You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228671187

Simulink implementation of adaptive control and multiple model network


control

Article · January 2001

CITATIONS READS

6 2,588

3 authors, including:

Gregor Gregorcic A. Mullane


qpunkt GmbH 25 PUBLICATIONS   2,022 CITATIONS   
18 PUBLICATIONS   362 CITATIONS   
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Gregor Gregorcic on 10 June 2014.

The user has requested enhancement of the downloaded file.


Simulink Implementation of Adaptive Control and
Multiple Model Network Control
Gregor Gregorčič, Alan Mullane and Gordon Lightbody
Department of Electrical Engineering
University College Cork
IRELAND
E-mail: gregorg@rennes.ucc.ie

Abstract — The Simulink implementation of two adaptive model


based control techniques will be presented in this paper. Implemen-
tation of the Recursive Least Squares algorithm for Pole–Placement
adaptive control will be shown. The parameters of the nonlinear
system can change over time due to different operating conditions.
When a single identification model is used, it will have to adapt it-
self to the operating condition, before appropriate control can be
taken. Slowness of adaptation may result in a large transient er-
ror. As an alternative, the Multiple Model framework can be used.
If models are available for different operating conditions, controllers
corresponding to them can be found in advance. The control strategy
is to determine the best model for the current operating condition
and activate the corresponding controller. A simulated example of
the control of a continuous stirred tank reactor will be shown, us-
ing the developed Adaptive Control and Multiple Model Network
simulation tools.
Keywords — Simulink, S–Function, Adaptive Control, Multiple
Model Control

I Introduction which has a corresponding linear controller. A su-


pervisor then determines from process data which
The most common adaptive control schemes are model best represents the process at a particular
based on self-tuning control techniques such as time, and then switches in its associated controller.
Pole–Placement [1], [2], [3] and minimum variance
control. In such controllers, a linear parameter- In terms of the structure of the overall model,
ized model of the process is continuously identified, this last approach is the simplest. The parameters
utilizing for example the Recursive Least Squares of the local models are identified locally. The Iden-
(RLS) algorithm [4]. This model is then used to tification can be done online in the closed–loop.
update the controller transfer functions. The idea is to bring the system to the desired op-
Narendra and Parthasaranthy [5], proposed the erating point using Pole–Placement adaptive con-
enhancement of model based control techniques, trol. When the closed–loop system reaches the
such as self–tuning controllers, by replacing linear steady state, (i.e. when the linear model represents
parameterized models with neural network based the controlled system accurately), the adaptation
nonlinear models. Two other approaches, based is turned off and the parameters of the model and
on local linear models have been also proposed. its corresponding controller are stored in to the
Such methods include Takagi–Sugeno Fuzzy mod- database.
els [6], and Local Model Networks [7], [8], both In order to facilitate teaching of adaptive con-
of which partition the nonlinear process into op- trol, an interactive simulation tool was required.
erating regimes, within which models are identi- This allow issues such as: sensitivity to the for-
fied. These models are then combined to form a getting factor, bumbless transfer, effect of the co-
nonlinear model of the process across its expected variance matrix reset, and the effect of the addi-
operating space. Controllers can be designed for tional excitation signal to be easily investigated.
each operating regime, and combined in a similar This paper shows the Simulink implementation of
fashion to provide a nonlinear control law, which such a tool, which was also extended for Multiple
would be applied, even for large set–point changes Model Network Control. After the Multiple Model
or disturbances, which drive the process across op- Network has been trained, it can be used as an in-
erating regimes. dependent control strategy. Using the Simulink
Narendra and Balakrishnan [9], [10], have pro- Implementation of Adaptive Control and Multi-
posed the use of multiple linear models, each of ple Model Network, the characteristics of these ap-
proaches have been examined and some limitations of R(q) is larger than, or equal to, the degrees of
of each of them have been pointed out. Continu- S(q) and T (q).
ous Stirred Tank Reactor (CSTR) has been used
as an example of the nonlinear system. R(q)u(k) = T (q)r(k) − S(q)y(k) (7)

II Adaptive Control Vs. Local Model This control structure can be represented as shown
Network Control in Fig. 1.

a) Pole–Placement Self–Tuning Control


PRBS

Pole-Placement Self–Tuning control utilizes the T(q)


R(q)
B(s)
A(s)
Scope

following ARX model of the process [4]. Here u(k) Set−Point Controller
Feedforward Nonlinear System

and y(k) are the input and output of the process, Controller
Feedback
S(q)
R(q) y(k)
with A(q) and B(q) polynomials in the forward u(k)
RLS theta(k+1)

shift operator, ε(k) is white noise. Recursive Least Squares Estimator Controller Update

