Proceedings of the 2006 IEEE
International Conference on Robotics and Biomimetics
December 17 - 20, 2006, Kunming, China
Research on the Kinematics and Dynamics of a 7-DOF
Arm of Humanoid Robot
Tie-jun ZHAO, Jing YUAN
Shenyang University ofTechnolog
Shenyang 110023, China
tj_zhao@tom.com.cn
Abstract-We have developed a highly flexible
anthropomorphic 7-DOF robotic arm for a mobile humanoid
robot. The kinematics of the arm such as workspace,
singularity, the number and physical nature of self-motion are
presented. The concepts and methodology of the inverse
kinematics base on the self -motion of the arm are described.
By this method the task and motion scheduling can simply be
done. The formulations of dynamics and the dynamic effect
due to the arm's dead weight were analyzed.
Keyword: Kinematics, Dynamics, Humanoid Robot, Arm
I. INTRODUCTION
In contrast with the industrial robots, the humanoid
robot must be able to cope with the wide variety of tasks
and objects encountered in dynamic unstructured
environments. So the arm of the humanoid robot should
be mobility, flexibility and anthropomorphic for better
adaptation to typical human environments and to allow
for human-like behavioral strategies in solving complex
tasks. For a human arm, it could be not only organize its
kinematics and dynamic in many different ways to be
appreciate for different tasks, but also exploit the
complex kinematics and dynamic in variety of ways in
dynamic unstructured environments. For example, we
often brace our arm on a hard surface to isolate excess
degree of freedom when performing a delicate task. We
always exploit our skeleton and tendon rather than our
muscle to strength our arm when we carrying loads. On
the other hand, the skeleton, tendon and muscle construct
a fairly complex mechanism in common to be a typical
super-redundancy system that has high flexibility. In
order to achieve a high degree of flexibility and to allow
the simple and direct cooperation with humans, the
humanoid robot arms should be anthropomorphic 7 DOF
redundant robots. There are several successful examples
of humanoid robots with two 7 DOF arms such as the
WABIAN biped robots and WENDY mobile robot of the
Waseda University in Japan[1]-[4], the Arnold mobile
robot of the Ruhr University Bochum, the HERMES
1-4244-0571-8/06/$20.00 C)2006 IEEE
Ming-yang ZHAO, Da-long TAN
Shenyang Institute ofAutomation of
the Chinese Academy of Science
Shenyang 110016, China
mobile robot of Bundeswehr University and the ARMAR
mobile robot of Karsruhe University in Germany[5]-[7],
the DAV a mobile humanoid robot of Michigan State
University in America, although the manipulation
capabilities and intelligence of these robots are still far
away from the human ability in solving complex
manipulation tasks[8]. The mechanical configuration of
the WABIAN, the WENDY, the HERMES and the
ANOLD 7DOF arm is: three in the shoulder, one at the
elbow, and three at the wrist. For the mechanical
construct, they all were adopted the method of a direct
driven mechanism with DC motors and harmonic drive
to realize high-power work, lightness of weight and
compaction. The newest WABIAN: WABIAN-R has a
completely humanoid figure with two arms, and it has
capabilities of dynamic dance waving arms and dynamic
carrying of a load. WENDY's arm Performances have
been evaluated by two requirements such as safety and
dexterity. It can fulfill tasks such as picking up objects,
breaking egg and cutting vegetable[6]. HERMES is
equipped with two arms, a bendable body on an
omni-directional wheelbase. Its work space can extend
up to 120 cm in front of it when it bends forward 1300,
and keep its balance even when the body and the arms
are fully extended to the front. The mechanical
configuration of the AMAR 7DOF arm is: three in the
shoulder, two at the elbow, and two at the wrist. The
weight of the arm is no more than 7kg because in each of
joints the power transmission is done via planet gears
and ropes. With a telescopic joint in its body the total
height of the robot can be adapted to enlarge the work
space. Arms of these humanoid robots have many
actuators, electron-magnetic brakes and sensors
comparison with conventional manipulators.
The author developed an autonomous mobile humanoid
robot YIREN (Fig.1) that is an experimental platform of
researching on implemental technology of humanoid robots.
In 2000, we began the research and development of the
humanoid robot. Our goal is to design, analyze and
implement a flexible and autonomous mobile robot with
handling skills for non-restricted environments. The robot is
a 23 DOF humanoid form robot with a 3 DOF mobile
platform, two 7 DOF redundancy Arms, two 1 DOF hands,
a 2 DOF waist and a 2 DOF neck, a simple gripper and a
head equipped with a stereo camera system. Specifications
of the arm such as workspace, angular velocity and angular
1 553
acceleration are determined with consideration of abilities of mechanism under the criterion of manipulability w and
of human being. The robot has been an anthropomorphic the theory of 3DOF planar manipulator [9].
model to be both easy and natural for humans to interact According to the theory of 3DOF planar manipulator,
with in a human like way. The upper body is fixed on a the limbs of humans can be modeled as mechanisms with
mobile robot with orthogonal wheel assemblies. The three moving links. The 3DOF planar mechanism is
maximum acceleration of the endpoint is about 10[m/sec2] fundamental to anthropomorphic arm, so we can adopt the
evaluation criteria of the 3DOF planar mechanism to
analyze the arm operational performance. The theory of
3DOF planar manipulator presents that: Suppose the three
link lengths are ri (i =1, 2, 3), the r1, r2, r3 are the link
lengths of upper arm, forearm and hand, respectively. If r1
r2= 1.0-1.2, r3 . r1 < 0.5, the dextrous workspace of the
mechanism is large, and the velocities, the payload, the
deformation of the mechanism are suitable. The maximum
length of the YIREN arm is about 1 m (350mm ofthe upper
arm, 350mm of the forearm and 180 of the hand) with a
maximum 2 kg of load.
Figure 2(a) shows the mechanism configuration of the
ARMRA arm. Figure 2(b) shows the mechanism
configuration of the WABIAN and the ARNOLD arm.
Figure 2(c) shows the mechanism configuration of the
YIREN arm.
III. THE KINEMATICS PROPERTIES OF THE ARM
A. The singularities.
TABLE 1
1r1g. i i ne llUmlanllUlU ruuDt Y ien
Denavit-Hartenberg parameters of the arm
and the maximum velocity of the endpoint is 1.0[m/sec]. joint as-1 a i-1 di 61 range
The mechanical system concept, the kinematics and i (mm) (mm)
dynamics properties of the arm are described. Differences SI 0 0 0 0 - 30° 150°
of the kinematics and dynamics properties from two other S2 0 -900 180 0 2+900 -90° 90°
arms of ARMAR humanoid robot and WABIAN humanoid S3 0 -900 0 0 3+900 -90 90°
robot were analyzed. E 300 -90o 0 04-900 0-150°
WI 0 -900 300 05 -90° 90°
II. DETERMINATION OF THE ARM CONFIGURATION W2 0 900 0 06 -90° 90°
From the mechanical point of view, the skeleton of W3 0 -900 0 07 -170° 170°
human's arm is composed of 7DOF mechanism while
others degree of freedom of the arm performing by tendon. Table 1 shows the reference and link coordinate systems
We developed a 7DOF anthropomorphic arm by synthesis of the 7-DOFarm using the Denavit-Hartenberg convention.
The letters S, E and W are the joints of shoulder, elbow and
wrist, and the S1S3, S3E, EW2 are the link lengths of
shoulder, upper arm and forearm, respectively.
Fig. 3. Coordinate system of the arm

