Professional Documents
Culture Documents
A Comparative Study of Modeling and Identification of The Pneumatic Artificial Muscle (Pam) Manipulator Based On Recurrent Neural Networks
A Comparative Study of Modeling and Identification of The Pneumatic Artificial Muscle (Pam) Manipulator Based On Recurrent Neural Networks
Abstract
The paper deals with the PAM manipulator modeling and identification based on recurrent neural
networks. The first time, the most powerful types of neural network-based nonlinear autoregressive
models, namely, NNARMAX, NNOE and NNARX models will be applied comparatively of the PAM
manipulator identification. Furthermore, the evaluation of different nonlinear Neural Network Auto-
Regressive models of PAM manipulator with various hidden layer nodes is completely discussed. On
this basis the merit and demerit features of each identified model of the highly nonlinear PAM
manipulator have been analyzed and compared. The results show that the nonlinear NNARX model
yields more performance and higher accuracy than the other nonlinear NNARMAX and NNOE model
schemes. These results can be applied to model and identify not only of the PAM manipulator but also
of other nonlinear and time-varied parametric industrial systems.
Keywords: pneumatic artificial muscle (PAM), PAM manipulator, recurrent neural networks,
nonlinear NNARMAX model, nonlinear NNARX model, nonlinear NNOE model, modeling and
identification.
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-98-
Track 1: AUTOMATIC CONTROL
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-99-
Track 1: AUTOMATIC CONTROL
An experiment system is illustrated in Fig.4. Fig.5 Experiment system for the PAM manipulator
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-100-
Track 1: AUTOMATIC CONTROL
e(t-1)
u(t-2)
yhat(t)
u(t-1)
y(t-1)
tested 2-axes PAM manipulator and the FITNESS CONVERGENCE OF NONLINEAR NNARMAX MODEL OF THE 2nd LINK 2-AXES PAM MANIPULATOR
(NNARMAX[2,2,2] ; Number of hidden nodes = 5)
0
10
0 2 4 6 8 10 12 14 16 18 20
2nd Link PAM manipulator Output and nonlinear NNARMAX Model Output
(NNARMAX [2,2,2] ; Number of hidden nodes = 5)
30
PAM manipulator response
NNARMAX model [2,2,2] response
IDENTIFICATION & MODELING NNARX MODEL of THE PROTOTY PE PA M MANIPULATOR
20
BSinput -[v]
5.5
5
Output response- [degree]
PR
4.5 10
0 5 10 15 20 25 30 35 40
30
0
20
10 -10
Joint Angle [deg]
0
-20
-10
-30
-20
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (samples)
-30 Error of 2nd Link PAM manipulator Output and predictive NNARMAX Model Output - (y-yhat)
0 5 10 15 20 25 30 35 40 10
Time [s]
Error - [degree]
-10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-101-
Track 1: AUTOMATIC CONTROL
u(t-1)
Network Model with the same typical
yhat(t-2)
yhat(t)
configuration (na=2, nb=2) for 3 typical NN
yhat(t-1)
Models (NNARX, NNARMAX and
1
FITNESS CONVERGENCE OF NONLINEAR NNOE MODEL OF THE 2nd Link 2-AXES PAM MANIPULATOR
(NNOE[2,2] ; Number of hidden nodes = 5) NNOE). It keeps all testing features like the
1st TEST except the hidden layer of all three
10
10
0
NN models composes 10 nodes.
