Академический Документы
Профессиональный Документы
Культура Документы
Xk+l
=/u(t,x,u)
with:
The x state vector is made of two parts; the first one represents the continuous state and the
second, the discrete one.
504 Appendices
Function called
~ C-MEXflle
mdlInitializeSizes
mdlGetTimeOfNext VarHit
3
2
1
9
mdlOutputs
mdlUpdate
mdlDerivatives
mdlTerminate
~Mflle
flat!: state
Initialization
Next sampling step calculation
(OPtional)
Outputs calculation
Discrete state updating
Continuous state updating
End of simulation
= function_name(t,x,u,flag,Pl,P2,
",)
default drive variables passed onto the S-function, through the system,
: optional variables wished to be passed onto the S-function.
The variables returned to the system, according to time, depend onto the x, u vectors and the
flag indicator.
flag = 0, the initialization function has to return to sys: the size-system
informations, to xo: the system's initial state; to str an empty chain (for the m functions)
and to ts the sampling instants. In this purpose the mdlInitializeSizes function
may be used in order to initialize a simsize-type structure containing the following
fields that will be affected onto the sys variable.
NumContStates
NumDiscStates
NumOutputs
Numlnputs
DirFeedthrough
NumSampleTimes
%
%
%
%
%
%
flag
flag
flag
flag
A = [-:.5
-1~25]
C =[0 1.5625]
=[~]
D=O
Calling this S-function is done without additional parameters passing. An input signal under
unit step form is applied.
Mux
The ex1 . m file tests the flag and calls a function for each case used.
flag = 0, S-functions parameters of initialization function call:
- number of continuous states,
- number of discrete states,
- number of inputs,
- number of outputs,
- number of sampling instants,
- initial state,
- sampling offset and time.
The DirFeedthrough field of the initialization structure is positionned at 1, when the D
matrix is used or when the case flag = 4 is being used.
506 Appendices
The (1,2) dimension ts vector, sampling offset and period is at zero for a [0 0] continuous
process. For a discrete use with a variable sampling (flag = 4 case use) this vector has to
be initialized with [-2 0].
flag = 1, call of the continuous state calculation function,
x=Ax+Bu
flag
y=Cx+Du
exl.mfile
function [8Y8,xO,.tr,t.}
% continuou. atat. mod.l
A= [-1.5 -1.25 1 ,
B
[2 , 0];
C
[0 1.5625};
D= 0,
=
=
0];
8Witch flag
0...
=OUtput.Calcul (x, u, c, D) ;
ca {2,',9}
Y.
=[] 1
% unu.e4 .tag
oth.nri
arror(['ODban d 1 .4 flag. ',num2atr(flag)});
and
% Initiali.ation function
funotion [.y.,xO,.tr,ta] Initialization(A,B,C,D)
if D 0
Dir" 0,
.1
Dir" 1;
and
.i a ~i ,
% .~iz type .tructure
.iz NUmContstat langtb(A); % oontinuou. atat.a number
aiz . NUmDiacstat 0;
% 4iacr.t tat number
.ize . HumDutput 1;
.ize RUmXnput 1;
.ize Dirreedthrough DirP;
.ize HumSampleT1me 1;
.Y tm.i C.ize.);
xO .ero.C.i CS;
.tr [) I
t [0 0]; "aontinuou. proae
3.5
.---.......----.--......,,......-.,...--r---.--~-_.._-__'T-___,
3
2.5
0.5
O~--~--~---L--~----~--~--~---L--~--~
5
Time In s
10
In ex2 . m we use a S-function to simulate the previous example discretized process with a Te
sampling period of 0.1 s.
508 Appendices
_ [1.815 - 0.86]
Ad -
Cd
=[0.237 0.225]
Bd -_ [0.25]
This S-function call is done with Te and Xi additional parameters passing. respectively
representing the sampling period and the state initial conditions. An input signal of unit step
type is being applied.
s _lun c toon2
EJ
ex2.mfile
% Discrete state med.1 simu1atiaD
Ad Bd -
IlWitcb flag,
cas. 0
% Xnitialization stag.
