Radial Basis Function (RBF) Neural Network

Control for Mechanical Systems
 

Jinkun Liu

Radial Basis Function (RBF)

Neural Network Control
for Mechanical Systems

Design, Analysis and Matlab Simulation

With 170 figures

 

Jinkun Liu
Department of Intelligent System
and Control Engineering
School of Automation
Automation Science
and Electrical Engineering
Beihang University
Beijing, China, People’s Republic

TUP ISBN 978-7-302-30255-1

IS
ISBN
BN 978-
978-3-
3-64
642-
2-34
3481
815-
5-0
0 IS
ISBN
BN 978-
978-3-
3-64
642-
2-34
3481
816-
6-7
7 (eBo
(eBook
ok))
DOI 10.1007/978-3-642-34816-7
Springer Heidelberg New York Dordrecht London

Library of Congress Control Number: 2012955859

# Tsinghua University Press, Beijing and Springer-Verlag Berlin Heidelberg 2013

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the mate
material
rial is concerned
concerned,, specifical
specifically
ly the right
rightss of tran
translati
slation,
on, reprinting
reprinting,, reus
reusee of illust
illustrati
rations,
ons,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or 
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar 

methodology
in connection nowwithknown
reviewsororhereafter
scholarlydeveloped.
analysis orExempted from thisspecifically
material supplied
material legal reservation
for theare brief excerpts
purpose of being
entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication
of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publis
Pub lisher
her’s
’s locati
location,
on, in its cur
curren
rentt versio
version,
n, and per
permis
missio
sionn for use mus
mustt always
always be obtain
obtained
ed from
from
Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center.
Violations are liable to prosecution under the respective Copyright Law.
The use of gen gener
eral
al des
descri
cripti
ptive
ve nam
names,
es, reg
regist
istere
eredd names,
names, tradem
trademark
arks,
s, servic
servicee marks
marks,, etc.
etc. in this
this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of 
publication
public ation,, neither
neither the authors
authors nor the editors
editors nor the publisher can accept any legal responsibili
responsibility
ty for 
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.

Printed on acid-free paper 

Springer is part of Springer Science+Business Media (www.springer.com)

 

Preface

Recent years have seen a rapid development of neural network control techniques
and their successful applications. Numerous theoretical studies and actual industrial
implementations demonstrate that artificial neural network is a good candidate for 
function approximation and control system design in solving the control problems
of complex nonlinear systems in the presence of different kinds of uncertainties.
Many control approaches/methods, reporting inventions and control applications
within the fields of adaptive control, neural control, and fuzzy systems, have been
published in various books, journals, and conference proceedings. In spite of these
remarkable advances in neural control field, due to the complexity of nonlinear 
systems, the present research on adaptive neural control is still focused on the
development of fundamental methodologies.
The advantage of neural networks is that a suitable number of neural network
functio
func tions
ns can model
model any (suf (suffici
ficient
ently
ly smoo
smooth)
th) continuo
continuous
us nonline
nonlinear
ar func
function
tion in
a compact set, and the modeling error is becoming smaller with an increase of neural
network
netw ork func
functio
tions.
ns. It is even
even possible
possible to model disc
disconti
ontinuou
nuouss nonline
nonlineari
arities
ties assuming
assuming
the right choice of discontinuous neural network functions. Thus, an adaptive neural
network approach is most suitable in an environment where system dynamics are
significantly changing, highly nonlinear, and in principle not completely known.
This book is motivated by the need for systematic design approaches for stable
adap
ad aptitive
ve co
contr
ntrol
ol sysyst
stem
em desi
design
gn usi
using
ng ne
neur
ural
al netw
network
ork appr
approx
oxim
imatatio
ion-
n-ba
based
sed
techniques. The main objectives of the book are to introduce the concrete design
method and Matlab simulation of stable adaptive RBF ( Radial Basis Function) Function)
neural control strategies.
It is our goal to accomplish these objectives:
• Off
Offer
er a cat
catalo
alog
g of imple
implement
mentabl
ablee neural
neural netw
network
ork control
control design
design method
methodss for 
engineering applications.
• Provi
Provide
de advance
advancedd neural net
network
work controller
controller design
design metho
methodsds and their stabilit
stability
y
analysis methods.
• For ea
each
ch neural network
network control
control algori
algorithm,
thm, we offer its
its simula
simulation
tion example
example and

