Trans. Neur. Syst. Rehabil. Eng., 17(1):2-8.
http://dx.doi.org/10.1109/tnsre.2008.2008280
Camomilla, V., Cereatti, A., Vannozzi, G., et al., 2006. An
optimized protocol for hip joint centre determination us-
ing the functional method. J. Biomech., 39(6):1096-1106.
http://dx.doi.org/10.1016/j.jbiomech.2005.02.008
Cempini, M., de Rossi, S.M.M., Lenzi, T., et al., 2013.
Self-alignment mechanisms for assistive wearable robots:
a kinetostatic compatibility method. IEEE Trans. Robot.,
29(1):236-250.
http://dx.doi.org/10.1109/TRO.2012.2226381
Esquenazi, A., Talaty, M., Packel, A., et al., 2012. The Re-
Walk powered exoskeleton to restore ambulatory func-
tion to individuals with thoracic-level motor-complete
spinal cord injury. Am. J. Phys. Med. Rehabil.,
91(11):911-921.
http://dx.doi.org/10.1097/PHM.0b013e318269d9a3
Farris, R.J., Quintero, H.A., Goldfarb, M., 2011. Preliminary
evaluation of a powered lower limb orthosis to aid
walking in paraplegic individuals. IEEE Trans. Neur. Syst.
Rehabil. Eng., 19(6):652-659.
http://dx.doi.org/10.1109/tnsre.2011.2163083
Fletcher, R., Powell, M.J., 1963. A rapidly convergent descent
method for minimization. Comput. J., 6(2):163-168.
http://dx.doi.org/10.1093/comjnl/6.2.163
Gamage, S.S.H.U., Lasenby, J., 2002. New least squares so-
lutions for estimating the average centre of rotation and
the axis of rotation. J. Biomech., 35(1):87-93.
http://dx.doi.org/10.1016/S0021-9290(01)00160-9
Gao, B., Conrad, B.P., Zheng, N., 2007. Comparison of skin
error reduction techniques for skeletal motion analysis. J.
Biomech., 40(s2):S551.
http://dx.doi.org/10.1016/S0021-9290(07)70541-9
Greenwald, A.S., O’Connor, J.J., 1971. The transmission of
load through the human hip joint. J. Biomech.,

Yang et al. / Front Inform Technol Electron Eng 2016 17(8):792-802
802
4(6):507-528.
http://dx.doi.org/10.1016/0021-9290(71)90041-8
Hidler, J., Nichols, D., Pelliccio, M., et al., 2009. Multicenter
randomized clinical trial evaluating the effectiveness of
the Lokomat in subacute stroke. Neurorehabil. Neur.
Repair., 23(1):5-13.
http://dx.doi.org/10.1177/1545968308326632
Jarrasse, N., Morel, G., 2012. Connecting a human limb to an
exoskeleton. IEEE Trans. Robot., 28(3):697-709.
http://dx.doi.org/10.1109/TRO.2011.2178151
Kang, M.J., 2004. Hip joint center location by fitting conchoid
shape to the acetabular rim region of MR images. Proc.
26th Annual Int. Conf. of the IEEE. p.4477-4480.
http://dx.doi.org/10.1109/iembs.2004.1404244
Kawamoto, H., Sankai, Y., 2005. Power assist method based
on phase sequence and muscle force condition for HAL.
Adv. Robot., 19(7):717-734.
http://dx.doi.org/10.1163/1568553054455103
Krupicka, R., Szabo, Z., Viteckova, S., et al., 2014. Motion
capture system for finger movement measurement in
parkinson disease. Radioengineering, 23(2):659-664.
Leardini, A., Cappozzo, A., Catani, F., et al., 1999. Validation
of a functional method for the estimation of hip joint
centre location. J. Biomech., 32(1):99-103.
http://dx.doi.org/10.1016/S0021-9290(98)00148-1
Lee, K.M., Guo, J., 2010. Kinematic and dynamic analysis of
an anatomically based knee joint. J. Biomech.,
43(7):1231-1236.
http://dx.doi.org/10.1016/j.jbiomech.2010.02.001
Lenzi, T., Vitiello, N., de Rossi, S.M.M., et al., 2011. Meas-
uring human–robot interaction on wearable robots: a dis-
tributed approach. Mechatronics, 21(6):1123-1131.
http://dx.doi.org/10.1016/j.mechatronics.2011.04.003
Menschik, F., 1997. The hip joint as a conchoid shape. J.
Biomech., 30(9):971-973.
http://dx.doi.org/10.1016/S0021-9290(97)00051-1
Nef, T., Riener, R., Müri, R., et al., 2013. Comfort of two
shoulder actuation mechanisms for arm therapy exoskel-
etons: a comparative study in healthy subjects. Med. Biol.