I z4 Izy
I67
5,7, e

xo- 4 1W y
y
Wi5W 7

Si

According to the coordinate systems of the arm (Fig.3),

the description of the coordinate transformation between
(a) (b) (c) fame i and i-i is given by the homogeneous transformation
Fig.2: Mechanism configuration of arms
matrix:

1554
 mechanism configuration of the YIREN arm. This curve
co, - caso, sabso, aicoi could be derived from the fact that the end point of the
i-sSi cao,iS - saicO, aisO, ( upper arm describes an ellipsoid center on the SI shoulder
d joint and that the starting point of the forearm describes a
O sai cai sphere center on the wrist. Since these two points are the
o o 0 1 same, the curve results from the intersection of the ellipsoid
Where s 9 i and c 9 i are sin 9 i and cos 9 i, respectively. and the sphere. Specially, once the elbow position on this
The position and orientation of the end gripper with curve is defined, an analytical-geometrical method was
respect to the base frame is described in the coordinate provided for a closed form solution of the arm inverse
transformation: kinematics problem. This method not only solves the
problem of approximation using iterative numerical method
T()=7
7F

itl
()=n
= O
o
0 0
a p
I
1L
but also makes the self-motion of elbow contact with the
arm inverse kinematics problem. The ellipsoid is given by:
Where 9 is the vector of the joint variables,
n ,o and a are x x + y
Y + Z =11 ((4)
the vectors of the frame attached to the end gripper, and p is (S1S3 + S3W)2 (S1S3 + S3W)2 (S3W)2
the position vector of the origin of the frame with respect to The curve is a circle resulted from the intersection of the
the origin of the reference frame. ellipsoid and the sphere, and its radius is given by:
The manipulability is defined:
w= det(jJT) (3)
I 7~~2 2 2
The manipulability is a measure of a robot's flexibility. A r |IS3E 2 _ IS3E -IEW
2 3
+ IS3W )2 (5)
large value of the manipulability ensures that the arm ~~~1~ 2 S3W
endpoint is within the area of the dextrous workspace. If
w-O, the arm motion is in a singular configuration.
According to the algorithm for the workspace evaluation, The position of its center is given by:
Figure 4 shows the reachable workspace of the YIREN arm
approximately from that of the WABIAN arm and the
ARMRA arm (show in Fig. 10-12) . From the value ofthe I 2 _
I 2+ I
2W 2
2
manipulability, we can draw a conclusion that singularities S3E
(6)
mostly occur at the boundary of the workspace. In contrast, 21S3W

there are many singularities in the workspace of others.