Matlab program.

v
 

vi Preface

This book provides the reader with a thorough grounding in the neural network
contro
con troll system
system des
design
ign.. Typ
Typica
icall neural
neural networ
network
k contro
controlle
llers’
rs’ designs
designs are verifie
verifiedd
using Matlab simulation. In this book, concrete case studies, which present the
result
resultss of neu
neural
ral net
networ
work k contro
controllellerr implem
implement
entati
ations
ons,, are used
used to ill
illustr
ustrate
ate the
successful application of the theory.
The book is structured as follows. The book starts with a brief introduction
of adaptive control and neural network control in Chap.   1, RBF neural network
algorithm and design remarks are given in Chap. 2 Chap.  2,, RBF neural network
network controller 
controller 
design based on gradient descent algorithm is introduced in Chap.   3, since only
local optimization can be guaranteed by using the gradient descent method, and
seve
severa
rall ad
adap
aptiv
tivee RBF
RBF neur
neuralal netw
networkork co
cont
ntro
roll
ller
er desi
design
gnss ar
aree gi
give
ven n based
based on
Lyapunov analysis from Chaps.   4,   5,   6,   7   and   8, which include simple adaptive
RBF neural network controller, neural network sliding mode controller, adaptive
RBF controller based on global approximation, adaptive robust RBF controller 
based
base d on local
local app
approxi
roximat
mation,
ion, and bac backste
ksteppin
ppingg adapti
adaptive
ve contro
controlle
llerr with
with RBF.
RBF.
In Chap. 
Chap.   9, digital RBF neural network controller design is given. Two kinds of 
discrete neural network controllers are introduced in Chap.   10 10.. At last, a neural
network adaptive observer is recommended and a speedless sliding mode controller 
design is given in Chap. 11
Chap.  11..
I would like to thank Prof. S. S. Ge for his fruitful suggestions. I wish to thank
my family for their support and encouragement.
In this book, all the control algorithms and their programs are described sepa-
rately and classified by the chapter name, which can be run successfully in Matlab
7.5.0.342 version or in other more advanced versions. In addition, all the programs
ca
can
n be down
downlo
load
aded
ed viviaa http
http:/
://l
/ljk
jk.b
.bua
uaa.e
a.edu.
du.cn/
cn/.. If you have
have quest
question
ionss abou
aboutt
algorithms
algorithms and simula
simulation
tion programs,
programs, pleas
pleasee E-mail [email protected]
[email protected]..
 