A(q)y(k) = B(q)u(k) + ε(k) (1)


Fig. 1: Simulink block diagram of the Pole–Placement
Adaptive Control.
Here

A(q) = q n + an−1 q n−1 + an−2 q n−2 + · · · + a0 (2) The closed–loop system can be represented as fol-
lows:
(A(q)R(q) + B(q)S(q)) y(k) = B(q)T (q)r(q) (8)
B(q) = Bm q m + bm−1 q m−1 +
Pole–Placement design is the choice of polynomi-
bm−2 q m−2 + · · · + b0 (3)
als R(q) and S(q), for estimated Â(q) and B̂(q)
polynomials, to obtain some desired closed–loop
This model can be continuously identified using,
characteristic equation Acl (q). This is represented
for example, the Recursive Least Squares algo-
by the following Diophantine equation:
rithm, with exponential weighting [4]:
Acl (q) = Â(q)R(q) + B̂(q)S(q) (9)
θ̂(k + 1) = θ̂(k) + P(k + 1)ψ(k + 1)
h i If polynomials Â(q) and B̂(q) do not have com-
y(k + 1) − ψ T (k + 1)θ̂(k) (4) mon factors, then this Diophantine equation can
be solved using matrix calculations, based on the
Sylvester matrix [2].
P(k + 1) = b) Multiple Model Network Control
" T
#
P(k)ψ(k + 1)ψ (k + 1)P(k) 1 When a single identification model is used, it will
= P(k) − (5)
λ + ψ T (k + 1)P(k)ψ(k + 1) λ have to adapt itself to the operating condition be-
fore appropriate control can be taken. After the
where the estimate of the parameter vector is de- operating condition of a system changes abruptly,
fined as: the original model (and hence controller) is no
longer valid. Slowness of adaptation may result
θ̂(k) = [ân−1 (k), · · · , â0 (k), b̂m (k), · · · , b̂0 (k)]T (6) in a large transient error.
If models are available for different operating
Exponential weighting implies that old measure- conditions, controllers corresponding to them can
ments are exponentially discounted, which gives be found in advance. The control strategy shown
an ability to track changes in time–varying and in [9], [10] is to determine the best model for the
nonlinear systems. The parameter λ in equation current operating condition at every instant, and
(5) is called the forgetting factor and is usually activate the corresponding controller. This control
chosen in range 0.95 < λ < 1. The choice of λ de- structure, as shown in Fig. 2, is based on N mod-
pends on how the properties of the system change. els which have been determined at various points
Smaller values of λ result in a faster forgetting, across the operating range of the process. A con-
which can be used to cope with nonlinear and fast troller is then designed for each model, using the
changing systems. Likewise values of λ close to Diophantine pole–placement algorithm. A super-
1 result in slower forgetting, and can be used for visor then compares the output errors for the N
systems that change gradually. models. A discrete equivalent of the performance
The control law of equation (7) is assumed, index proposed in [10] is given in equation (10), for
where R(q), S(q) and T (q) are polynomials in the the ith model:
forward shift operator. u(k) and y(k) are the plant M
X
input and output, r(k) is the set–point. To have Ji (k) = αe2i (k) + β exp(−jλi )e2i (k − j) (10)
a causal controller, it is necessary that the degree j=1
However, for a nonlinear process operating over
its full range, the steady–state control input at
each operating point is usually a nonlinear func-
tion of the desired output. It is therefore difficult
to provide zero mean data without online estima-
tion of the means. It is then preferable to use the
following second order discrete time incremental
model [12]:

q 2 + a1 q + a0 ∆Ca (k) = (b1 q + b0 ) ∆qc (k) (14)



Fig. 2: A general architecture of the Multiple Model
Network Control.
Here

Here ei (k) is the output error for the ith model. ∆Ca (k) = 1 − q −1 Ca (k)

(15)
The parameters α, β and λ are chosen to yield
a desired combination of instantaneous and long-
term accuracy. The forgetting factor λ determines
∆qc (k) = 1 − q −1 qc (k)

(16)
the memory of the index, and will have direct effect
on the decision speed of the supervisor. The model This will provide the additional benefit of intro-
with the lowest cost is selected, and its associated ducing integral action into a resulting controller.
controller is switched in the loop. For good parameter estimation, the system should
III The Nonlinear System be excited with a signal of rich frequency content.

