Copyright © IFAC Control Science and Technology

(8th Triennial World Congress) Kyoto. Japan . 1981

ANALYSIS AND CONTROL OF ARTICULATED

ROBOT ARMS WITH REDUNDANCY

H. Hanafusa, T. Yoshikawa and Y. Nakamura

Automation Research Laboratory, Kyoto University, Gokasho, Uji, Kyoto 611, japan

Abstract. Industrial robot applications have seen remarkable growth in num-

bers and in diversity of uses. In configuration, it is also slowly changing
from cartesian and cylindrical coordinate types to articulated one mainly due
to the fact that articulated robot arms are more flexible and versatile.
From the view point of control, however, arms of this type possess certain
inherent problems. One concerns the nonlinear relation between the position
and orientation of the robot hand and the rotational angles of the joints of
the arm. Another problem concerns the existence of singular points in the
state space of joints, which reduces the robot's degree of freedom.
Although it has sometimes been claimed that redundant arms can effectively
overcome the problem of singularity and increase flexibility and versatility,
these have yet to be studied in full. In this paper, the manipulatability
and redundancy of multi-articulated robot arms are analyzed and utilization
of redundancy is discussed. Experimental results are also presented concern-
ing the utilization of redundancy for trajectory control with provisions for
an obstacle.

Keywords. Robots; manipulators; redundancy; obstacle avoidance;

multivariable control system; computer control; d.c. motors.

INTRODUCTION This paper will analyze the manipulatability

and redundancy of multi-articulated robot arms
With increasing utilization of industrial and discuss utilization of redundancy. First-
robots in various manufacturing areas, demands ly, the concept of manipulation variable is
for more flexible and versatile robots have introduced as the variable suitable for de-
intensified. Articulated robot arms composed scribing a given task of the robot. Secondly,
of rotational joints have recently drawn at- utilizing the manipulation variable, the ma-
tention because of their potentiality for nipulatability and redundancy of the robot are
flexibility and versatility. analyzed. Thirdly, based upon the analysis of
redundancy, the utilization of redundancy is
One of the main problems in controlling an considered and the basic equations are derived
articulated robot arm is a kinematic one for the control of manipulation variables in
arising from the nonlinearity of the arm con- cases where there are orders of priority.
figuration. In other words, the nonlinearity And then, after a brief introduction of "UJI-
between the state of the robot hand (i.e., its BOT", a 7 axes articulated robot arm driven by
position and orientation) and the state of the d.c. servomotors, the basic equations of the
joints (i.e., their rotational angles) compli- utilization of redundancy is applied to a
cates their coordinate transformation [1] [2] trajectory tracking problem with an obstacle
[3] [4]. In particular, because of this non- in the working area and the experimental re-
linearity, there exist some singular points sults are presented.
in the state space of joints where the degree
of freedom of the robot hand is reduced [5].
This further complicates the problem. MANIPULATION VARIABLES

It has sometimes been pointed out that redun- In preparation for the analysis of manipula-
dancy can effectively overcome the problem of tability and redundancy of articulated robots,
singularity and increase the robot's flexibil- the concept of manipulation variable will be
ity and versatility [5] [6]. Also several re- introduced as the variable suitable for de-
searches on the redundancy of 7 axes robots scribing a given task.
with configuration similar to human arms have
been reported [7] [8]. Nevertheless, the anal- When the position and orientation of the robot
ysis and the study of the utilization of re- hand are controlled in 3 dimensional space,
dundant arms with more general configurations tasks have often been described by 6 variables
have not been carried out sufficiently. on the basis of an absolute coordinate system