Contents

1 In
Intr
trod
oduc
ucti
tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Neururaal Ne
Network Cont ntrrol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.
1.1.
1.1
1 Wh Why yN Neu
eurarall Net
Netwo
work rk Cont
Contro rol?
l? . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Re Revview of Neural Networ orkk Contro roll . . . . . . . . . . . . . . . . 2
1.
1.1.
1.3
3 Re Revi
view
ew of RBF RBF Ad Adap apttive
ive Con
Contrtrol
ol . . . . . . . . . . . . . . . . . . 3
1.2 Review of RBF RBF Neural N Neetwork . . . . . . . . . . . . . . . . . . . . . . . . 3
1.
1.3
3 RBF
RBF Ad Adapapttive
ive Co
Contntro
roll fo
forr Robo
Robott Man anip
ipul
ulat
ator
orss . . . . . . . . . . . . . 4
1.
1.4
4 S Func
Functtion
ion Des
Desig ign
n fo
forr Cont
Contro roll Sysyste
tem
m. . . . . . . . . . . . . . . . . . . . 5
1.4.1 S Function IIn ntrod
roduction . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4.2 Ba Basisicc Paramete terrs in S Funct nction . . . . . . . . . . . . . . . . . . . 5
1.4.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5
1.5 An Ex Examamplplee of
of a Simp
Simple le Ad
Adapapti
tive
ve CoContntro
roll Syst
Systemem . . . . . . . . . . . 7
1.5.1 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.2 Adaptive Cont ntrrol Law Desi siggn . . . . . . . . . . . . . . . . . . . . 7
1.5.3 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Append
ndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Refere
ren
nces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 RBF N
Neur
eural
al NNetw
etwork
ork D Desi
esign
gn an
and d Simul
Simulatiation
on . . . . . . . . . . . . . . . . . 19
2.1
2.1 RBF
RBF N Neu
eurarall Net
Netwo
work rk Desi
Design
gn anandd Sim
Simul ulat
atio
ionn. . . . .. . . . . . . . . . 19
2.1.1 RBRBF F Algor oriithm
thm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.
2.1.
1.2
2 RBRBF F DeDesi
sign
gn ExExamampl
plee wi
with
th Ma
Matltlab
ab SiSimu
mula lati
tion
on . . . . . . . . 20
2.2 RBF Neur
Neural al Ne
Netwo
twork rk App
Approx
roxima
imatio
tionn Based
Based on Gradie
Gradient nt
Desce
scent MeMethod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.
2.2.
2.1
1 RBRBF F NeNeur
ural
al Ne
Netwtwor
ork
k App
Approroxi
xima
matition
on . . . . . . . . . . . . . . . 22
2.2.2 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Effect
Effect ooff Gau
Gaussia
ssiann Functi
Functionon PaParam
ramete
eters
rs on RBF
RBF
Approxi
roximmation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4
2.4 Effe
Effect
ct of Hidd
Hiddenen Ne
Nets ts Nu
Numb
mber er on RB
RBF F Appr
Approx oximimatatio
ionn. . . . . . . 28

vii
 

viii Contents

2.5
2.5 RBF
RBF Ne Neurural
al Ne
Netwtwororkk Tra
Traininin
ingg fo
forr Sys
Systetemm Mod
Model elin
ingg. . . . . . . . . 33
2.
2.5.
5.11 RB RBF FN Neeur
ural
al Netetwo
workrk Trarain
inin
ingg. . . . . . . . . . . . . . . . . . . . 33
2.5.2 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.6
2.6 RBF
RBF N Neueura
rall Net
Netwoworkrk Appr
Approx oxim
imatatio
ionn. . . .. . . .. . . .. . . .. . . .. 36
Append
ndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Refere
ren
nces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3 RBF NNeur
euralal Ne
Netwo
twork rk Contr
Control ol Based
Based on Grad Gradien ientt
Descent Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1
3.1 Supe
Supervrvis
isor
oryy Cont
Contro roll Bas
Baseded on RB RBF FN Neueura
rall Net
Netwoworkrk . . . . . . . . . . 55
3.
3.1.
1.11 RB RBF FS Sup
uper
ervi
viso
sory
ry Cont
Contro roll . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.2 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2
3.2 RBFN
RBFNN- N-BaBase
sedd Mod
Model el Re
Refefere
renc
ncee Ada
Adaptptiv
ivee Con
Contr trol
ol . . . . . . . . . . 58
3.2.1 Co Con ntroller De Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.2 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.
3.3
3 RBF
RBF S Sel
elf-
f-Ad
Adjjust
ust C Con
ontrtrol
ol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.1 System De Descripti tio
on . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.2 RB RBF FC Coont
ntrroller DeDesisign
gn . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.3 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Append
ndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Refere
ren
nces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4 Ada
Adapti
ptive
ve RBF
RBF N Neur
euralal NNetw
etwork
ork Contro
Controll . . . . . . . . . . . . . . . . . . . . . 71
4.1
4.1 Adap
Adapti tive
ve Co
Cont ntro
roll BBas
ased
ed on Ne Neurural
al Ap
Apprprox
oximimatatio
ionn. . . . . . . . . . . 71
4.1.1 Problem D Deescripti tioon . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.
4.1.
1.22 Ad Adapaptitive
ve RB
RBF FC Con
ontrtrol
olle
lerr Des
Desig ign
n. . . . . . . . . . . .. . . . . . 72
4.
4.1.
1.33 Simu
Simullat atio
ionn Exa
Exampmple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 Adapti
Adaptive ve Contro
Controll Based
Based on Neur Neural al Ap
Approx
proximaimatio
tionn
with
ith UUnk
nknonow wn P Paara
ramemetter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.2.1 Problem D Deescripti tioon . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.
4.2.
2.22 Ad Adapaptitive
ve Co
Cont ntro
roll
ller
er Des esiign . . . . . . . . . . . . . . . . . . . . . . 79
4.
4.2.
2.33 Simu
Simullat atio
ionn Exa
Exampmple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3
4.3 A DirDirecectt M
Metetho
hodd ffor
or Robu
Robust st Adap
Adapti tive
ve Cont
Contro roll by
by RBF
RBF . . . . . . . . 83
4.3.1 System De Descscrripti
tio
on . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3.2
4.3 .2 DesDesireired
d Feedbac
Feedback k Con
Controtroll and
and Functi
Function on
Approximat atiion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.
4.3.
3.33 Co Contntro
roll
ller
er Desi
Designgn anand d Pe
Perfrfor
orma
manc ncee Anal
Analys ysis
is . . . . . . . . . 87
4.3.4 Simulation Ex Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Refe
Re fere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
112
5 Neu
Neural
ral Net
Networ
work kS Slid
liding
ing Mode
Mode Cont
Control rol . . . . . . . . . . . . . . . . . . . . . . 113
5.1
5.1 Typi
Typica
call Sli
Slidi
ding
ng Mod
Modee CCon
ontr
trol
olle
lerr Des
Desig ign
n. . . . . . . . . . . . . . . . . . 114
114
5.2 Slidin
Sliding
g Mode C Cont
ontrol
rol Ba
Based
sed on RBF RBF for SeconSecond-Or
d-Order 
der 
SIS
ISOONNononli
line
near
ar Syste
ystemm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
116

