Robot Wrist Configurations, Mechanisms and Kinematics

A. Romiti, T. Raparelli, M. SorliZ

Department of Mechanics
Politecnico di Torino, Italy
2 Department ofindustrial Engineering

University of Cassino (FR), Italy

Abstract: This paper discusses a general method for detennining the kinematic performance of spherical
robot wrists. Different wrist types considered include R-P-R (roll-pitch-roll) and P-Y-R (pitch-yaw-roll) wrists.
Singularity conditions are indicated for both cases. For R-P-R wrists, singularity is defined by the ratio of the
angular velocity of each motor to the velocity around the degeneracy axis. For P-Y-R wrists, singularity is
identified both by analyzing the Jacobian matrix and by analyzing the relative velocities of the wrist
components.

1 Introduction

Robot wrists are designed to provide orientation to the end effector. They should preferably be centered
around a point. that means to constitute "spherical joints", because the resulting configuration is more
dexterous and less cumbersome than the other configurations.
Robot wrists should have a low "degeneracy level", that represents the region where some rotations
around certain fixed axes in the cartesian space are forbidden, or require very high speeds of the actuators.
Degeneracy occurs when the three axes of the spherical wrist joints are contained in one plane. The
degeneracy condition implies that it is not possible to follow the shortest path in producing certain
orientations according to a specified sequence in the work envelope without exceeding the maximum
possible velocity for certain joints [Paul and Stevenson (1983)], [Huang and Milenkovic (1987)], [Tre-
veljian et al. (1986)].
They are two main principles concerning the make of wrists. One is the R-P-R (roll-pitch-roll) method;
the other is the P-Y -R (pitch-yaw-roll) method [Rivin (1988)]. R-P-R wrists are based on the application,
through conical gears, of the Euler formulas for successive rotations of bodies with respect to the others
[Litvin and Zhang (1986)], [Romiti and Sorli (1992)]. P-Y-R wrists act in different way. Two axes (p
and Y) are fixed in space (with respect to the last robot arm): the adjiustment of rotations to comply to
kinematic constraints is obtained by free rotations of bodies one with respect to the others [Milenkovic
(1987)], [Rosheim (1989)], [Romiti and Raparelli (1993)].

2 R-P-R wrists

R-P-R wrists consist of three arms connected in sequence via three rotary joints. The kinematic connection
of each arm with the preceding arm can be described using the transfer matrices ,·1 Ai; if a..1 is the angle
which brings joint axis Zi.I to Zi through a rotation around axis Xi.!> and if e i is the angle which brings axis
Xi.I to Xi through a rotation around axis Zi we have:
cosei -sinei o
:-IAi = Rot(xi_l,ai_I)Rot(Zi,e.> = sineicosai _1 coseiCOSai _1 - sinai_I (1)
sin e i sin a i _I cosei sin ai-I cos ai_I
Angular velocities of the successive arms are given by the expression:
(2)

where Wci -I.i) describes the angular velocity of body i relative to body i-I. Assuming that the three sets
are arranged so that the three degrees of freedom e i occur around axes Zi' we have:
w=9),>9l2 +9,k3 (3)
Remembering (1), we then have (where Ci=cos(e,) and Si=sin(e,) with i=I,2):
0), 0 -SI CIS2 91
0)= O)y 0 CI SIS2 92 (4)
0), 0 C2 93

P. Kopacek (ed.), Robotics in Alpe-Adria Region

© Springer-Verlag/Wien 1994
 45

We will examine a spherical wrist developed at our laboratory. It has three axes which intersect at right
angles (Figure la). Roll movement of arm I is controlled by the motor with angular velocity ~1' Arm 2
is rotated around axis q by the bevel gear set consisting of gears N1 (connected to the motor with angular
velocity~:J and N2 (connected to arm 2). The roll movement takes place around axis ~ and is produced
via bevel gears N3• N4. Ns and N6•
Indicating the axis perpendicular to q and Z3 with x' and using the notation shown in figure lb which
defmes the configuration assumed by the three sets of cartesian axes. we can write:
Z.
OOx' = C100x + Sl00y
OOy' = C100y - Sl00x (5) cl- e•
COz ' = c.oz
As the wrist has a single center.
we have:

N2 N4 Ns
----- (6)
Nl N3 N6
N2
$2 -$1 = N/J2 ZI

N.
$,-$1 = N3 {I. (7)

$1 = {II

.,
Fig.la. Kinematic scheme of a R-P-R wrist

The common angular speed of wheels N4 and Ns

around axis Z2 is {I•. The angular speed of wheel N3•
{I,. is then given by: Z. // X,X, ,X. ,X,
/
/
/
/
/
(8) ./
,/
,/

The relations between velocities {Ii and $i with i = o

1,2,3 are thus: z.z, z.
Fig. lb. Wrist reference systems

{I, = $,
N,
{l2 = N2 ($2 - $1) (9)

Ns[N3 N, ]
{l3 = N6 N4 ($3 - $,) - N2 ($2 - $1)

Introducing equations (5) in (7) gives:

C2
$1 = 00, - S2 OOx'

C2 N2
$2 = 00, - S2 OOx' + Nl OOy' (10)

I (N.N6 ) N4
$3 = 00, +s,: N,Ns -C2 00,.+ N3 OOy'
Fig.2. R-P-R wrist constructive example
46

Note that possible degeneracy conditions are shown by the fact that it is not possible to rotate around
a a
axis x' for 2=0. In other words, with 2 tending towards zero, the motors must have angular velocities
«Pi tending towards infinity in order to achieve a finite velocity 00",
The ratios «PJoo,. define the degeneracy level of this R-P-R wrist as a function of angle 2•a
A constructive solution for a wrist R-P-R, whose kinematic behaviour corresponds at the scheme offigure
la, is shown in figure 2.
Similar considerations should be made for other types of R-P-R wrist as discussed by Romiti and Sorli
(1992).

3 P-Y-R wrists

The wrist analyzed here is a pitch-jaw-roIJ type. Unlike the majority of wrists of this type however, all
movements are actuated by motors which can be mounted on the robot base and by belt and chain drives.
Configuration is spherical.
Figure 3 shows the reference systems used and the wrist's kinematic layout: x,y,z is a set of Cartesian
axes located at the end of the last robot arm, while 1,7, k are the respective versors. Axis x coincides with
the arm longitudinal axis, while axes y and z are transverse to the arm. The origin of the set of axes is
the wrist center.
The pitch motion (in plane xz) and yaw motion (in plane xy) are transmitted to the members carrying the
end effector, to which the roll motion is communicated.
In figure 3, the end effector coincides with bevel gear shaft 4, which is associated with versorv; rotation
around v is indicated with 8 3 and identifies the roll motion.
Yaw and pitch motions are identified by e,k and eJ respectively.
The axis of gears 2 and 3 identified by versor Xis connected by body 8 via rotating fits to the end effector.
Ii.
The axis of the articulation carried by toothed quadrant 6 is identified by versor This articulation is
connected by body 7 to the end effector again through a rotating fit. This permits a relative adjustment
motion between members.
Ii. Ii
Axis v is thus always normal to vcrsors Xand Versors Xand move in planes xz and xy respectively.

--L - - X
e.

Fig.3. Kinematic scheme of a P- Y-R wrist

In a configuration obtained following a pitch and yaw motion, a. is the angular velocity of body 8 relative
Ii.
to X, and ~ is the angular velocity of 7 relative to Thus:
- - - C2 C1I + S2C
V=I1I\A=
J- SlC,k
(11)
-Vc~+s~ci
given that:
1i=-s,1 +CJ (12), (13)
 47

I\ngular velocities of bodies 7 and 8 are:

ID,=ai+~1i (14),(15)
v
l1ie angular velocity of body 8 relative to body 7 must be directed along because of the constraint of
rotary joints; we have:
00sn = Ois - ~ = (a.s\ +~S2)7 +(a, - CiC2)} +(Ctc, - a2>""£ = IDsnv (16)
and then one has:
· C2C,
as,+~S2=IDsn.1
vcr+src]
a S2C,
,-CiC2=IDsn--JC~+S~Ci (17)

· a s,c2
aC,- 2=--olsn--Jcr+srci
which give the following two expressions:

a- c , 2
s,c J
· --.!..(a -IDsn--Jcr+srci 2 --.!..(a -S2C ,
~-C2 ' IDsn--Jcr+s~ci J (18),(19)
By stating values for CJ)S{7 in (18) and (19), we can obtain a and ~ expressed only as a function of a, and
a2• Le.:

a=a(a" a 2) ~ = ~(9" 92 ) (20), (21)

The absolute angular velocity of the end effector can be written in either of the following ways (93 ' and
93" differ only by CJ)S{7):
ro = 9,} +iii +9 .v 3 (22)

ro = a;£ +~Ii +a··v 3 (23)

