1999 IEEE International Fuzzy Systems Conference Proceedings

August 2 - 2 5 , 1999, Seoul, Korea

Synchronous Operation for DC Servo-motors with Harmonic Drive Gearing

Yeon Tae Kim, Jeong Won Choi, Suk Gyu Lee
School of Electrical and Electronic Eng., Yeungnam University
214-1 Dae-dong, Gyongsan Gyongbuk 712-749, Korea

Abstract J , L+Rq. + [ N + n , I [ K ( q , q , ) z + K , l + T , = U (2)

Synchronous operation is a control strategy that is
widely applied to industrial field for high efficiency and
precision. This paper proposes an active synchronous
control algorithm by considering the weight value which is
alteration of traditional synchronous operation algorithm,
and simulates it for five DC servo-motors with harmonic
drive. The simulation results using the proposed algorithm
show better performance than that oftraditional algorithm.
where, M(q) is inertia matrix of links, V,(q, q)is centrifugal
and Coriolis term, G(q) is gravity term, and F(q) is friction
term caused by mainly bearing of links, J, is inertia matrix
of motors, B, is damping term concerned with bearing of
motors, and U is a vector of control inpiiu of system.
3. Controller Design
3.1 Integrator Backstepping Method

1. Introduction Integrator Backstepping method constructs pseudo

control laws recursively using Lyapunov concept to
In industrial field, synchronous operation of multi- synthesize final control laws[''. It offers a design tools for
actuators enhances the efficiency and precision of the accommodation of uncertain nonlinearities and can avoid
system operation comparing with individual operation of cancellations of meaningful nonlinearities.
each actuator. I. H.Yun proposes the serial and parallel This paper uses Integrator Backstepping method to
synchronous operation algorithm for multiple DC motors. design nonlinear controller for I-DOF manipulator which
However its simple structure causes trade-off between serial contains DC servo-motor with harmonic drive, and
and parallel type synchronous operation algorithm. This intermediate controller U,., uq and the final control input U
paper proposes a generalized algorithm by applying the is as follows[7'.
weight value which is modification of their algorithm. Our
new algorithm determines the weight values and modifies
the reference inputs of each motor to reduce the error of U,, = K(q,q,J'[k,r, +w,,l~W'"" (3)
each motor. U, = [det{M(q)Jl-'[k,q + wlIE W"" (4)
On the other hand, one of the most desirable abilities of
robot lies in dealing with heavy load. Hence, the speed U = [det{M(q))[l -diog(-)]]-'[kxn % + w.] E W"'" (5)
reduction device is commonly used. However, large aq.,
backlash of the traditional gear system makes it dificult to
operate precisely. Comparatively, harmonic drive gearing is where, kL, k, and k. are the controller gains with positive
a backlash free torque transmission with compact size and value and I is mxm identity matrix. Fig. 1 shows the
low noise, But, it should be pointed out that the biggest composition of one DC motor with harmonic drive and
disadvantage of Harmonic Drive is the inherent nonlinearity controller for simulation.
caused by its structural characteristics. Recently, many
researches['-'1[51[']focus on some desirable characteristics of ...... .............................................
,_
the harmonic drive. This paper references their results about
nonlinearities of harmonic drive to get the dynamic model
of the whole robot system for effective nonlinear controller
design.

2. Modeling of Control System

Fig. I Conceptual Structure of a DC motor with harmonic drive
The precise dynamic equation of m I-DOF manipulators gearing system
including DC servo-motors with nonlinear characteristics of
harmonic drive"] has the form of
3.2 Synchronous Control Algorithm
M ( q ) = i j + V , ( q , q ) q + G ( q ) + F ( q ) - K , = K ( q , q , ) z (1)

0-7803-5406-0/99/$10.00 L1999 IEEE 111-1738

 This paper proposes a new algorithm which is input variables. The variables e , , and e,, as well as d, and
modification of synchronous operation algorithm proposed
by Yun et@. Their serial type synchronous operation e,, have the same membership function
algorithms cannot avoid the piling up the error, and parallel
type cannot compensate for the difference of output from
other motors when load of the motor changes. These
S MO BO I
problem of serial and parallel type algorithm can be
improved by designing each motor's controller to modify
their own reference input actively.
The proposed synchronous operation algorithm with
weight value multiplies weight value w,, between 0 and 1 to
reference input, and multiplies w,, to the output of the
previous motor, then summing them to determine reference
input of i-th motor. The weight value w,, is determined in
real time as equation (6),
SM BG
w,(k) = w,(k - I)+ &w,,(k) (6)

where, w,(k) is the value of w,, at k-th step. The value of

w,,(k) is determined simply 1 - w,,(k).