[sys,xO,str,ts) Xnitia1ization(Ad,Dd,T.,Xi);
cas. 2
% State calculation stage
sya Derivat Ca1cu1 (x,u,Ad,Bd);
cas. 3
% OUtput calculation stage
sya OutputsCalcu1 (x,u,Cd,Dd);
ca (1,4,9)
% unus.d stag.s
ays [);
otherwise
error(['0Dband1ed flag ',n~str(flag)]);
end
Xniti.11 tion(Ad,Dd,T.,X1)
.1
DirJ'
end 1
.i tm.i ,
. i. HUmContSt.t 0,
. i. JlIUDT!\P'lt 1;
. i. HumSamp1.Ttm. 1,
~ tm.iz (.iz ),
xO XiI
% .t.t. initi.l cODditiona
.tr [11
t [T. 0]; % offtl T ampling period
3,5 .----r---r-__,.----r---r---,r--"T"""-""T"---,----,
~ ~---------
>
0,5
/"l,J
011--~ , 5~-~-~--L--~-~--L--~-~-~L-~
5
Time In s
10
510 Appendices
Example 3, use 0/ a So/unction to simulate a discrete state model with variable sampling
period
We use a ex3 . m S-function to simulate the previous example discrete process to which we
add an input defining the process sampling period.
This S-function call is done with Xi additional parameter passing, representing the state initial
conditions. The sampling period, measured in seconds is chosen as being equal to the voltage
amplitude (measured in volts), applied onto the S-function second input.
. s.funchon3
s-runcUon example
Variable sampling, discrete stale system
~~~----------------~M~
The flag = 4 case is used to calculate the future sampling instant. In order to validate the
using of this case, it is necessary at position to 1 the DirFeedThrough element of the Sfunction initialization structure and to initialize ts = [-2 0].
ex3.mfile
awitch flag
c... 0
% ~tiali.ation atag.
(aya,xO tr.ta) ~tialization(Ad,Xi);
ca.. 2
% State calculation atag.
aya Derivat.aCalcul (x,u,Ad,ad);
ca.. 3
% OUtput calculation atage
aya OUtputaCalcul (x,u,C4,D4);
c... ,
% Next T. oalculation atage
aya NextT.Ca1cul (t,u);
ca (l,9) % enua.d atag.a
ay [);
oth.:nd
% Initi.lis.tion function
function [..v.,zO,.tr,t.] - XDiti.li tion(Ad,Xi)
.i - .tm.i ,
.i NUmContSt.t - 0;
.i NUmDi.cSt.t - length(Ad),
.i BUmOutput 1;
. i . RUmXnput. - 2;
3.5r---'---~---'~--.----r--~----~--'---~--~
3
2.5
>
.!:
"
!!
>
1.5
(;
0.5
o
-0.5 '--__.......__-'-__--''--__.L-__-L._ _---..1____' ' -__- ' -__- ' -__---l
023
4
5
6
7
8
9
10
Time in s
512 Appendices
ex4.cfile
'*
SetRumCoatState.(S,O) ,
'* Coatiauou tate. DUmber *'
SetRUmni.cState.(S,2),
'* ni.crete .tate. DUmber *'
SetNUmIDput.(S,l) ,
Input. ZlUIIIber
*,
SetBUmOutput.(S.l) ,
OUtput. mDIIber
SetoirectreedThrough(S.O) ,
direct feedthrough flag *'
Se~DDBampleTtm..(S,l)1
1* .~ling inatant. mDIIber
SetRumBrCDPar. . . (S,O) ,
Se~ork(8,0)1
setHumIWork(S,O) ,
SetNumPWork(S,O) ,
'*'*'*
*'
*'
SetSampleTime(S, 0, 0.1);
SetOff.etTime(S, 0, 0.0),
)
int il
for (i 0, i<21 i++)
(
.0++ 0,
)
)
'*
{
*'
c
)
'*
Ien41f
In order to compile this file, it is necessary to specify the type of compiler and its installation
directory. For this, you have to execute the cmex -setup command and to answer the
questions asked. In return, the mexop t s . ba t file is created and contains the paths,
commands and options of compilation and linkage.
514 Appendices
The .C file may then be compilated, thanks to the help of the cmex.ba t file to be found in
the <MATLAB>\BIN directory where <MAT LAB> designates the MATLAB installation
directory.
Appendix 2
= a e-k t sin(wt-tp)
G~in
kt
Oot ProduGl:
...
516 Appendices
The sinus and exponential operators are defined by the same MATLAB Function
block of the Nonlinear library.
After mouse selecting the whole content of the figure, the Create Subsystem option of
the Edi t menu enables to create the equivalent subsystem.