1927
1928 H. Hanafusa, T. Yoshikawa and Y. Nakamura

fixed in the working area [9]. But in welding, Manipulatability is here used to describe the
for example, since the rotational angle around faculty of robot for performing tasks . This
the weld torch is not essential, only 5 varia- depends not only upon the number of joints but
bles are needed to describe the task. Further- also upon 8 because of the geometrical non-
more the position and orientation of the weld linearity of the robot. Linearizing Eq . (1)
torch are conveniently expressed in terms of around 8, the equation between or and 08 , which
the following 3 axes : the direction of the weld are the small deviations of rand 8 respective-
line, the direction of the normal vector of the ly, is derived as follows :
weld plane and the direction orthogonal to
these two directions. These 3 axes constitute or = J(8 )08, (2)
a curved coordinate s ystem in the case of a
curved weld line. To give another example, in where J(8 ) = df/d8 is the Jacobian of f(8) with
the case of a task performed by two robot arms respect to 8 .
in coordination, it is more desirable to ex-
press the state of one hand with an absolute The manipulatability means whether or not the
coodinate system and that of the other hand robot can realize or at 8 to perform the given
with the coordinate system fixed to the former task. From Eq. (2) it can be seen that or
hand, rather than to express both hands with which can be produced by 08 at 8 belongs to
an absolute coordinate system . R(J), the range of J(8). Conversely if or is
an element of R(J), then it can be produced by
Thus the variables adopted to describe the 08. Hence, we define manipulatable space and
tasks of a robot should not be restricted to degree of manipulatability (d.o . m. ) as follows:
one absolute coordinate system or any other
orthogonal coordinate system fixed to the space . DEFINITION 1
Further, not only the number of variables but
the properties should be changeable according R(J) and its dimension are cal l ed the manipu l a-
to the given task. These variables are called table space and the degree of manipulatabi l i t y
manipulation variables and are expressed by a at 8 respectively.
vector r ERm . The vector r is assumed to be a
function of 8ERIl (n = the number of degree of When the d.o.m. at 8 is equal to the essential
freedom (d . o . f . )) whose components indicate the dimension of the manipulation variables, all
rotational angles of the joints . That is, the tasks required as or can be realized. In
this case, we say that the robot arm is manipu-
r=f(8). (1) latable at 8. This is an extension of the
maneuverability of the robot hand defined by
It should be noticed that there is no restric- Kinoshita and co-workers [4].
tion on the magnitudes of n and m since it is
permitted to describe r redundantly. Redundant Space and Degree of Redundancy

In order to express the orientation of the Redundancy has often been expressed by "degree
robot hand Euler angles are often used . But of redundancy" (d.o.r.) which means the differ-
the relation between Euler angles and 8 is ence between the d.o.f. of the robot and the
often very complicated. Furthermore, Euler essential dimension of the manipulation varia-
angles do not express the actual hand orienta- bles. It is not clear, however, whether or not
tion in an easily comprehensible way. Accord- this "degree of redundancy" expresses exactly
ingly it would be more desirable to express the the redundancy for robots of general configura-
hand orientation by 2 unit vectors fixed to the tion including articulated ones. Also the
hand as shown in Fig. 1. That is to say, 3 "degree of redundancy" is not by itself suffi-
d.o.f. of the hand orientation are redundantly cient for the utilization of redundancy. In
described by 6 variables. Although this de- this subsection we will define the redundant
scription method has the drawback of increasing space which expresses the redundancy qualita-
the number of variables to be handled, it has tively, and the d.o.r. suitable for robots of
the following merits: general configuration.
a) the correspondence between the 2 unit
vectors and the actual hand orientation To begin with, we consider the redundancy of
can be easily seen, the cartesian coordinates robot shown in Fig. 2.
b) the hand orientatior. can be expressed by Taking the hand position as the manipulation
simpler equations of 8 , variables, the relation between r = col. (x,y)
c) the coordinate transformation for the and 8 = col. (8 ,8 ,8 ) is given by
1 2 3
hand orientation is simplified in cases
where another orthogonal coordinate system r = H8 (3)
is introduced .