(b) e,z and en

Fig. 3 Membership function of input variables

m W" MoUv i
We defined fuzzy rule for the system as follows.

1. If ep,is SM and e,, is SM then Aw,~is ZR

2. If e,, is SM and e , is MD then Aw,, is INC

b L L y 3 . If e,, is SM and e,, is BG then Aw,, is GDEC

Fig. 2 Synchronous operation with weight value
'The structure of proposed algorithm is effective €or both
of serial and parallel type. For example, in case of w,! is 1,
w , ~becomes 0, which means the structure of system is serial
type. Conversely, if the weight value w,! becomes 0, it
means that the stmcture of system is similar with parallel
type.
In the simulation, we used fuzzy inference to determine Aw,,
which is the change of w,~,for the active modification of
reference input for any motor itself. The inputs of fuzzy
inference system are the position error e, between i-th
motor and (i- I)-th motor, the position error e,j between i-th
motor and the reference input, and their derivatives with
respect to time ep,, p,! the output is Aw,,.
lQIz&iN
We used the position error en,,e,,, and their derivatives
with respect to time, e>,, p,r for the input of fuzzy
inference system. Fig. 3 shows the membership function of
4. If e,,, is MD and e, is SM then Aw,, is DEC
5. If e,,, is MD and e,, is MD and in is SM then Aw,, is DEC
6. If e,, is MD and e,j is MD and e,, is SM and pH is BG
then Aw,~is INC
7. If e, is MD and e,, is MD and d , is BG and is BG
then Aw,, is GINC
8. If e,, is MD and e,j is BG and 6, is SM then Aw,~is DEC
9. If e,, is MD and e,j is BG and eP,is SM and km is BG
then Aw,, is INC
IO. If e,, is MD and e , is BG and p , is BG and pn is BG
then Aw,! is GINC
11. If e,, is BG and e,j is SM then Aw,~is GINC
12. If eu, is BG and e,, is MD and Aw,, is SM Aw,) is DEC
13. If e,, is BG and e,, is MD and ep, is SM and kn is BG
then Aw,, is INC
14. If e,, is BG and e,, is MD and p, is BG and inis BG
then Aw,, is GINC
15. If e , is BG and e,, is BG and p,, is SM then Aw,, is DEC

111-1739
16. If ep, is BG and e,8 is BG and gp, is SM and B,, is BG
then Aw,, is INC
17. If e, is BG and e,, is BG and io,is BG and B,, is BG
then Aw,, is CINC

Fig. 4 shows membership function of output variable. In

this research, we use Min-Max method for fuzzy inference,
-p+j
O

Eo
I
o 5 7

which is computationally simple and has been used widely. -0.05

0 5 10 15
[sec]
G
m
l
(a) Position error between motor I and motor 2

05
0 05
I
0
.2 -15 ., U5 0 03 1 15
XI04
2

Fig. 4 Membership function of output variable Aw,,

We select the Center-of-Gravitv method which

[sec]
represents quite good results and most widely used by
control engineers, among the defuzzification method of (b) Position error between motor 1 and motor 3
commonly used.

4. Simulation Results

4.1 Simulation System

Simulation is performed to verify the validity of

proposed synchronous operation algorithm, with five DC
servo-motors which is separated mechanically. We compare -0.05
the results of traditional algorithm with that of the proposed 0 5 10 15
algorithm. Table 1 shows the value of parameters used in [sec]
simulation. The values of the load inertia, motor inertia and (c) Position error between motor 1 and motor 4
spring constants of harmonic drive k, and k2 are slightly
different.