By double-clique onto the obtained block, we retrieve the system's detail, to which are added
the input and output ports. It is necessary, like shown here above, to name these
ports with variables names used, so a, phi, w, k et x in our case.
Subsystem
"
kt
Dol Product
Ready
In this example, the variables characterising the signal are constant inputs of the subsystem
block. The mask edition enable dialog box creation through which the user may enter the
different values of the input variables, editing a help text or drawing curve.
In the previous subsystem, we replace the constant inputs values by variable names; those of
the signal to be generated, in the same constant blocks.
1------"1l>1w
kl
Subryslom
After selecting this subsystem, after suppressing the To Workspace block, we create
another by Crea te Subsys tern of the Edi t menu. This system inputs-less, has one output
named Outl. We double-clique onto this subsystem to change its label in x.
~----~'~I----~
Su~om
In this block, the used variables and the type of the used function don't appear.
Its masking may to personalise it by drawing onto it a sinusoidal curve and creating a dialog
box in order to specify the values of the used variables.
After selecting this icon, we choose Mask Subsys tern option of the Edi t menu.
We obtain the Mask Editor window, with its 3 menus, Icon, Initialization and
Documentation. In this window, we define the mask parameters.
518 Appendices
Iconmenu
Mask Type:
Drawing commands:
1. Icon frame
In our example, the curve is drawn for the time values going from
the Autoscale option .
Menu Documentation
In this menu, we define the variables whose will be entered in the dialog box.
Mask type:
help text.
lconlran.~
Icon lotaliorr
~=-=~
The help block text always appears when we clique onto the help button of the dialog box.
In order to validate it, we must clique onto the apply button.
Initialization menu
This menu enables defining the variables, as wall as their signification (prompt) which we
retrieve in the dialog box in which we may enter numerical, chains or Boolean values.
The numeric value of the variable may be directly entered in the edition field or chosen in a
pop menu. The Boolean type is expected in marking (1 logic) or keeping empty (0 logic) the
correspondent check box.
A variable may have chain type value. MA TLAB interprets it using the eval command.
The variable type is controlled by the Control type button by what we can choose:
Edit
Checkbox
check box,
Popup
popup menu.
520 Appendices
The prompt corresponds to the variable signification and in the Variable filed we define
the variable name which will take the specified value in the dialog box.
When we fill the Prompt and Variable fields, this variable signification, name and type
are noticed in the end of parameter list delimited zone
The display order is that it appears like it appears in the dialog box ..
The Down and up buttons are expected for modifying and organising this order. If we want to
delete a variable, we must clique onto the Delete button.
The Assignment button bears 2 choices ; Evaluate signifies one that MATLAB will
directly evaluate and Li teral when this variable is character chain type.
Each variable entering (prompt and name) is done after selecting the Add button.
The following figure illustrates these several options of entering different types variables will
be appear in the dialog box when we have to use the edited mask block.
M d ~k
r r~ltOi
H ra ~
JI I~ II' D~ I
Icon
l r ,"
II.
Add
Type
.-
PIoqIt
'oJ.
p\lUlltiOD
I pbue
II
'l&!
II
Down
I~J [
dulplDg
r.r.
~,~.'
f~
edit
edit
ed1t
rJ
If D. . .
"4IiIbIe
coefficient
:-!:
......
Praqj: J~
Conbal \?PC
jEdl
.f
"... Jo
JEvoUotll
.E
ro,
I,..... COIM\rId$:
....
;.
"
Ilr
:1;1
' OK
'I I!
,'"
Creel
-,
J II
.
Umd.
I lL
Help
J II
AA*.>
The prompt and the variable name definition is made after clicking onto the Add button.
If we want to enter a list of variables whose will be available in a popup menu, we must enter
them in Popup s tr ings zone separating each of them by the I symbol.
An example of popup menu using is provided in the following paragraph in which we
generate a pseudo-random binary sequence (PRBS). The popup menu enables the choice of
sequence number corresponding to its length.
When the mask is created and we want to modify one of the 3 menus properties, we must
execute Edi t Mask option of the Edi t menu.
The Look Under Mask option enables displaying the whole masked blocks.
If we want to generate the following signal :
x(t)=5 e -O.2t
sin(wt-~),
we double-clique onto the masked block in order to inquire the different dialog box fields
(variable values specification).
Ei
P. __ en ~~--~--~~~~~~,--~~~~
~
15
r
piI2
The generated signal by the mask goes to the workspace through the x global variable.
