International Conference on Intelligent Control and Information Processing August 13-15, 2010 - Dalian, China

Design of a Redundant Arm for Yiren Humanoid Robot

Tiejun Zhao
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. WENDYs 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 130, 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. Our goal that we began the research and development of the humanoid robot is to design, analyze and implement a flexible and autonomous mobile robot with handling skills for non-restricted environments. Conceptual design for Yiren are as follows: Enough DOFs, range, torque and speed of joints to support mobility, manipulability. We want to organize the basic functions of humans and try to realized these on a humanoid platform. The platform should have dexterous arms to support corresponding intelligence developing, and it has a non-holonomic drive base to develop its autonomous navigation ability. Modular structure for the ease of maintenance and enhancement. The controller of each joint is implemented by exactly same methods. To realize humanoid robot using electrical actuators, 23 DC servo motors integrate with encoders. Sensors are necessary to build up its safe manipulating and mobiliting experiences. There are three gyrosensors, three accelrration sensors and two force/torque sensors to be used in the humanoid robot. Two color cameras are integrated into head and two microcameras are are integrated into the end of arms. Computer architecture in Yiren consist of several embedded PC, DSP , counter, D/A and CAN-Bus(Fig.2). The PCs like PC/104 are connected via ethernet among each other by CAN-Bus system. By using the DSP,the servo control of

AbstractWe have developed a highly flexible anthropomorphic 7-DOF robotic arm for a mobile humanoid robot. This paper presents the kinematics of the arm such as workspace, singularity, the number and physical nature of self-motion. 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 arms dead weight were analyzed.

I. INTRODUCTION 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 mobile robot of Bundeswehr University and the ARMAR mobile robot of Karsruhe University in Germany[5]-[7], the DAVa 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. In the mechatronic construct, they all were adopted the method of a direct driven
N This study is supported by the National Advanced Technology Research and Development Project (863 project 2001AA422170), and Natural Science Foundation of Liaoning Province (20032036). Zhao Tie-jun is with Shenyang University of Technology; e-mail: [email protected]

978-1-4244-7050-1/10/$26.00 c 2010 IEEE

597

the motors is doneby monitoring the revolution speed of the motor, the angle of the axis and the torque applied to the axis.

The mechanical system concept, the kinematics and dynamics properties of the arm are described in this paper. Differences of the kinematics and dynamics properties from two other arms of ARMAR humanoid robot and WABIAN humanoid robot were analyzed. II. DETERMINATION OF THE ARM CONFIGURATION From the mechanical point of view, the skeleton of humans arm is composed of 7DOF mechanism while others degree of freedom of the arm performing by tendon. We developed a 7DOF anthropomorphic arm by synthesis of mechanism under the criterion of manipulability w and the theory of 3DOF planar manipulator [9]. According to the theory of 3DOF planar manipulator, the limbs of humans can be modeled as mechanisms with three moving links. The 3DOF planar mechanism is 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 = 1.01.2, 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 of the upper arm, 350mm of the forearm and 180 of the hand) with a maximum 2 kg of load.
S1 S2 S1 S2 S1 S2 S3

Fig.1. The humanoid robot YiRen

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 acceleration are determined with consideration of abilities of human being. The robot has been an anthropomorphic model to be both easy and natural for humans to interact with in a human like way. The upper body is fixed on a mobile robot with orthogonal wheel assemblies. The maximum acceleration of the endpoint is about 10[m/sec2] and the maximum velocity of the endpoint is 1.0[m/sec]. Encoder Crystal oscillator

E1

E1

E2

E2

E W1

W1

W1

Counter

P2

W2 W3

W2 W3

W2 W3

MCU
CAN controller A/D SIO PWM Servo Amp D/A CAN driver

(a)

(b)

(c)

Fig.3. Mechanism configuration of arms

Fig.3. (a) shows the mechanism configuration of the ARMRA arm. Fig. 3. (b) shows the mechanism configuration of the WABIAN and the ARNOLD arm. Figure 3(c) shows the mechanism configuration of the YIREN arm. III. THE KINEMATICS PROPERTIES OF THE ARM A. The Singularities Table 1 shows the reference and link coordinate systems 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.