1 000
If the position of the wrist is expressed in spherical
coordinates:
500 I w = (ow, (oIW Irw O(7
The position of the elbow can be written as:
0

SIE (R . RYSW Re a ex) * R +ro + ISlS3 (8)

-500

Where, R y, R z are the rotation matrix around yo, zo axis in

-1000
-1000 -500 0 500 1000 the o-xoyo zo coordinate system. a is the redundancy angle
x
Fig. 4. The reachable workspace of the YIREN arms of the elbow. e x is the basis vector in xo axis. R is the radius
of the sphere whose center is S3.
B. The flexibility and inverse kinematics.
Redundant arm could be in infinite distinct configurations
with the same hand position and orientation. This
movement of redundant arms joints that does not cause any F
hand motion is referred to as its self-motion. The
self-motion of redundant arm provides the capability for S1
criteria optimization, obstacle and singularity avoidance and
this is especially important for applications in hazardous
environments. The self-motion of YIREN arm is defined as
the trajectory describable by the rotation of the elbow point
about the shoulder-wrist line while the position and
orientation of hand remain fixed. It is because the feasible Fig. 5. Self-motion angle of the arm
positions of the elbow could be defined by a curve in the

1555
 The self-motion angle a can be arbitrarily chosen
within the interval [0, 2 w ] in the condition of the absence figure 1 (a): r (IS2 +IEW) (17)
of any constrains. When the arm must be avoided the
obstacle, singularity or joint limit on the determined figurel(b): r (ISE+IW) (18)
trajectory, it usually has to make use of the self-motion. It is figure I (c): r (IS3E +IEW) (19)
quite easy to calculate the remaining range simply by The self-motion range of the configuration (b) is bigger,
subtracting the corresponding angle segment of the and that of the configuration (c) is smaller.
obstacle, singularity or joint limit from the self-motion
circle. Then, we can determine the objective function that
minimizes the joints angle change to choose the value of IV. THE DYNAMICS PROPERTIES OF THE ARM
aO: The arm's dead weight is the main load of motors.
7
Theoretical analysis and computer simulations indicate the
E (0) = L(0i cur -i,cal )2 YIREN arm was provided with better mechanical
i=l (9) characteristics. Euler-Lagrange equations describe the
Where 9i,cur and 9i,cal represent the current and the dynamics of a n-DOF robot:
calculate value of the joint angles, respectively.
Once the a is determined the elbow position (Xe, Yen Ze) r = D(6)6 + h(6,O )O + G(6) + E (20)
is known and the vectors ISE, lEwcan be directly calculated.
Based on a system of shoulder joint S1 coordinates o° - X1, where D is a 7x7 inertia matrix, h is a 7xl centrifugal
and Coriolis matrix, G is a 7xl gravity matrix, E is a load
yi, z1, the joint angle 0 l can be calculated: matrix.
01 = atan2(SlEY,IslEx) (10) The dynamics of 7DOF robot present complex
characteristics of nonlinear and coupling. According to the
In the shoulder joint S2 coordinates 0 2 X 2, Y 2, Z 2, the property of the YIREN arm configuration, the inertia energy
joint angle 0 2, 0 3 can be calculated: of the 3DOF wrist can be simplified as5 =0 6 =07 =0,
62 = atan2(/s3EY,IS3Ex) (11) and merge its mass into forearm. The Euler-Lagrange
equations are simplified as:
63 = a co S(E )
IS3E (12)
T1= D111 + D12 2 + D13 63 + D144 +
In the elbow joint coordinates 0 4 X 4 Y4 Z4, the joint h122 22 + h133 32 + 2h,126162 + 2h,13 103 +
angle 0 4 can be calculated:
64 =a tan 2(UEWY,lIEWX)
2h1146l64 + 2hI23 6263 + 2hI246264 + 2hI346364 +
(13) GI +El
Based on a system of wrist joint coordinates 05 X 5 Y5, (21)
z 5,the joint angle 0 5, 0 6 can be calculated: In order to estimating the dynamics properties of these
three arms configuration, here mainly analyzed the
65 W aWtan2(lwHY,wH) (14) influence of the arm's dead weight. If we define the mass of
T the forearm, the upper arm, the wrist and hand all to be m,
and the velocity and the acceleration are zero, the joints
06 = a cos(tH ) moment overcoming the arm dead weight shows in the
IWH (15) figure 6 to 8. The results of analyzing show the moment of
Since hand vector IG is orthodoxy with vector I WH, we dead weight: the arm configuration of figure 2 (a), figure 2
can evaluate 0 7: (b) is bigger and that of the configuration 2 (c) is smaller.
67 =atan2(lGYIGx) (16)
Having derived the closed form equation for each 0 l to
0 7, we have implemented the inverse kinematics on the
YIREN arm. The performance of the arm may be judged in
terms of various performance criteria such as avoiding
obstacle or avoiding singularity. In additional, if the
distance between the position of the shoulder joint and the
point of the wrist joint is expressed as vector r there are two
kind of the self -motion according to the change of r:
self-motion during the abduction of the upper limb (except
rmax = ISE + IEW because of singularity) self-motion during
the adduction of the upper limb (except rmin = ISE - 'EW Fig.6. Moment of the aim's dead weight
because of the interference of joints) . The results of
analyzing show the arm configuration of: V. CONCLUSION