522 Appendices
S ubsystlm
Ir..Q.OXl
i'0d&45
Curve drawing
plot(x)
,. grid
title('4ampe4 .inu. - 5; . . 2 rad/.;\phi \pi/2;k-O.2')
xlahel ( 'time' )
damped sinus - a=5; w = 2 rad/s;
cI>
= rr/2; k=0.2
/\
I\
(\
/ \ ! \\V
I \/
\j
-1
-2
/'--.
-----
IV
-3
-4
-5
10
20
30
40
50
60
time
~ o'h;; - d,;",~cdSln
----
- --
-~
- --- ----,.r-T1E3
Dol P,oduot
This model, more compact that the previous one, bears only one input, the phase ({J. The
other parameters: the pulsation w, the dumping coefficient k and the amplitude a are defined
by the gain blocks.
Like previously, the same dialog box may be used for the correspondent of the obtained
subsystem, obtained by this model.
To initialise the sequence first values to 1 (the 10 firsts of the sequence of length 1023), we
use transfer function blocks, for what we define the initial state and transform them to delays.
These Discrete Transfer (with initial states), are provided in the
Blocksets & Toolboxes / Simulink Extras / Additional Discrete library.
Each sequence is transformed to a subsystem.
524 Appendices
Fiil IE3
To select one of the 6 sequences, we use the sf_select S-function for which we pass the
Te sampling time and k representing the sequence number which we specify in a popup menu
of the final mask dialog box. This S-function bears 6 inputs corresponding to the 6 PRBS
sequences and one output which be the kth sequence.
sf-select.m
% S-function : PUS quenc. ..1.ction
.witch nag
ca.. 0
k 1;
[.y.,xO,.tr,t.] xnitiali.ation(T.);
ca.. 3
.y u(k);
ca (l,2,4,9)
.y [],
oth.rw1
flag
,n~.tr(flag)]);
The outputs of the 6 sequences blocks are multiplexed then multiplied by the a gain
representing the amplitude, before their passing to the sequence selection S-function.
We select the set of blocks that we transform into subsystem which then be masked.
PRSS
526 Appendices
loan
11 IriiIIuIian I,
. .1Ik.
Yr
I PA BS sequence gene!5Of
Tp
-Aaa
~r o
Ilr
O~
VIriIbIe
-,
...---I a.pl1tude a:
DUlbe~
lJlji
-I
Iad o pu_tel:
11lt
I rFI)"_~ 1 n r
Variable:
I Te
Pi
Intialzalion CClI'IWNrldr.
Ir
OK
1 n lAnce! -I If
Umilk
Ir
:!!
JEeft
:':::
.a
IU
. :,..!!: J
ConIroII,pp
Ii
Ito-I
popup
ecl1t
o elqQeDCl
I II
AAJIIo
The obtained masked subsystem is used to generate the 6th sequence of length 1023.
U~----+I~I
ll!!!!JJJ
PlbrJlquoncel
T. f ill
PR8S
In this window, we recognise the popup menu used for the number sequence choice.
The following command lines enable drawing the generated PRBS sequence.
~
o
1
.3
'--
10
l...
L...
20
30
40
50
samp les
60
70
80
90
100
528 Appendices
nIl
PRes
~1------1~~1
~
Aq3'JI'...eJ
ToFIl.
Cown!
.tair.(t,prb.)
r-
5.5
" .5
3.5
10
20
30
40
50
samples
80
70
80
90
100
Bibliography
Anderson B. D.O., Moore 1. B., Optimal Filtering, Prentice-Hall, Englewood Cliffs, NJ,
1979.
Anderson, C. W., Learning to control an inverted pendulum using neural networks, IEEE
Control Magazine, pp. 31-37, April 1989.
Arago 14, Logiquefloue, Observatoire
Fran~ais
Astrom K J., Wittenmark B., Computer-Controlled Systems, Theory and Design, PrenticeHall International Editions, 1984.
Bellanger M., Traitement numerique du signal, Theorie et Pratique, Masson, 1996.
Biran A., Breiner M., MATLAB for Engineers, Addison-Wesley Publishers Ltd., 1995.
Brossier 1. M. , Egalisation adaptative et estimation de phase: Application aux
communications sous-marines. These de Docteur de l'Institut National Polytechnique de
Grenoble.
Brown M., Harris C. 1., Fuzzy Logic, Neural Network and B-Spline for Intelligent Control,
IMA 1. Math. Control and Information Theory, 8, pp. 239-265, 1991.