CAN
Fig.2. Controller structure

598

z0,1 y x 0,1

y 3

z6 y 2 z4 x,z 3 2 x 4 y 4 x 5,6,7 y 5,7 z5,7,y 6

0,1

z3 x 2

S2 S1

S3

W1

W2

W3

Fig. 4. Coordinate system of the arm

According to the coordinate systems of the arm (Fig.4), the description of the coordinate transformation between fame i and i-1 is given by the homogeneous transformation matrix:

ci s i 1 T = i i 0 0

c i si c i c i s i 0

s i si s i ci c i 0

ai c i ai si di 1

(1)

Where si and ci are sini and cosi, respectively. The position and orientation of the end gripper with respect to the base frame is described in the coordinate transformation:
7 n o a T ( ) = Ti (i ) = i =1 0 0 0

p 1

(2)

Where is the vector of the joint variables, n ,o and a are the vectors of the frame attached to the end gripper, and p is the position vector of the origin of the frame with respect to the origin of the reference frame. The manipulability is defined: (3) The manipulability is a measure of a robots flexibility. A large value of the manipulability ensures that the arm endpoint is within the area of the dextrous workspace. If w=0, the arm motion is in a singular configuration.
1000

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 hand motion is referred to as its self-motion. The self-motion of redundant arm provides the capability for 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 positions of the elbow could be defined by a curve in the mechanism configuration of the YIREN arm. This curve could be derived from the fact that the end point of the upper arm describes an ellipsoid center on the S1 shoulder joint and that the starting point of the forearm describes a sphere center on the wrist. Since these two points are the same, the curve results from the intersection of the ellipsoid and the sphere. Specially, once the elbow position on this curve is defined, an analytical-geometrical method was provided for a closed form solution of the arm inverse kinematics problem. This method not only solves the problem of approximation using iterative numerical method but also makes the self-motion of elbow contact with the arm inverse kinematics problem. The ellipsoid is given by: x y z + + =1 (4) 2 2 ( S1 S3 + S3W ) ( S1 S3 + S3W ) ( S3W ) 2 The curve is a circle resulted from the intersection of the ellipsoid and the sphere, and its radius is given by:

w = det( JJ T )

r = lS 3 E 2 (

lS 3 E 2 lEW 2 + lS 3W 2 2 lS 3W

)2

(5)

The position of its center is given by:

500

ro =

l S 3 E l EW

+ l S 3W

2 l S 3W

(6)

0 Y

If the position of the wrist is expressed in spherical coordinates:

-500

l
-500 0 X 500 1000

= (W , W , rW )

(7)

-1000 -1000

The position of the elbow can be written as:

Fig. 5. The reachable workspace of the YIREN arms

According to the algorithm for the workspace evaluation, Figure 5 shows the reachable workspace of the YIREN arm approximately from that of the WABIAN arm and the ARMRA arm. From the value of the manipulability, we can draw a conclusion that singularities mostly occur at the boundary of the workspace. In contrast, there are many singularities in the workspace of others.

l S 1E = ( R z

Ry

R z e x ) R + ro + l S 1S 3 (8)

Where, R y, R z are the rotation matrix around y0 , z0 axis in the o-x0 y0 z0 coordinate system. is the redundancy angle of the elbow. e x is the basis vector in x0 axis. R is the radius of the sphere whose center is S3. The self-motion angle can be arbitrarily chosen within

599

the interval [0, 2] in the condition of the absence of any constrains. When the arm must be avoided the obstacle, singularity or joint limit on the determined trajectory, it usually has to make use of the self-motion. It is quite easy to calculate the remaining range simply by subtracting the corresponding angle segment of the 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 :

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 = lSE + lEW because of singularityself-motion during the adduction of the upper limbexcept rmin = lSE - lEW because of the interference of joints. The results of analyzing show the arm configuration of: figure 1 (a): r = (lS 2 E + lEW ) figure 1 (b): r = (lSE + lEW ) (17) (18) (19)

E S1 S3 W H
Fig. 6. Self-motion angle of the arm 7
i =1

figure 1 (c): r = (lS 3 E + lEW )

The self-motion range of the configuration (b) is bigger, and that of the configuration (c) is smaller. IV. THE KINEMATICS PROPERTIES OF THE ARM The arms dead weight is the main load of motors. Theoretical analysis and computer simulations indicate the YIREN arm was provided with better mechanical characteristics. Euler-Lagrange equations describe the dynamics of a n-DOF robot:

(9) Where i,cur and i,cal represent the current and the calculate value of the joint angles, respectively. Once the is determined the elbow position (xe, ye, ze) is known and the vectors lSE, lEW can be directly calculated. Based on a system of shoulder joint S1 coordinates o1 - x1, y1, z1, the joint angle1 can be calculated: (10) In the shoulder joint S2 coordinates o 2 x 2 , y 2 , z 2, the joint angle2,3 can be calculated:

E ( ) = ( i , cur i , cal ) 2

1 = a tan 2(lS1E y , lS1E x )