5.
5.2.
2.1
1 Prob
Proble
lem
m Des
Descr
crip
ipti
tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
116
 

Contents ix

5.2.2
5.2.2 Sli
Slidin
ding g Mo
Modede C Cont
ontrol
rol Based
Based o on
n RBF
RBF
Unknown  f ðÞ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
for Unknown f  117
5.
5.2.
2.33 Simu
Simullat atio
ionn Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
118
5.3 Slidin
Sliding g Mod
Modee ConControtroll Based
Based o onn RBF for Unkno Unknown wn f  f ()
and g
and  g(() . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.3.1 Int Introdu
oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.
5.3.
3.22 Simu
Simullat atio
ionn Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
122
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Refe
Re fere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
132
6 Ada
Adapti
ptive
ve RBF
RBF Cont Controlrol Based
Based on Glob Global al Approx
Approxima imatiotionn . . . . . . . . 133
6.1 Adaptive
Adaptive Control
Control with RBF Neural Network Network C Compensa
ompensation tion
fo
forr Ro
Roboboti
ticc Mani
Manipu pula
lato
tors
rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
134
6.
6.1.
1.11 Prob
Proble lemm Des
Descrcrip
ipti
tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
134
6.
6.1.
1.22 RB RBF FA Appppro
roxi
xima
mati tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
135
6.1.3
6.1 .3 RBF C Cont
ontrol
roller
ler and
and Adapt
Adaptive ive LLawaw Desi
Designgn
and An Analysi siss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.
6.1.
1.44 Simu
Simullat atio
ionn Exa
Examp mple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
140
6.2 RBF Neur Neural al Rob
Robot ot Contro
Controlle llerr Desi
Design gn with
with Sliding
Sliding ModeMode
Robus
bust TeTerm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.
6.2.
2.11 Prob
Proble lemm Des
Descrcrip
ipti
tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
144
6.
6.2.
2.22 RB RBF FA Appppro
roxi
xima
mati tion
on . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
147
6.
6.2.
2.33 Co Cont ntro
roll L
Law
aw De Desisign
gn and
and S Sta
tabi
bili
lity
ty Anal
Analysysisis . . . . . . . . . . . 147
147
6.
6.2.
2.44 Simu
Simullat atio
ionn Exa
Examp mple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
148
6.3
6.3 Robus
Robustt Cont
Controlrol Base
Based d on
on R RBF
BF Ne Neura
urall NNet
etwo
workrk with
with HJIHJI . . . . . . 153
153
6.3.1 Foundation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.
6.3.
3.22 Co Cont ntro
roll
ller
er Desi
Design gn anand d Ana
Analy lysi
siss . . . . . . . . . . . . . . . . . . . 153
153
6.
6.3.
3.33 Simu
Simullat atio
ionn Exa
Examp mple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
156
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Refe
Re fere
renc
nceses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
191
7 Adapti
Adaptive
ve Robust
Robust RBF Contro Controll BaseBased d on Local Approximati
Approximation on . . . 193
7.1 Robust
Robust C Cont
ontrol
rol Based
Based o onnNNomi
ominalnal Mod
Model el
fo
forr Ro
Robo
boti
ticc Mani
Manipu pula
lato
tors
rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
193
7.
7.1.
1.1
1 Prob
Problelemm DesDesccri
ript
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . 193
193
7.
7.1.
1.2
2 CoContntro
roll ller
er Desi
Designgn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
194
7.
7.1.
1.3
3 Sta
Stabi
bili
litty AAna
nallys
ysis
is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
195
7.
7.1.
1.4
4 Simu
Simullatatioion
n Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
196
7.2 Adapti
Adaptiveve RBF ControControll Based
Based on Local Local Model
Model Approxi
Approximat mationion
fo
forr Ro
Robo
boti
ticc Mani
Manipu pula
lato
tors
rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
197
7.
7.2.
2.1
1 Prob
Problelemm DesDesccri
ript
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . 197
197
7.
7.2.
2.2
2 CoCont
ntro
roll ller
er Desi
Designgn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
199
7.
7.2.
2.3
3 Sta
Stabi
bili
litty AAna
nallys
ysis
is . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
200
7.
7.2.
2.4
4 Simu
Simullatatioion
n Exa
Exampmple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
203
 