Brown M., Harris Ch., Neurofuzzy adaptive modelling and control, Series in Systems and
Control Engineering, Prentice-Hall International, 1994.
Brown R. G., Hwang P. Y. C., Introduction to random signals and applied Kalman filtering
with MATLAB exercises and solutions, 3rd Edition, John Wiley & Sons, 1997.
Candy J. V., Signal Processing: The Model-Based Approach, New York: McGraw-Hill,
1986.
Carayannis G., Manolakis D. G., and Kalouptsidis N., A Fast Sequential Algorithm for Leastsquares Filtering and Prediction, IEEE Trans. Acoustics, Speech, Signal Processing,
ASSP-31, pp. 1394-1402, 1983.
Catlin D. E., Estimation, and the Discrete Kalman Filter, Springer-Verlag, 1989.
Chui C. K, Chen G, Kalman filtering with real-time application, Springer-Verlag, 1987
Cioffi, J. M., The Fast Adaptive Rotor's RLS Algorithm, IEEE Trans. Acoustics, Speech,
Signal Processing, ASSP-38, pp. 631-653, 1990.
Connor J., Atlas L., Recurrent Networks and Time Series Prediction, International Joint
Conference on Neural Networks, Seattle, IEEE Publications, pp. 1301-1307.
Davalo E., Nairn P., Des Reseaux de Neurones, Eyrolles, 1990.
Dieulesaint E., Royer D., Automatique appliquee, tome 2, Systemes linea ires de commande
signaux echantillonnes, Masson, 1990.
Domany E., Van Hemmen 1. L., Schulten K, Models of neural networks, Springer-Verlag,
1991.
530 Bibliography
Dorf R. C., Bishop R. H., Modem Control Systems, 7th Edition., Reading MA : AddisonWesley, 1995.
Dubois D., Prade H., Fuzzy Sets and Systems : Theory and Applications, Orlando, FL:
Academic Press, 1980.
Duvaut P., Traitement de signal, Hermes, 1994.
Elbert T. F., Estimation and Control of Systems, New York: Van Nostrand Reinhold, 1984.
Fausett L., Fundamentals of Neural Networks, Architectures, Algorithms and Applications,
Prentice-Hall International Editions, 1994.
Foulloy L., Commande qualitative et commande floue: vers une methode d'ecriture, Actes
des douziemes journees internationales sur I'intelligence artificielle, les systemes
experts et Ie langage naturel, Avignon, France, pp. 473-485, Juin 1992.
Foulloy L., Galichet S., Topology of fuzzy controllers, in Theoretical Aspects of Fuzzy
Control, Nguyen H. T., Sugeno M., et al., Wiley, Inter-Science, 1995.
Gelb A., Applied Optimal Estimation, Cambridge, MA : MIT Press, 1974.
Glover J.R., Adaptative noise cancelling applied to sinusoidal interferences, IEEE Trans.
Acoust. Speech Signal Process., vol. ASSP-25 pp. 484-491, 1977.
Grewal M. S., Andrews A. P., Kalman Filtering Theory and Practice, Englewood Cliffs, NJ :
Prentice-Hall, 1989.
Grimble M. J., Optimal control and stochastic estimation, Theory and Applications, John
Wiley & Sons, 1988.
Haykin S. S., Adaptive Filter Theory, Englewood Cliffs, NJ : Prentice-Hall, 1996.
Landau I. D., Identification et commande des systemes, Hermes, 1988.
Landau I. D., Dugard L., La commande adaptative - Aspects Pratiques et TMoriques,
Paris: Masson, 1986.
Landau I. D., M'Sirdi K. N., M'Saad M., Techniques de Modelisation Recursives pour
I 'Analyse Spectrale, Traitement du Signal, Vol. 3, n 4-5,1986
Jaume D., Thelliez S., Verge M., Commande des systemes dynamiques par calculateur,
Eyrolles, 1991.
Jezequel L., Active control in mechanical engineering, Hermes, 1995.
Jodouin J. F., Les rbeaux de neurones, principes et definitions, Hermes, 1994.
Johnson c., Process control instrumentation technology, Prentice-Hall International, 1988.
King P. J., Mamdani E. H., The Application of fuzzy control systems to industrial processes,
Automatica, Vol. 13, pp. 235-242, 1977.
Kosko B., Neural Networks and fuzzy systems, A dynamical systems approach to machine
intelligence, Prentice-Hall International, 1993.