A chemical system, known as a Continuous Stirred IV Blockset Implementation


Tank Reactor (CSTR), was utilized as an example, For the simulation, three Simulink block diagram
to illustrate the use of the proposed adaptive con- windows have to be opened. The Control Panel
trol tool. In the CSTR, two chemicals are mixed, can be seen in Fig. 5. From the Control Panel the
and react to produce a product compound with Adaptive Control block diagram, seen in Fig. 3,
concentration Ca (t). The temperature of the mix- can be controlled. Also through the Control Panel,
ture is T (t). The reaction is exothermic, produc- the model and controller parameters can be sent
ing heat which acts to slow the reaction down. By from the Adaptive Control block diagram to the
introducing a coolant flow–rate qc (t), the temper- Multiple Model block diagram seen in Fig. 6. In
ature can be varied and hence the product concen- this case, the structure of the Multiple Model Con-
tration controlled. This system can be described trol network has to be defined in advance and it
by the following nonlinear simultaneous differen- can not be changed online. As such the Multiple
tial equations1 : Model block diagram effectively acts as a database
for the model and controller parameters.
Q
Ċa (t) = (Ca0 − Ca (t)) − a) Adaptive Control block diagram
V  
E The Adaptive Block diagram is shown in Fig. 3. As
k0 Ca (t) exp − (11)
RT (t)
1

PRBS PRBSsw

1 0.00019z−1+0.00016z−2
PRBS @ Input
  1−1.59z−1+0.71z−2

Q E 0 PRBS @ Set Point


Model

Ṫ (t) = (T0 − T (t)) − k1 Ca (t) exp − + 688.34


Delta Operator1 qc_ini Constant1

V RT (t) 0.08

Constant
1+0.54z−1
+1.28e−007z−2

Filter
2
1
1−z−1
PLANT

   Sw Control Open/Close Scope


Integrator

k3 P Matrix: 0
Switch
Delta Operator
CSTR

k2 qc (t) 1 − exp − (Tc0 − T (t)) (12)


0.82 0.48 −2197.97 1912.26
0.48 0.61 −921.12 634.86 Constant2 Set Point: 0.08
−2197.97 −921.12 21125804.67 −17987627.25 RLS: on

qc (t) 1912.26 634.86 −17987627.25 19367091.03

Condition number = 202719881.88


Determinant = 1.58271041002e+013
y(k)
RLS
PRBS: on\Input
PRBS lambda: 1
PRBS Gain: 0.01
PRBS Offset: 0
Estimator theta(k+1)
on P Matrix: Update
u(k) Forggeting Factor: 0.956
P Matrix Display Loop: Closed
RLS Controller Auto Set Point Change: on

Consider the flow–rate qc (t) as the input and prod- 4408.81−6062.35z−1


+2341.88z−2
Tuning Multiple Model Tuning: on
Sample Time: 0.1

1+0.54z−1
+1.28430871649e−007z−2 Control Panel Display

uct concentration Ca (t) as the output of the sys- Feedback

tem. The following second order difference equa-


tion model of the CSTR process was proposed: Fig. 3: Adaptive Control block diagram

q 2 + a1 q + a0 Ca (k) = (b1 q + b0 ) qc (k)