Eng. Comput., 51(7):781-789.
http://dx.doi.org/10.1007/s11517-013-1047-4
Ren, Y.P., Kang, S.H., Park, H.S., et al., 2013. Developing a
multi-joint upper limb exoskeleton robot for diagnosis,
therapy, and outcome evaluation in neurorehabilitation.
IEEE Trans. Neur. Syst. Rehabil. Eng., 21(3):490-499.
http://dx.doi.org/10.1109/tnsre.2012.2225073
Schiele, A., van der Helm, F.C.T., 2006. Kinematic design to
improve ergonomics in human machine interaction. IEEE
Trans. Neur. Syst. Rehabil. Eng., 14(4):456-469.
http://dx.doi.org/10.1109/TNSRE.2006.881565
Stienen, A.H.A., Hekman, E.E.G., van der Helm, F.C.T., et al.,
2009. Self-aligning exoskeleton axes through decoupling
of joint rotations and translations. IEEE Trans. Robot.,
25(3):628-633.
http://dx.doi.org/10.1109/TRO.2009.2019147
Suzuki, K., Mito, G., Kawamoto, H., et al., 2007.
Intention-based walking support for paraplegia patients
with Robot Suit HAL. Adv. Robot., 21(12):1441-1469.
Valiente, A., 2005. Design of a Quasi-Passive Parallel Leg
Exoskeleton to Augment Load Carrying for Walking. MS
Thesis, Massachusetts Institute of Technology, Boston,
USA.
Veneman, J.F., Ekkelenkamp, R., Kruidhof, R., et al., 2006. A
series elastic- and bowden-cable-based actuation system
for use as torque actuator in exoskeleton-type robots. Int.
J. Robot. Res., 25(3):261-281.
http://dx.doi.org/10.1177/0278364906063829
Wang, D., Lee, K.M., Guo, J., et al., 2014. Adaptive knee joint
exoskeleton based on biological geometries. IEEE/ASME
Trans. Mech., 19(4):1268-1278.
http://dx.doi.org/10.1109/TMECH.2013.2278207
Wu, G., Siegler, S., Allard, P., et al., 2002. ISB recommenda-
tion on definitions of joint coordinate system of various
joints for the reporting of human joint motion—part I:
ankle, hip, and spine. J. Biomech., 35(4):543-548.
http://dx.doi.org/10.1016/S0021-9290(01)00222-6
Yan, H., Yang, C., Zhang, Y., et al., 2014. Design and valida-
tion of a compatible 3-degrees of freedom shoulder exo-
skeleton with an adaptive center of rotation. J. Mech. Des.,
136(7):071006.
http://dx.doi.org/10.1115/1.4027284
Zakani, S., Smith, E.J., Kunz, M., et al., 2012. Tracking
translations in the human hip. ASME Int. Mechanical
Engineering Congress and Exposition, p.109-115.
http://dx.doi.org/10.1115/IMECE2012-87882
Zoss, A.B., Kazerooni, H., Chu, A., 2006. Biomechanical
design of the Berkeley lower extremity exoskeleton
(BLEEX). IEEE/ASME Trans. Mech., 11(2):128-138.
http://dx.doi.org/10.1109/TMECH.2006.871087

doi:10.1631/FITEE.1500286
题目：基于人体髋关节转动中心分析的髋关节外骨骼仿生设计
概要：为了改善外骨骼穿戴舒适性，本文提出了一种基于人机物理交互优化的外骨骼设计方法。该方法通
过设计外骨骼髋关节，使其保证人体髋关节运动时外骨骼髋关节转动中心能跟随人体髋关节转动中
心的运动轨迹。当人体髋关节运动时，通过实验测量和计算可以得到其转动中心轨迹。本文设计的
外骨骼髋关节运动机构能在人体髋关节屈曲/伸展和外展/内收时，保证转动中心都能够包容人体髋关
节转动中心运动范围。同时，所设计的外骨骼髋关节被应用到下肢步行康复训练外骨骼中。通过人
机接触力实验可知，与传统设计外骨骼髋关节进行相比，本文设计的仿生髋关节外骨骼在髋关节屈
曲/伸展和内收/外展时分别可以减小 24.1%和76.0%的人机接触力。这一结果证明仿生设计髋关节外
骨骼更具穿戴舒适性，更符合人机工程学的设计要求。最后，本文通过建立人机闭式链模型进一步
分析了仿生设计对于人体髋关节内力的影响，并验证该设计能减少关节内力作用。
关键词：髋关节外骨骼；髋关节中心；柔顺关节；人机交互力