x Contents

7.3 Adaptive
Adaptive Neural
Neural Network
Network Control
Control of RobotRobot Manipulator
Manipulatorss
in TTas
askk Spa
Spacece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
205
7.3.1 Coordin
Coordinationation Transformatio
Transformation n from Task Space
to Jo
Join
intt Spa
Spacce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
208
7.3.
7.3.22 Ne Neurural
al Ne
Netwo
tworkrk M Mode
odelin
lingg of Robo
Robott ManManip ipula
ulato
tors
rs . . . . . . 208
208
7.
7.3.
3.33 Co Contntro
rollle
lerr Des
Desigignn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
210
7.
7.3.
3.44 Simu
Simullat atio
ionn Exa
Exampmple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
213
Append
ndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Refe
Refere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
249
8 Bac
Backste
ksteppi
pping
ng Control
Control with with RBF . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
8.1 Introduc
duction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
8.2
8.2 Back
Backst step
eppi
pingng CCon
ontrtrol
ol for
for Inv
Invererte
ted
d Pend
Pendul ulum
um . . . . . . . . . . . . . . . 252
252
8.
8.2.
2.1
1 Sy Syststem
em Des esccri
ript
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
253
8.
8.2.
2.2
2 Co Contntro
rollle
lerr Des
Desig ignn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
253
8.
8.2.
2.3
3 Simu
Simullat atio
ionn Exa
Examp mple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
254
8.3 Backst
Backsteppepping
ing Contro
Controll BaseBased d on RBF for Inverte Inverted d Pend
Pendulum
ulum . . . . 255
8.
8.3.
3.1
1 Sy Syststem
em Des esccri
ript
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
255
8.
8.3.
3.2
2 Ba Backckststep
eppi
ping
ng CoContntro
rollller
er De
Desisign
gn . . . . . . . . . . . . . . . . . . . 25
256
6
8.
8.3.
3.3
3 Ad Adapaptitive
ve La
Law wD Desesig
ignn. . . . . . . . . . . . . . . . . . . . . . . . . . 257
257
8.
8.3.
3.4
4 Simu
Simullat atio
ionn Exa
Examp mple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
259
8.4 Backst
Backsteppepping
ing Contro
Controll forfor Singl
Single-Le-Link
ink Flexibl
Flexiblee Joint
Joint Robot
Robot . . . . . 260
8.
8.4.
4.1
1 Sy Syststem
em Des esccri
ript
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
260
0
8.
8.4.
4.2
2 Ba Backckststep
eppi
ping
ng CoContntro
rollller
er De
Desisign
gn . . . . . . . . . . . . . . . . . . . 262
262
8.5 Adaptive
Adaptive Backste
Backstepping
pping Control with RBF for for Single-
Single-Link
Link
Flexi
lexibl
blee Joi
Jointnt Ro
Robo bott . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
265
8.5.1 Backst
Backstepping
epping Control
Controller ler Design with Function
Function
Esttim
Es imatatio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
265
8.5.2
8.5 .2 BacBackstkstepp
epping
ing C Cont
ontrol
roller
ler Desi
Designgn with
with RBF
RBF
Appr
Ap proxoxim
imatatiion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
269
8.
8.5.
5.3
3 Simu
Simullat atio
ionn Exa
Examp mple less . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
272
Append
ndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
Refe
Refere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
291
9 Dig
Digita
itall RBF
RBF Ne Neura
urall NNetw
etwork
ork Contro
Controll . . . . . . . . . . . . . . . . . . . . . . . 293
9.1
9.1 Adap
Adapti tive
ve Ru
Rung nge–
e–KuKutttta–
a–Me
Mers rson
on Me
Meth thod
od . . . . . . . . . . . . . . . . . . 293
293
9.1.1 Int Introdu
oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
9.
9.1.
1.2
2 SimuSimullatatio