2 = a tan 2(lS 3 E y , lS 3 E x )

(11) (12)

3 = a cos(

lS 3 E z lS 3 E

= D( ) + h( , ) + G ( ) + E (20) where D is a 7x7 inertia matrix, h is a 7x1 centrifugal and Coriolis matrix, G is a 7x1 gravity matrix, E is a load matrix. The dynamics of 7DOF robot present complex characteristics of nonlinear and coupling. According to the property of the YIREN arm configuration, the inertia energy of the 3DOF wrist can be simplified as5 =6 =7 =0, and merge its mass into forearm. The Euler-Lagrange equations are simplified as: 1 = D111 + D12 2 + D13 3 + D14 4 +
h122 22 + h133 32 + 2 h1121 2 + 2 h1131 3 + 2 h1141 4 + 2 h123 2 3 + 2 h124 2 4 + 2 h134 3 4 +
(21)

In the elbow joint coordinates o 4 x 4 , y 4 , z 4 , the joint angle4 can be calculated:

4 = a tan 2(lEW y , lEW x )

(13)

Based on a system of wrist joint coordinates o5 x 5 , y 5 , z 5 ,the joint angle5,6 can be calculated:

5 = a tan 2(lWH y , lWH x )

(14) (15)

6 = a cos( WH )
lWH

G1 + E1 In order to estimating the dynamics properties of these three arms configuration, here mainly analyzed the influence of the arms dead weight. If we define the mass of the forearm, the upper arm, the wrist and hand all to be m, and the velocity and the acceleration are zero, the joints moment overcoming the arm dead weight shows in the figure 7 to 9. The results of analyzing show the moment of dead weight: the arm configuration of figure 3 (a), figure 3 (b) is bigger and that of the configuration 3(c) is smaller.

Since hand vector lG is orthodoxy with vector l WH, we can evaluate7:

7 = a tan 2(lG y , lG x )

(16)

Having derived the closed form equation for each1 to 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

600

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. REFERENCES
[1] [2] Jinichi Yarnagucku, Daisuke Nishino, Atsuo Takanishi. Realization of dynamic biped walking varying joint stiffness using antagonistic driven joints. IEEE ICRA 1998, pp.2022-2029. Jin-ichi YAMAGUCHI, Atsuo TAKANISHI, Ichiro KATO. Development of a Biped Walking Robot Compensating for Three-Axis Moment by Trunk Motion. In Proc. of IEEE/RSJ Int. conf. on Intelligent Robots and Systems ( IROS 93), pp. 561-566, 1993. Toshio MORITA, Hiroyasu IWATA, Shigeki SUGANO. Human symbiotic robot design based on division and unification of functional requirements. ICRA 2000, 2229-2234. Toshio MORITA, Koji SHIBUYA, Shigeki SUGANO. Design and control of mobile manipulation system for human symbiotic humanoid: Hadaly-2. ICRA 1998, 1315-1320 K. Berns, T. Asfour, R. Dillmann. ARMAR: An Anthropomorphic Arm for Humanoid Service Robot. ICRA1999 International Conference on Robotics & Automation, Detroit 1999 Thomas Bergener, Carsten Bruckhoff, Percy Dahm, Herbert Janen, Frank Joublin, Rainer Menzner. Arnold: An Anthropomorphic Robot for Human Environments. Internal Report 97-12, extended version of SOAVE paper Rainer Bsichoff, Tamhant Jain. Natural communication and interaction with humanoid robots. Second ISHR1999 J.D.Han, S.Q.Zeng, K.Y Tham,M.Badgero and J.Y.Weng. Dav Humanoid: A Robot Platform for Mental Development. Feng Gao, Fabrice Guy, William A. Gruver. Criteria based analysis and design of three degree of freedom planar robotic manipulators. ICRA 1997, 468-473 Hiroyasu IWATA, Hayato HOSHINO. A Physical Interference Adapting Hardware System Using MIA Arm and Humanoid Surface Covers. IROS 1999,1216-1221 Richard M. Voyles. Terminatorbot: A robot with dual-use arms for manipulation and locomotion. IEEE ICRA 2000, 61-66 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 Percy Dahm, Frank Joublin. Closed form solution for the inverse kinematics of a redundant robot arm. Internal Report IRINI97-08

Fig.7. Moment of the arms dead weight

[3] [4] [5] [6]

[7] Fig.8. Moment of WABIAN arm configuration [8] [9] [10] [11] [12] [13] Fig.9. Moment of ARMRA arm configuration

Fig.10. Arms of humanoid robot YIREN

V. CONCLUSION In this paper 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

601