Kosko B., Neural Networks and Fuzzy Systems, Englewood Cliffs, NJ: Prentice-Hall, 1992.
Kosko B., Neural Networks for Signal Processing, Englewood Cliffs, NJ: Prentice-Hall,
1992.
c., Fuzzy Logic in Control Systems, Fuzzy Logic Controller: Parts 1 and 2, IEEE
Trans. Sys. Man and Cyb., vol. 20, No.2, pp. 404-435.
Lee C.
Bibliography 531
Ljung L., Soderstrom T., Theory and Practice of Recursive Identification, Boston: MIT Press,
1983.
Luenberger D. G., Introduction to dynamic systems, Wiley, 1979.
Mamdani E. H., Application of fuzzy algorithms of simple dynamic plant, Proc. IEEE, 121,
pp. 1585-1588, 1974.
Marven c., Ewers G., A Simple Approach to Digital Signal Processing, Texas Instruments,
1993.
Meditch 1. S., Stochastic Optimal Linear Estimation and Control, New York: McGraw-Hill,
1969.
Meunier B., La logiquefloue, Que sais-je n 2702, P.U.F., Paris, 1993.
Miller W. T., Sutton R. S., & P. 1. Werbos, eds, Neural Networks for Control, Cambridge,
MA: MIT Press, 1990.
Minkler G., Minkler 1., Theory and Application of Kalman Filtering, Palm Bay, FL:
Magellan
Book Co., 1993.
Minsky M., Papert S., Perceptrons, Cambridge, MA: MIT press, 1969.
Mokhtari M., Mesbah A., Apprendre et Maftriser MATlAB, versions 4 & 5 et SIMUUNK,
Springer-Verlag, 1997.
Murat K., Techniques de traitement numerique des signaux, Collection Traitement de
l'Information, Presses Techniques et Universitaires Romandes.
Narendra K. S., Adaptive control using neural network, in Neural Networks for Control
CW. T. Miller et al. Eds), Cambridge, MA : MIT Press, pp. 115-142.
Nie 1., Linkens D., Fuzzy-Neural Control, Principles, Algorithms and Applications,
Prentice-Hall, 1995.
Oppenheim A. V., Digital Signal Processing, Prentice-Hall International, 1975.
Oppenheim A. V., Schafer R. W., Discrete-Time Signal Processing, Englewood Cliffs, NJ:
Prentice-Hall,1989.
Pedrycz W., Fuzzy Control and Fuzzy Systems, 2nd Edition Research Studies Press, Taunton,
John Wiley and Sons, 1993.
Porat B., Digital Processing of Random Signals, Theory & Methods, Prentice-Hall
Information and System Sciences, 1994.
Soderstrom T, Stoica P., System Identification, Englewood Cliffs, NJ: Prentice-Hall, 1985.
Sorenso H. W., Kalman Filtering: Theory and Application, New York: IEEE Press, 1985.
Stengel R. F., Stochastic Optimal Control- Theory and Application, New York: Wiley, 1986.
Sugeno M., Industrial Applications offuzzy control, Amsterdam: North-Holland, 1985.
Sugeno M., Special issue onfuzzy control, Fuzzy Sets and Systems, 26, pp. 149-223, 1988.
Sutton R. S., Two Problems with Backpropagation and other Streepest-descent Learning
Procedures for Networks, Proceedings of the Eighth Annual Conference of the
Cognitive Science Society, pp. 823-831, 1986.
532 Bibliography
Vemuri V., Artificial Neural Networks, Theoretical Concepts, Washington, DC: IEEE
Computer Society Press, 1988.
Widrow B., Lehr M. A., 30 Years of Adaptive Neural Networks,' Perceptron, Madaline, and
Backpropagation, Proceedings of IEEE, yol. 78, n 10, pp. 1415-1441, 1990.
Widrow B., Steams S., Adaptive Signal Processing, Prentice-Hall, 1985.
Zadeh L. A., Fuzzy sets, Information and Control, Vol. 8, pp. 338-353, 1965.