As a matter of convenience the number of vari-

010
where H = [ 1 °1 ].
In this case the redundancy is explained by the
ables which are essentially independent among following equation.
manipulation variables will be called "essen-
tial dimension of the manipulation variables". r = H (8 + k 8 ) = H8 (4 )
r
Here k is an arbitrary scalar and 8r col.(l,
MANIPULATABILITY AND REDUNDANCY 0,-1). In Eq. (4) the vector k 8 mapped to
o by H expresses the redundancy r qualitatively
Manipulatable Space and Degree of Manipulatability and the dimension of the space spanned by such
 Analysis and Control of Articulated Robot Arms 1929

vectors refers to the extent of redundancy. dynamic redundancy is expressed by mapped e

into 0 by J(0) . Hence they are related to the
Following the same line of argument, the re- characteristic of J(0) in the same way as in
dundancy of articulated robots can be discussed the static case. Therefore DEFINITION 1 and
based on Eq. (2). The redundancy is now quali- DEFINITION 2 are valid also for the dynamic
tatively expressed by the vectors mapped to 0 manipulatability and redundancy.
by J(0 ), i.e., the elements of N(J), the null
space of J(0) . And the extent of the redundan-
cy is given by the dimension of N( J). Hence UTILIZATION OF REDUNDANCY IN CONTROL
the following definition: OF MANIPULATION VARIABLE WITH ORDER
OF PRIORITY
DEFINITION 2
Task Described by the Manipulation Variables
N( J) and its dimension are called the redundant with Order of Priority
space and the degree of redundancy at 0 respec-
tively. Not to speak of welding, there are various
kinds of tasks where the command for the hand
In the case where the robot is not manipulata- position is more significant than that for the
ble at 0 , the d.o.r . at 0 expresses the extent hand orientation. Contrary to these is the
left for 00 as redundancy when robot makes a task of pointing a camera attached to the robot
move of an element of the manipulatable space. hand to a certain direction where the hand ori-
In this sence the degree of redundancy in DEF- entation is more important than its position.
INITION 2 is an extension of that as commonly If in these cases it is impossible to realize
understood. the both commands, it would be adequate to re-
alize the command for the more significant var-
It is well known that the range and the null iables of the two. We try to realize the less
space of a matrix M satisfy the following: significant command as much as possible only
after realizing the more significant command.
l = dim R(M) + dim N(M) (5) In this section we regard these tasks as the
tasks described by the manipulation variables
where l is the number of columns of matrix M. with order of priority.
Applying Eq . (5) to J(0 ), we have the follow-
ing proposition. We can also deal with more complicated tasks by
this task description. For example, tasks in
PROPOSITION which the hand position and orientation should
track the desired trajectory and the arm should
Between the d. o.m. and the d.o . r . of the robot avoid the obstacles in the working area as much
with n d.o.f., tha following relation holds as possible, can be treated by assigning higher
regardless of 0 . priority to the hand position and orientation
and lower priority to the motion of avoiding
d. o.m. + d. o .r = n (6) obstacles .

This proposition indicates that at singular Derivation of Basic Equations

points of J(0) where the d.o.m. decreases, the
d.o.r. increases by the number the d.o.m. de - Suppose that the first manipulation variable
creases. rl €Rml are given higher priority than the sec-
Dynamic Manipulatability and Redundancy ond manipulation variable r2 €Rm2. Then the
equations corresponding to Eqs. (1) and (2)
The above discussion was solely based on the are expressed as:
geometric relation between or and 00 , and did
i
not consider the dynamics of the robot. Thus r =fi (0) (i 1,2) (8 )
it was an analysis of static manipulatability i i
and redundancy. In this subsection we will or = J (0) 00 (i 1,2) (9)
briefly comment on dynamic manipulatability and i i
redundancy . In this case the manipulatability where J (0) = af la0 .
and redundancy can be considered on the basis
The general solution of 00 satisfying Eq. (9)
of the relation between r, the a~celerations of
the manipulation variables, and 0, the acceler- for i = 1 is given by:
ations of joint angles. By differentiating 00 = Jl # (0) orl + { I - Jl # (0) Jl (0) } Y (10)
Eq. (1) twice with respect to time, we obtain:

r - j (0) El = J (0) i:i . (7)