The results concern the structure of each NN
Criterion of fitness
10
-1
0 10 20 30 40 50 60 70 80 90 100
output are showed in following Fig.11-12-13:
Iteration
2nd Link PAM manipulator Output and nonlinear NNOE Model Output
(NNOE [2,2] ; Number of hidden nodes = 5)
30
PAM manipulator response
a) NNARMAX Model:
NNOE model [2,2] response
20
e(t-2)
Output response - [degree]
10 e(t-1)
u(t-2)
yhat(t)
0 u(t-1)
y(t-2)
-10
y(t-1)
-20
FITNESS CONVERGENCE OF NONLINEAR NNARMAX MODEL OF THE 2nd LINK 2-AXES PAM MANIPULATOR
3
(NNARMAX[2,2,2] ; Number of hidden nodes = 10)
10
-30
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (samples) 2
10
Error of 2nd Link PAM manipulator Output and predictive NNOE Model Output - (y-yhat)
1
2
Error - [degree]
Criterion of fitness 10
0
0
10
-2
-1
0 200 400 600 800 1000 1200 1400 1600 1800 2000 10
time (samples)
-2
10
0 10 20 30 40 50 60 70 80 90 100
Iteration
2nd Link PAM manipulator Output and nonlinear NNARMAX Model Output
(NNARMAX [2,2,2] ; Number of hidden nodes = 10)
c) NNARX Model: 30
20
u(t-2)
10
utput response- [degree]
u(t-1)
yhat(t)
y(t-2)
0
y(t-1)
O
-10
FITNESS CONVERGENCE OF NONLINEAR NNARX MODEL OF THE 2nd LINK 2-AXES PAM MANIPULATOR
(NNARX[2,2] ; Number of hidden nodes = 5)
3
10 -20
2
10 -30
0 2 4 6 8 10 12 14 16 18 20
Error of 2nd Link PAM manipulator Output and predictive NNARMAX Model Output - (y-yhat)
2
Error - [degree]
1 1
10
Criterion of fitness
-1
0 -2
10 0 2 4 6 8 10 12 14 16 18 20
time - [s ec ]
-1
10
Fig. 11 PAM manipulator NNARMAX Model results
-2
10
-3
10
b) NNOE Model:
0 10 20 30 40 50 60 70 80 90 100
Iteration
u(t-2)
u(t-1)
yhat(t)
2nd Link PAM manipulator Output and nonlinear NNARX Model Output
(NNARX [2,2] ; Number of hidden nodes = 5) yhat(t-2)
30
PAM manipulator res ponse
yhat(t-1)
NNARX model [2,2] respons e
20
2nd Link PAM manipulat or Output and nonlinear NNOE Model Output
(NNOE [2,2] ; Number of hidden nodes = 10)
10 30
utput response-[degree]
10
eg
se-[d
O
-10
on
0
tp sp
ut re
-20
Ou
-10
-20
-30
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (samples)
Error of 2nd Link PAM manipulator Output and predictive NNARX Model Output - (y-y hat) -30
0 2 4 6 8 10 12 14 16 18 20
0.2
Error- [degree]
0.1 Error of 2nd Link PAM manipulat or Output and predict ive NNOE Model Output - (y -yhat)
2
gree
]
0 1
rror-[de
-0.1 0
E
-1
-0.2
0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 2 4 6 8 10 12 14 16 18 20
time (samples) time - [s ec]
Fig. 10: PAM manipulator NNARX Model results Fig. 12: PAM manipulator NNOE Model results
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-102-
Track 1: AUTOMATIC CONTROL
10
3
(NNARX[2,2] ; Number of hidden nodes = 10) utilized in modeling and identification of PAM
10
2
manipulator in this paper, have overcome
10
1
successfully the nonlinear characteristic of the
Criterion of fitness
0
prototype PAM manipulator system and
10
-1
resulting recurrent neural networks models
10
would surely enhance the control performance
-2
10
of PAM manipulator, due to the extraordinary
10
-3
0 10 20 30 40 50
Iteration
60 70 80 90 100 capacity in learning nonlinear characteristic.
2nd Link PAM manipulator Output and nonlinear NNARX Model Output
(NNARX [2,2] ; Number of hidden nodes = 10)
These results can be applied to model and
30
PAM manipulator response
NNARX model [2,2] response
identify not only of the PAM manipulator but
20
also of other nonlinear and time-varied
parametric industrial systems.
Output response - [degree]
10
-10
Acknowledgment
This research was supported by the BK21
-20
(Republic of Korea)
-30
0 200 400 600 800 1000 1200 1400 1600 1800 2000
time (samples)
0.2
Error of 2nd Link PAM manipulator Output and predictive NNARX Model Output - (y-yhat)
REFERENCES
Error - [degree]
0.1
-0.1
International Symposium on Electrical & Electronics Engineering 2007 - Oct 24, 25 2007 - HCM City, Vietnam
-103-