4.2 Simulation Results

In the simulation, the first motor follows sine-wave

reference position input and the remainding motors follow
each corresponding reference input determined by using the
position output of first motor and the error between the -0.05 I I
position output of previous motor. Fig. 5 shows the 0 5 10 15
simulation results of traditional serial type synchronous [sec]
operation, and fig. 6 shows the simulation results of
(d) Position error between motor I and motor 5
proposed synchronous operation. The results show that
notable improvement in both transient and steady state
Fig. 5 Position errors for serial type synchronous operation
response when used the proposed algorithm compared with
that of traditional synchronous operation algorithms.

Ill-1740
 5. Conclusion

This paper proposed an active synchronous operation

algorithm by considering weight value to improve
traditional synchronous operation method of serial and
parallel type. The simulation is performed for five DC
servo-motors with harmonic drive gearing. The results of
simulation using the proposed algorithm shows much better
fesults compared with that of traditional algorithm.
Also, this paper designed a nonlinear controller for DC
[sed servo-motors with harmonic drive using Integrator
(a) Position error between motor I and motor 2 Backstepping method, and the simulation results proves the
validity of proposed nonlinear controller.

0.05
References
- [ I ] 7’. Marilier and J. A. Richard, “Non-linear Mechanic
go ~ .... :
and Electric Behavior of a Robot Axis with a Harmonic
Drive Gear”, Robotics & Computer Integrated
Manufacturing, Vol. 5, No. 213, pp. 129 - 136, 1989.
-0.05 [2] T. C. Lin and K. H. Yae, “Use of Spatial Velocity in
Recursive Dynamic Formulation of a Manipulator
[sed Driven by Harmonic Drives”, Mech. Struct. & Mach.,
(b) Position error between motor 1 and motor 3 22 (3), pp. 343 - 370, 1994.
[3] Tsuug-Chieh Lin and K. Harold Yae, “The Effects of
Harmonic Drive Gears on Robot Dynamics”, Advances
in Design Automation. Vol. 2, DE-Vol. 32-2, pp. 515
- 522. 1991.
[4] Miroslav Krstic, Ioannis Kanellakopoulos, Peter
Kokotovic, Nonlinear and Adaptive Confrol Design,
John Wiley & Sons, Inc., New York, NY, 1995.
[5] Tuttle, T. D., “Understanding and Modeling the
Behavior of a Harmonic Drive Gear Transmission”,
-0 05 Master’s Thesis of MIT, 1 9 9 2 , 4 1 3 ~
0 5 10 15 [6] J. H. Yun, I. M. Suh, Y. S. Shin, G. H. Kim,
[sed “Synchronizing Control of Multiple DC Motors”, KIEE
(c) Position error between motor 1 and motor 4 Summer Annual Conference ‘90, pp. 475 - 419,1990.
[7] Yeon Tae Kim, Jeong Won Choi, Dong Gi Hong, Suk
Gyu Lee, “A Study on Nonlinear Controller of DC
Servo-motors with Harmonic Drive Gearing and Its
0.05 ~

Synchronous Operation”, Journal of the Korean Society

of Precision Engineering. Vol. 15, No. 8, pp. 60- 70,
August 1998.

-0.05 I I
0 5 10 15
[sec]
(d) Position error between motor 1 and motor i

Fig. 6 Position errors for synchronous operation with weight value

111-l741
 Table 1. Parameters for simulation oftwo DC motors

Parameter Meaning Value

Sd Dynamic friction coefficient 0.007 Kg-cm-sec
01 Kinematic error 1 0.000001 rad
a, Kinematic error 2 0.000001 rad
JL Load inertia 1 .O Kg-cm'
J", Motor inertia 0.0 I 1 9 Kg-cm2
4b Phase angle 0.03 rad
b, Motor damping constant 0.007 Kg-cm-sec
b, Harmonic drive friction preload 0.0 Kg-cm
b2 Harmonic drive coulomb friction coefficient 0.2747
b3 Harmonic drive viscous damping coefficient 1 0.0047 Kg-cm-sec
ba Harmonic drive viscous damping coefficient 2 0.01 Kg-cm-sec'
bs Periodic friction torque of harmonic drive 0.01 Kg-cm
kl Harmonic drive stiffness constant 1 399.1 Kg-cm/rad
kz Harmonic drive stiffness constant 2 1 5 . 3 4 8 7 IO6
~ Kg-cm/rad3
n Gear reduction ratio 0.02
m Mass of load 5 Kg
L Length of load 24 cm

ill-1742