In matrix form we have:

0 S,
--Jc~+s~Ci
CJ)x 9, 9,
S2C,
CJ)= CJ)y 0 a = [1'] a (22')
CJ), --JC~+S~Ci a/ 93 '
-S,C2
0 C,
--Jc~+s;ci
C,C2
0 -S2
--Jcr+s~ci
CJ)x
S2C ,
CJ)= CJ)y 0 C2 (23')
CJ), --JC~+S~Ci
-S,C2
o
--JC~+S~Ci
The values of angles 6, and 62 at which the determinant of the Jacobian matrix is zero identify the
singularity configurations. There are thus two possible constraint configurations. one for the first me-
chanism and one for the second mechanism. For the first mechanism, we have det[J']=O for ~=O, i.e.
v
for 62=1t!2. Angular velocity ~ cannot be given: axis cannot rotate around x. For the second mechanism,
a v
det [J"]=O for C,=O, i.e for 6,=1t!2. Rotation is impeded: axis cannot rotate around axis x. In both
cases, the singularities could have been determined directly from (18) and (19), which show that ~ tends
a
to zero for 62 tending to 1t!2, and that tends to zero for 6, tending to 1t!2. In any case, the axis of the
degeneracy cones is perpendicular to the wrist's main longitudinal axis (directed along x), so that the
wrist shows no degeneracy along the axis of the spherical surface which defines the size of the work
space.
48

The relations between a and ~i with i = 1,2,3 are thus:

i

Ns NlN3 N/V3 NlN3

a = ~l
l a = N6 ~2
2 a = N~4 ~l + N~4 ~2 + N~4 ~3
3
(24)

where the number of teeth on the i-nth gear is indicated with N;. By introducing equations (20)
and (21) in (22) or (23) and allowing for (24), it is possible to obtain a procedural analysis similar to that
conducted for R-P-R wrists and thus indicate degeneracy levels. Figure 4 shows the constructive solution
of the new P-Y -R wrist.

1!O"I-_h~..... 8
-I;>
x

B'

Fig.4. P-Y -R wrist constructive example

4 Conclusions

The kinematic performance of two different types of orthogonal axis spherical wrists with three degrees
offreedom are analyzed and compared. The configurations of the wrists presented in the paper are original,
but the analysis method can be readily extended to other types of R-P-R and P-Y-R wrists. Wrist dege-
neracy conditions are discussed, with particular reference to the position of the degeneracy cone and
degeneracy level. The kinematic relations are then found which link the angular velocities of the motors
actuating the three degrees of freedom to the component of the end effector absolute velocity in the set
of three axes located on the terminal robot arm.
In this way, we think to offer an overview of wrist configurations and possibilities, and an instrument
both for designing wrists and for forecasting and comparing their performances.
This research project was supported by National Research Council C.N R . ofItaly as part ofthe "Progetto Finalizzato
Robotica".

Bibliography
Huang ,B., Milenkovic, V. (1987). On an algorithm negotiating wrist singularities. Proc. of 17th Int. Symposium on
Industrial Robots, pp.13-I ,13-6.
Litvin,F.L., Zhang Yi.(1986). Robotic bevel gear differential train. International Journal of Robotic Research, vol.5,
n.2, pp.75-81.
Milenkovic,V.(1987). New non-singular robot wrist design. Proc. of 17th Int. Symposium on Industrial Robots,
pp.13-29,13-42.
Paul,R.P., Stevenson,C.N.( 1983). Kinematics of robot wrists. International Journal of Robotic Research, vol.2, n.l,
pp.31-38.
Rivin,E.1.(1988). Mechanical design of robots. McGraw-Hili, pp.285-303.
Romiti,A.,Sorli, M.(1992). Wrist analysis and degeneracy evaluation and comparison. Proc.of 2th Int. Symp. on
Measurement and Control in Robotics, Tsukuba, Japan, pp.437-444.
Romiti,A.,Raparelli,T.(1993). A new spherical robot wrist with remote degeneration regions. Submitted to 6th
ICAR, Tokyo, Japan.
Rosheim,M.E.(1989). Robot wrist actuators. Wiley.
Treveljian,J.P.,Kovesi, P.O.,Ong, M.,Elford, 0 .(1986). ET: a wrist mechanism without singular positions. Inter-
national Journal of Robotic Research, vol.4 , n.4, pp.71·85.