where J 1# (0 ) is the generalized inverse matrix
of Jl(0) and y€R n is an arbitrary vector [10].
Instead of Eq. (2) we should consider Eq. (7)
the basic equation here. In Eq. (7) the second From R( J1#) = N(Jll-, R(I -} #Jl) = N( } ) [10],
term of the left-hand side refers to the effect it is understood that the second term of the
of the present state 0 and El on r ...To realize right-hand side of Eq. (10) is an eleQent of
the required r, we need to produce 0 satisfying the redundant space at 0 of the first manipu-
Eq. (7). In other words, dynamic manipulata- lation variables . So choosing y properly, the
bility means whether such i:i exists or not, and redundancy can be utilized.
1930 H. Hanafusa. T. Yoshikawa and Y. Nakamura

The vector y which minimizes 11 or2 - J2 (0) 00 11. system for tracking the desired trajectory rO(t)
the euclidean norm of or2 - J 2 (0 ) 00 , is given by : has been designed as follows. If we take the
feed forward compensation for the third and
y = {J2( I _ }#Jl)}# (or 2 _ J2 J l#or l ) (11) fourth terms of the right-hand side of Eq . (131
the system can be regarded as linear and decou-
pled. Furthermore, because of P « Q in Table 1
From Eqs . (10) and (11). 00 can be computed for i i
l it can be seen that the time constants for ve-
or and or2 in consideration of the priority.
locity command is very small. Hence, we de-
signed a servo system in which the modified
For the trajectory tracking problem in working
space with obstacles, as stated in the preced- trajectory r * (t) is generated according to the
ing subsection, if the motion for avoiding ob- deviation from rO(t), and the 8 necessary for
stacles have been previously described in terms
2 realizing r*(t) is commanded to the open loop
of 0, we can make r = 0 . Then, using J 2 (0) = I
and the properties of generalized inverse ma- controller which was constructed with due con-
sideration to the dynamics. Figure 5 shows the
trices, (I _Jl#Jl)#= I _ Jl#JlandJl#=Jl#JlJl#, conceptual block diagram of the servo system.
in terms of Eqs. (10) and (11) 00 can be ex- For example, at time to a modified trajectory
pressed by the following simpler form:
r* (t) (t ~ to) is generated as follows.
00 = Jl# (0) or l + {I - Jl# (0 ) Jl (0)} or2. (12)
r * (t) = r 0 (t)
+ exp{-H (t-t )} {r(t ) - r (to)} (14)
o
TRAJECTORY CONTROL OF UJIBOT WITH O o
PROVISIONS FOR AN OBSTACLE where a constant matrix H=diag . (H.) €Rmxm,H .> O.
1 1