1556
 We present the Kinematics and Dynamics properties of a
7-DOF Arm of Humanoid Robot. Comparing with arms of
WABIAN humanoid and ARMRA humanoid, the arm
configuration in this paper has many kinematics and
dynamics properties: approximate workspace, a few
singularities within workspace, moderate range of
self-motion, moderate moment of dead weight. We also
show a closed form solution for the inverse kinematics of
the robot arm. In the future work we will concentrate on
fitting our model in the complex control problem.

rig. IU. Ite reaclnaoe workspace o1 tne AKItV1IEA arms

ACKNOWLEDGMENTS
This study is supported by the National Advanced
Technology Research and Development Project (863
project 2001AA422170), and Natural Science Foundation
of Liaoning Province (20032036).
REFERENCES
[1] Jin'ichi Yarnagucku, Daisuke Nishino, Atsuo Takanishi.
Realization of dynamic biped walking varying joint
stiffness using antagonistic driven joints. IEEE ICRA 1998,
2022-2029.
[2]Jin-ichi YAMAGUCHI, Atsuo TAKANISHI, Ichiro
KATO. Development of a Biped Walking Robot
Compensating for Three-Axis Moment by Trunk Motion. In

1557
Proc. of IEEE/RSJ Int. conf on Intelligent Robots and
Systems ( IROS' 93), pp. 561-566, 1993.
[3] Toshio MORITA, Hiroyasu IWATA, Shigeki
SUGANO. Human symbiotic robot design based on
division and unification of functional requirements. ICRA
2000, 2229-2234.
[4] Toshio MORITA, Koji SHIBUYA, Shigeki SUGANO.
Design and control of mobile manipulation system for
human symbiotic humanoid: Hadaly-2. ICRA 1998,
1315-1320
[5]K. Berns, T. Asfour, R. Dillmann. ARMAR: An
Anthropomorphic Arm for Humanoid Service Robot.
ICRA1999 International Conference on Robotics &
Automation, Detroit 1999
[6] Thomas Bergener, Carsten Bruckhoff, Percy Dahm,
Herbert JanBen, Frank Joublin, Rainer Menzner. Arnold: An
Anthropomorphic Robot for Human Environments. Internal
Report 97-12, extended version of SOAVE paper
[7] Rainer Bsichoff, Tamhant Jain. Natural communication
and interaction with humanoid robots. Second ISHR' 1999
[8] J.D.Han, S.Q.Zeng, K.Y Tham,M.Badgero and
J.Y.Weng. Dav Humanoid: A Robot Platform for Mental
Development.
[9] Feng Gao, Fabrice Guy, William A. Gruver. Criteria
based analysis and design of three degree of freedom planar
robotic manipulators. ICRA 1997, 468-473
[10]Hiroyasu IWATA, Hayato HOSHINO. A Physical
Interference Adapting Hardware System Using MIA Arm
and Humanoid Surface Covers. IROS 1999,1216-1221
[11 ]Richard M. Voyles. Terminatorbot: A robot with
dual-use arms for manipulation and locomotion. IEEE
ICRA 2000, 61-66
[12]Manja V. Kircanski, Tatjana M. Petrovic. Inverse
kinematic solution for a 7 dof robot with minimal
computational complexity and singularity avoidance. IEEE
ICRA 1991, 2664-2669
[13]Percy Dahm, Frank Joublin. Closed form solution for
the inverse kinematics of a redundant robot arm. Internal
Report IRIN197-08

1558