ionn Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
295
9.2
9.2 Digi
Digitatall Adapt
Adaptiv ivee Cont
Controroll fo
forr SISO
SISO Sy Syststem
em . . . . . . . . . . . . . . . . . 295
295
9.2.1 Int Introdu
oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
9.
9.2.
2.2
2 SimuSimullatatio
ionn Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
297
9.3
9.3 Digi
Digitatall Adapt
Adaptiv ivee RBF
RBF Cont
Controlrol fo
forr Two-
Two-Li Linknk Man
Manipu ipulalato
tors
rs . . . . . 298
298
9.3.1 Int Introdu
oduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
9.
9.3.
3.2
2 SimuSimullatatio
ionn Exa
Exampmple le . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
299
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Refe
Refere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
309
 

Contents xi

10 Discre
Discrete
te Neu
Neuralral Net
Networ
work kC Cont
ontrolrol . . . . . . . . . . . . . . . . . . . . . . . . . . 311
10.1 Introdroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
10.2
10.2 Direct
Direct R RBF
BF Co Contrntrol
ol for
for a Cla
Class ss of Discre
Discrete- te-Tim
Timee
Nonl
onlinea
inearr Sys
Systtem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
312
10.2
10 .2..1 Sys yste
temm Des
Descrcriipt
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . 312
312
10.2
10 .2.2
.2 Co Contntro
roll
ller
er Des
Desigignn and
and S Sta
tabi
bilit
lity
y Anal
Analysiysiss . . . . . . . . . . . 312
312
10.2
10 .2..3 Simulimulat atio
ionn Exa
Exam mpl ples
es . . . . . . . . . . . . . . . . . . . . . . . . . 316
316
10.3
10.3 Adapti
Adaptive ve RBF C Contr
ontrol
ol for a ClasClasss of Discrete
Discrete-Ti -Time
me
Nonl
onlinea
inearr Sys
Systtem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
319
10.3
10 .3..1 Sys yste
temm Des
Descrcriipt
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . 319
319
10.3
10 .3.2
.2 Tr Trad
adit
itio
iona
nall Con
Contr trol
olle
lerr Des
Desigignn. . . . . . . . . . . . . . . . . . . 320
320
10.3
10 .3.3
.3 Ad Adapaptitive
ve Ne
Neura
urall N Net
etwo
workrk Contr
Control olle
lerr Des
Desig ignn. . . . . . . . 320
320
10.3
10 .3..4 Sta tabi
billit
ity
yAAnanaly
lysi
siss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
322
10.3
10 .3..5 Simulimulat atio
ionn Exa
Exam mpl ples
es . . . . . . . . . . . . . . . . . . . . . . . . . 324
324
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Refe
Refere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
337
11 Adapti
Adaptiveve RBF Obser Observer ver DeDesig
signn and Slidin
Sliding g Mode Co Contr
ntrolol . . . . . . 339
11.1
11.1 Adap
Adapti tive
ve RB
RBF F Obs
Obsererve
verr Des
Desig ignn. . . . . . . . . . . .. . . . . . . . . . .. 339
339
11.1
11 .1..1 Sys yste
temm Des
Descr criipt
ptio
ionn. . . . . . . . . . . . . . . . . . . . . . . . . . 339
339
11.1
11 .1.2
.2 Ad Adapaptitive
ve RB
RBF FO Obse
bservrver
er Desi
Designgn and
and Anal
Analysysisis . . . . . . . 340
340
11.1
11 .1..3 Simulimulat atio
ionn Exa
Exam mplples
es . . . . . . . . . . . . . . . . . . . . . . . . . 343
343
11.2
11.2 Sl Slidi
idingng Mo
Modede CoContr
ntrol
ol Ba
Basesedd on
on RRBFBF Adapt
Adaptiv ivee Obse
Observ rver
er . . . . . 347
347
11.2
11 .2.1
.1 Slid
SlidininggMMododee Con
Contrtrol
olle
lerr Des
Desig ignn. . . . . . . . . . . . . . . . . 347
347
11.2
11 .2..2 Simulimulat atio
ionn Exa
Exam mplplee . . . . . . . . . . . . . . . . . . . . . . . . . . 349
349
Appendndiix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Refe
Re fere
renc
nces
es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
362