Dynamics of UJIBOT Example of Trajectory Control with Provisions

for an Obstacle
Figure 3 shows a comprehensive view of UJIBOT,
an articulated robot of 7 d.o.f . driven by d.c. Based on the controller designed above, we shall
servomotors. UJIBOT is equipped with a poten- consider an example of the trajectory control
tiometer and a tachogenerator at each joint in with provisions for an obstacle.
order to measure 0 and 0. Figure 4 is the dis-
tribution of d . o.f . of UJIBOT, which shows that The task is as follows: from the initial state
unlike human arm type robots UJIBOT is charac- 0 = (2.48 -24.6 -22.1 15.9 -29.1 3 . 30 -49 . 4) T
terised by the general distribution of d.o.f . . 1
The length of each link is (ll lz l3 l4 ls l6 (deg.) shown in Fig. 6 the hand position is to
l7) = (0.0 0.30000 0.27810 0.21190 0.25000 be moved in the direction of Y-axis at the con-
0.06853 0.23147) (m). The mass altogether is stant rate of 0.03 m/s for 24 seconds. There
about 134 kg. is an obstacle in the working space as shown in
Fig. 6. The arm must avoid colliding with this
The distinctive feature of the driving system obstacle.
of UJIBOT is its large reduction ratio. Since,
compared with the dynamics of the arm itself, The hand position expressed by the absolute
the dynamics of the transmission is emphasized coordinate are choosen as the first manipula-
by the square of the reduction ratio, we shall tion variables, and 0 for avoiding collision
have to consider only the inertia force, viscous as the second manipulation variables. That is,
friction force, Coulomb friction force of the 1
transmission of UJIBOT and the gravity force of r ( X Y Z /
arm which is important statically. Furthermore, 2 (15)
since the inductance of armature controlled
r o
d.c. servomotor is very small, we shall neglect The desired trajectory is expressed as follows.
the inductances of driving motors of UJIBOT.
Considering the above, the dynamics of UJIBOT rlO(t) rlO(O) + t (0 0.03 0) T
is represented as follows: 2O (16 )
r (t) 0
r
V = P8+ Q0+ RSignG + STg(0) (13)
where 0 refers to the desirable constant arm
7 r
where V€R is the control voltage vector of posture for avoiding collision, which had been
7 taught in advance. Selecting the modified traj-
motors, Tg(0)€R is the torque vector generated
by gravity force, sign0=col.(sign0 . )€R 7 , P= ectorys r 1* (t) and r 2* (t) as Eq. (14), the ve-
1
diag. (P. )€R7x7, Q = diag. (Q . ) €R7x7, R = diag. (R.) .
lOClty r.1* an d r.2* at tIme
. . ca lcuIate
to IS d as
7 7 1 71 7 1 follows.
€R x and S = diag,(Si) €R x for constants, Pi'
1* 10 1 1 10
Qi' Ri and Si' Table 1 shows the numerical r (to) = r (to) - H {r (to) - r (to)}
values of Pi' Qi' Ri and Si' r2* (to) = - H2 { 0 (t ) - 0 } (17)
O r
Design of the Trajectory Control System Regarding Eq. (12) as the relation among the
Based upon the discussion above, the control small changes 0f 0-, r 1 an d r 2 In
. In
. f Inlteslma
" . 1
 Analysis and Control of Articulated Robot Arms 1931

time at, the following equation is obtained. It was shown that the sum of the degree of ma-
nipulatability and the degree of redundancy is
El = Jl# (0) rl + { I - Jl# (0) }(O) } r2 (18) equal to the degree of freedom of the robot
without regard to 0.
From Eqs. (17) and (18) the velocity command El
can be calculated. The block diagram of the 3) Based on the analysis of the redundancy, a
control system is shown in Fig. 7. control scheme was developed for a task having
several objectives with order of priority.
Figure 8 shows the experimental result without This is a typical example of the utilization
of redundancy.
provisions for the obstacle, where H~ = 3 (l/s)
2 1
for i=l, 2, 3 and Hi =0 (1/s) for i=l, ··· , 7. 4) Experimental results were presented concern-
ing the utilization of redundancy for an artic-
It can be seen that the arm posture changes ulated robot arm with 7 axes driven by d.c. ser-
smoothly and the desired trajectory of the vomotors. The hand followed the desired trajec-
first manipulation variables is tracked almost tory successfully while the arm avoiding the
perfectly. The obstacle, however, interferes collision with an obstacle in the working space.
with the movement of the arm as shown in the
photograph of Fig. 8 .
REFERENCES
Teaching Or as shown in Fig. 9, the experimen-
tal result with provisions for the obstacle is (1] Whitney, D. E. (1969). Resolved motion rate
control of manipulators and human prosthe-
given in Fig. 10, where H~ = 3 (l/s) for i=1,2,3 ses. IEEE Trans. Man-mach. Syst., 10,47-53.
2 1
and Hi =0 . 1 (l/s) for i=l, .. · , 7. In compar- [2] Stepanenko, Y. A. (1970) . In M. M. Gavrilovic
and A. B. Wilson, Jr. (Ed.), Advances in
ison with Fig. 8 it can be seen that the arm External Control of Human Extremities,
posture changes greatly in first 10 seconds to Yugoslav Commit. Electr. Autom., Belgrade .
approach Or and thus the arm successfully avoids [3] Roth, B., Rastegar, J. and Scheinman, V.
the obstacle. The control system in Fig . 7 is (1973). On the design of computer controlled
all software. Using the minicomputer NOVA manipulator. In On Theory and Practice of
model 03 with a floating point processor unit, Robots and Manipulators, Vol . I . Springer-
it requires 47 ms of sampling time. Verlag, Wien-New York. pp. 93-113.
[4] Kinoshita, G., Kobayashi, H. and Hiraoka,
Y. (1978). Study on maneuverability of
CONCLUSION manipulators. Proc. 21st J. A. C. C. in
Japan, 229-230.
Results obtained in this paper can be summa- [5] Uchiyama, M. (1979). Study on dynamic con-
rized as follows: trol of artificial arm. Trans. JSME, 45,
314-345. --
1) Manipulation variables were defined as vari- [6] Freund, E. (1977) Path control for a redun-
ables suitable for describing a given task of dant type of industrial robot. Proc. 7th
the robot. Based on the manipulation variables 1. S. 1. R., 107-114.
the manipulatability, which is the faculty of [7] Nakano, E. (1976). Mechanism and control
the robot for performing the task, was analyzed. of articulated manipulator. J. Soc. Instrum .
The manipulatable space and the degree of ma- & Control Eng., ~, 637-644.
nipulatability, which are important concepts [8] Mizukawa, M. and c0-workers (1975) . Torque
in understanding the manipulating faculty of position control of articulated artificial
the robot, were also defined. arm. Proc. 4th Biomechanism Sympo., 242-253.
[9] Whitney, D. E. (1972). Mathmatics of coor-
2) The concept of redundancy was clarified by dinated control of prosthetic arms and ma-
regarding it as the remaining faculty of the nipulators. Trans. ASME Ser. G, 94,303-309.
robot under the condition that it must perform (10] Boullion, T. L. and Odell, P. L. (1971).
the task described by the manipulation varia- Generalized Inverse Matrices, Jhon Wiley
bles as much as possible. The redundant space & Sons, New York. Chap. 1, pp. 1-11.
and the degree of redundancy were also defined. For Discussion see page 1940