Index
A
a posteriori, 222
a priori error, 222, 225, 226, 247
acceleration, 93
acker, 357, 362
Ackennan, 62,63, 68, 70, 75,356
acoustic wave, 463
Adaline, 150,445
adaptation coefficient, 237, 454, 468
adaptation gain matrix, 79
adaptation gain, 211, 224, 225, 226, 227,
245,247,498
adaptation speed, 223
adaptive control, 200
adaptive equalizer, 498
adaptive filter, 220, 453, 460, 463, 493
adaptive neural networks, 150
adaptive predictor, 227, 229
adaptive, 219
add button, 521
add,520
aggregation, 332
air gap, 276
algorithm, 219
amplifier, 257
analog, 1, 15
analog and digital control, 2
ARMA,152
attenuation, 238,454,463
autocorrelation matrix, 221, 225, 226, 247
autoregressive model, 227, 228
autoregressive process, 227
axis, 488, 493
B
backpropagation algorithm, 170
backpropagation error, 165
balancing coefficient, 441
bias vector, 166
bias, 179
bilinear transformation, 15
binary sequence, 49
bipolar sigmoid, 166
bode, 7,18,66,235,260
C language, 503
C MEX file, 504, 512
C++ compiler, 512
c2d, 20, 90, 285, 356, 361
canon, 59, 60
canonical, 56
Cart with inverted pendulum, 304
cart, 303
case, 497
Cayley-Hamilton theorem, 62
ceil,457
centered white noise, 227
central frequency, 231, 234, 239, 244,
245,251
centre of gravity method, 331
centre of gravity, 304
channel, 481
characteristic equation, 287, 411
characteristic polynomial, 63, 524
checkbox, 520
closed-loop, 2, 3,14, 15, 17,22,26,32,
61
closed loop stability, 184
coefficients vector, 220, 221, 225, 226,
247,250
coil, 276
companion, 56
compensation, 352
complex poles, 60
conduit, 30, 463
continuous state model, 505
continuous system, 51
Control System TOOLBOX, 18,410
controllability, 53
convergence, 238,249,461
coordinates, 341
534 Index
counternoise, 463
ctrb,59,69,74,357
D
damping coefficient, 32, 39
damping factor, 281
damping ratio, 3
dcgain,22
dead zone, 274
defuzzification, 318, 321
delete, 520
deltalin, 170
deltalog, 170
deltatan, 170
dendrite, 149
denormalization, 373
derivative action, 3, 4, 7, 9, 23, 24, 272
derivative corrector, 1, 3
derivative return corrector, 1, 4
derivative, 423
deterministic process, 54
dialog box, 517
differential equation, 48
differential equations, 429
digital filter, 219
digital signal processor, 219, 466
direct chain, 67
direct current motor, 47, 50, 57, 58, 65,
74,79,90
dirFeedthrough, 504
discrete state model, 507
discrete system, 51, 74
discrete Transfer, 524
discretisation, 1, 15, 20
disnormaiisation, 195
disp, 5, 6, 7, 8, 22, 26
dispersion channel, 484
distortion, 454
distortion factor, 484
distortion ratio, 258
disturbance rejection, 416
disturbance, 340
dlqr, 368, 369
documentation, 519
DSP, 219,329,466
E
echo cancellation, 453
edit Mask, 521
eigenvalues, 55, 60
electrical time constant, 66
electromagnet, 275
electromagnetic levitation, 276
equalization, 482
equalizer, 481
estimation, 229
estimation error, 55
evolution matrix, 55
exact least squares criterion, 222
exponential, 230 '
exponential envelope, 237
F
factor, 226
feedback,22,23, 26,82,411
filter, 11,39,41
filter coefficients, 220
filter order, 459
filter time constant, 223, 468
find,14,26,33,290
flag indicator, 504
flag, 524
fliplr, 235, 240, 248
floor, 456, 457
follower system, 13, 14
forecastings, 251
from file, 496
function, 455, 460, 465, 484, 497,504
fundamental frequency, 463
fuzzification, 291, 303
Fuzzy control, 422
fuzzy controller, 303
fuzzy corrector, 291
fuzzy matrix, 334
Fuzzy regulation, 315
fuzzy regulator, 427
fuzzy rules, 329, 424
G
gain margin, 55, 79, 262
gain, 523
gantry, 429
gap, 184
gaussian membership, 298
generalised Bell curve, 316
gensurf, 320
gensurf, 294
global,497
Index 535
gradient algorithm, 222, 454, 464
Graphic animation, 418
graphic animations, 503
graphic editor, 316