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
 

Table of Notation

Notation Meaning
 R   The set of real numbers
numbers
n
 R The set of all n
all  n-dimensional
-dimensional real vectors
nm
 R all   n   m-dimensional
The set of all   m-dimensional real matrices
jaj   The absolute
absolute value of scalar   a
detð AÞ   The determinant
determinant of matrix   A
k xk   The norm of vectovector  r   x
T 
 A The transpose of   A
1
 A The inverse of   A
 I    An identi
identity ty matrix
 I n   The identity
identity matrix of dimen sion   n   n
dimension
 λi ð AÞ   The i
The  ith
th eigenvalue of   A
 λð AÞ   The set of eigen
eigenvalue
valuess of   A
 λmin ð AÞ   The minimum eigenvalue of   A  where   λðAÞ  are real
 λmax ð AÞ   The maximum eigenvalue of   A  where   λð AÞ  are real
 xi   The i
The  ith
th element of vector   A
aij    The ij 
The  ij th
th element of matrix   A
()
^
The estimate of   ðÞ
()   ðÞ-()
~ ^

sup   α ðt Þ   The smallest

smallest number
number that is larger than or equal to the maxim um value of   α ðt Þ
maximum
diag½     Diagonal matrix with given diagonal elements
h   Output
Output vector
vector of Gaussian
Gaussian func
function
tion of RBF
c j    Center
Center vector
vector of the j 
the j th
th net of Gaussian function of RBF
b j    Width of the j 
the j th
th net of Gaussian function of RBF
W    Weight
Weight vector
vector of RBF

xiii
 

List of Acronyms

BP back-propagation
GL Ge–Lee matrix
LIP
LIP li
line
neaar-
r-in
in-t
-the
he-p
-par
aram
ameete
ters
rs
MFN
MF N mult
multililay
ayer
er fe
feed-
ed-fo
forw
rwar
ardd netw
network
ork
MIMO
MIM O multi-
multi-inp
input
ut mul
multi-
ti-outp
output
ut
MNN
MN N mult
multililay
ayer
er ne
neura
urall netw
network
orkss
NN neural networks
RBF
RB F radi
radial
al ba
basi
siss fu
func
ncti
tion
on
RKM
RK M Ru
Rung
nge–e–Ku
Kutttta–
a–Me
Mers rson
on
SISO
SISO singl
single-
e-inp
input
ut si
sing
ngle
le-o
-outp
utput
ut
SMC
SM C slid
slidin
ingg mode
mode co cont
ntro
roll

xv