standard
posture f'--,*~

Fig. 1 Description of hand orientation Fig. 2 Cartesian coordinate robot Fig. 3 Comprehensive vie" of UJIBOT.
by two unit vectors. with redund ancy .

eST 4 - 0
 1932 H. Hanafusa, T. Yoshikawa and Y. Nakarnura

.
r (t ) j- I tI

,t cp 1 gelleratlon of modifi('J tr :t.il·\... t 0 r~ ·.

~t cl': coo rdinate tr31l..-.ft"'lr:Ti ;Itlon .l n J lItill ::l tio :l Clf' r l'J:lT:J~!:h::'
:-:.tC'p:; cOC'l rJi lutC' tra!l:-:.t" I.'f:-:::lt l C'il .

Fig. Seno system for trajectory control of UJ IBOT.

Estimated Parameters of UJ I BOT

2
P . vs
I
Qi vs R. v
1
S. v / Nm
1

10. 76 189 .9 1.623 0.0

40. 80 108 7 .2 .2.070 4 . 475 .10- 3
.676 233 0.950. .118)( I 0- 2
2
653 128 0.9207 2. 067,10-
2
491 181 .7 1. 086 3. 323x10-
10.92 47.09 1. 274 2. 85 I , 10- 1
from X-oxis fro ~ v-axIs
0 . 714 5.686 1.014 0
Fig. 6 Initial position 0 1 and an obstacle.

Fig. 4 The d . o.f. of UJIBOT.

r
10 (t)I......-~~

1
r

,0
r~ (tJ~=*;~
or
from X-axis from Y-oxls
Fig . Block diagram of trajectory control with provisions for obstacles. Fig. 9 Desirable arm posture 0 in order
to avoid the collision. r

3cmls x 24s

from Y-axis from Y-axiS

taught t;;::q / / / /
arm posture "" .... ""
/
6r / /
/
/
I
I
I

arm posture

from Y-oxls
==~> 3cm/s x 24S

Fig. 8 Trajectory of th e arm without provisions Fig. 10 Trajectory of the arm with provisions for the obs tacle.
for the obstacle.