Graphical animation, 303, 340
gravity center, 294
gtext,491
H
Hadamard product, 172
hands-free telephone, 443
harmonics, 465
heating resistor, 347
hidden layer cells, 176
hydraulic friction coefficient, 47
hyperbolic tangent, 166
identification, 31, 51
impulse, 20, 21
impz,20
indicial response, 70
inductance, 47
induction, 276
inertia moment, 47
Inference rules, 294, 303
informative signal, 234
initial conditions, 49
instrumentation amplifier, 347
integral action, 23, 24
integral control, 1, 3
integral correction, 414
integration, 57, 198
intercorrelation, 79
intercorrelation matrix, 81, 221
interference, 231, 481
inverted pendulum, 303
K
kalman, 81, 364, 369
Kalman filter, 78
Kalman predicter, 54
Kalman reconstructor, 364
Kinetic energy, 303, 304
L
Lagrange equation, 303
Laplace transform, 49, 277, 348
loglog, 9, 11
logsig, 167, 170, 374
logspace, 261, 265
Look Under ~ask, 521
look-up counts, 463
loudspeaker, 464
LQ, 1,27,29,43,368
LQI control, 441
LQI controller, 429
LQI monovariable control, 440
LQI, 1,27,28,30,31,32,34,36,38,39,
40,41,43
Iqr,368
Isim,84
Luenberger observer, 54
M
293, 318
margin, 22, 261
mask, 520
~ask Subsystem, 518
masking, 515
masking a set of blocks in SI~ULINK,
516
matrix algebra, 47
mdIDerivatives,512
mdlTerrninate,512
mdlUpdate,512
mean square deviation, 482
measurement noise, 78
~easurement Update, 79
membership degree, 422
membership functions, 329, 422
microphone, 463
modeling noise, 78, 79, 84, 86, 90
moment of inertia, 306
~amdani,
536 Index
monovariable,47
motor, 30
multivariable, 1,27,28,30,36,41,43,47
Multivariable RST control, 435
N
naturrufrequency, 11,32,39,41,281
network, 373
network training, 189
Neurru Network TOOLBOX, 152,445
neurru networks, 149, 187, 189
neuron, 149
neuronru control, 378
neuronru inverse model, 371
noise cancellation in a conduit, 464
noise power, 472
non-linear organ, 469
non-stationary signru, 461, 462
normruization, 373
nurn2str,5,6,7,8,22,26
NumContStates, 504
NumDiscStates, 504
Nurnlnputs, 504
NumOutputs, 504
NurnSampleTimes, 504
nyquist, 21
Nyquist criterion, 1, 17
Nyquist diagram, 17
observability,53
observation equation, 49, 53
observers, 54
open loop gain, 258
operationru amplifier, 257
optimru control, 27, 28, 30, 39
optimru observer, 55
optimization criterion, 220
originru, 52
oscillator, 257
oven, 347
oven control, 348
overshoot, 203, 439
overshooting, 353
Q
quadratic linear control, 27, 368
p
pade,25
parrulel, 82
R
randn, 90, 229
Index 537
sampling period, 74
saturation, 3, 166,351
save, 482, 498
scalar, 219, 222, 223
selectivity, 231
semilogx, 5, 6, 7, 9, 11
semilogy, 230
sensor,89,282,347
series, 23, 25
set, 34, 37,42,45,290,494
set-point, 427
S-function, 449, 454, 464, 472, 503, 525
showrule, 319
signal prediction, 213
simsizes, 342, 456, 461
SIMULINK, 449, 496, 503
sinusoidal interference, 234, 236
sizes, 525
sizes,DirFeedthrough, 3, 525
sizes.NumContStates, 34, 525
sizes.NumDiscStates, 34, 525
T
tachometer, 5
tansig, 170
telephone, 445
telephone line,453
text,288
tf, 22, 57, 235, 240, 248, 265
thermocouple, 347
538 Index
threshold, 166,472
time constant, 2, 3, 5, 11,23,230
time response, 439
Time Update, 78
To Workspace, 517
trace,227,248,251,254,255,256
training law, 151
transfer function, 2, 14, 16, 22, 24, 29, 32,
41,258
transformation matrix, 56
transmission delay, 454
transmission line, 453
transparent, 519
transport delay, 454, 464, 496
trapezoidal membership, 291
traprnf, 329
traprnfs, 291
triangular membership function, 318
Tustin, 15,20
u
uniform random number, 340
unipolar sigmoid, 166
uparrow, 261
v
variable sampling period, 510
variance matrix, 78, 472
variance, 93,218,234,483
view, 294
w
weight, 179
weighting coefficient, 33, 34, 36
white noise, 469
Widrow-Hoff,154