(13) well as the standard Simulink blocks it contains the
following, which were built using S–Functions [13]:
In the Pole–Placement algorithm the parameters PRBS block: This is used to add excitation to the
a0 , a1 , b0 and b1 are identified on–line using the the control signal, to insure that all system modes
RLS algorithm. It is essential that the input and are excited. Since the estimated model is linear,
output data to the RLS algorithm are zero mean. the amplitude of the PRBS signal must be small
1 The CSTR parameters of (11) and (12) can be found in comparison with set–point changes. The PRBS
in [11]. signal is produced using the combination of an 11
bit shift register and an XOR gate [14]. The mini- tion number and determinant are displayed on the
mum length of the pulse, amplitude, offset and the icon of the P Matrix Display.
sampling time can be set in the Subsystem Mask A more informative indicator of covariance ma-
window. trix behavior is given by the P Matrix Viewer.
RLS Estimator block: This is a Simulink imple- This uses the Matrix Viewer to display a color coded
mentation of the Recursive Least Squares algo- covariance matrix. Each segment of window rep-
rithm. Inputs to the RLS Estimator are the input resents an element of the covariance matrix, with
and output of the system or their increments. The dark color (black) corresponding to a small num-
output is the vector of the estimated parameters ber, and light (white) color corresponding to the a
θ̂(k + 1). large number. The condition of the matrix can be
To be able to calculate the new parameter vector deduced from the pattern. The pattern on the left
θ̂(k + 1), it is necessary to store the previous val- hand side of the Fig. 4 shows ill–conditioned co-
ues of the covariance matrix P(k), regressor vector variance matrix, whereas than on the right is well
conditioned.
ψ(k) and parameter vector θ̂(k). (See (4),(5)) This
is done by splitting the covariance matrix in to a
vector. The regressor vector and the parameter
vector are then appended to this vector. This new
vector is used as the state vector in the S–Function.
In the Update section of the S–Function the covari-
ance matrix, regressor vector, and the parameter
vector are rebuilt from the state vector. Then the
new P(k + 1), ψ(k + 1), θ̂(k + 1) are calculated and
the state vector is updated.
The estimated model parameters are “sent” to Fig. 4: P Matrix Viewer. Left: ill–conditioned covariance
the Model transfer function using the set param matrix, right: well conditioned covariance matrix
function [15]. The following parameters can be set
in the RLS Estimator Subsystem Mask window:
• Structure of the model (m and n in (6)) Control Panel Display: This reads values from the
• Forgetting factor λ Control Panel block diagram, using the get param
• Covariance matrix initialization factor α function [15], and sets the values in the Adaptive
• Name of the target model Control block diagram using the set param func-
• P Matrix Reset (on/Reset, off/Update) tion. These are displayed on the block icon. This
• Name of the P Matrix Display S–Function is responsible for communication be-
• Switch (on/off) tween the Control Panel and the Adaptive Control
• Sample time block diagram window.
Controller Tuning block: This takes as an input b) Control Panel block diagram
the parameter vector θ̂(k + 1). From the model
The simulation can be controlled from the Con-
parameters and the desired closed–loop behavior
trol Panel block diagram shown in Fig. 5. Various
specified in the desired characteristic polynomial
parameters can be controlled online, while the
Acl , the controller parameters are calculated using
simulation is running. Most of the controls
the Diophantine equation (9). The controller is
are built using the Manual Switch and Slider
factorized in to the two transfer functions, Filter
Gain. The Terminator block is used to avoid
and Feedback. These transfer functions are up-
the warning in the Matlab command window.
dated using the set param function. The step re-
The following simulation properties can be con-
sponse of the desired closed–loop behavior is plot-
trolled via the Control Panel Block diagram:
ted on the block icon. The structure of the model,
• Set–point setting
desired characteristic polynomial, name of the con-
• Switch between manual and preprogrammed
troller transfer functions and the sample time can
set–point changes
be specified in the Controller Tuning Subsystem
• RLS algorithm on/off switch
Mask window.
• PRBS signal on/off switch
P Matrix Display: A tool for observing the covari- • PRBS parameters settings (λ, amplitude and
ance matrix is also provided. Online monitoring off–set)
of the covariance matrix can be useful as a diag- • PRBS signal can be injected at the set–point
nostic of the parameter vector convergence. Es- or at the input of the system
timated model parameters are close to real pa- • Covariance matrix reset switch
rameters when the covariance matrix is well condi- • Forgetting factor setting
tioned. This can be seen from the condition num- • Switch between the open and closed–loop
ber, or from the determinant of the covariance ma- • Sample time setting
trix. The covariance matrix, along with its condi- • Multiple Model Tuning switch
1 0.08 1 0.956 Model Control Network are identified and stored
Constant Set Point Terminator Constant1 Forgetting Factor Terminator5
in to the Multiple Model block diagram. Then the
1 1 1
RLS ON Constant2 PRBS Lambda Terminator6
Multiple Model block diagram is now redy to be
0 Terminator1 used.
RLS ON−OFF 1 0.001
RLS OFF Constant3 PRBS Gain Terminator7
1 a) Adaptive Control and Parameter Estimation
1 0
PRBS ON
Constant4 PRBS Offset Terminator8
for Multiple Model Control Network
0 Terminator2
PRBS ON−OFF
PRBS OFF 1
Closed Loop
Since the CSTR system is highly nonlinear, the
1
@ Set Point 0 Terminator10 adaptive control is not trivial. To allow the RLS
Open−Close Loop
0 Terminator3
Open Loop Estimator to identify the initial model at the cho-
PRBS Set P.−Input
@ Input 1
sen operating point and for the Controller Update
On
1
0 Terminator11
block to set the controller parameters to their right
Reset Auto Sp Change
0 Terminator4
Off values, it is necessary to run the simulation in the
P Matrix Reset
Update open–loop for a few seconds. After the model is
1
Source Model: CSTR_tuning/Model
Target Model: CSTR_switch/Model_0.08
identified and controller is set, the loop can be
Source Feedback: CSTR_tuning/Feedback
Update Model
Target Feedback: CSTR_switch/Feedback_0.08 closed. The decision, when the parameters are
0 Source Filter: CSTR_tuning/Filter
No Update
Multiple Model
Tuning ON/OFF
Target Filter: CSTR_switch/Filter_0.08 close enough to real parameters can be made ac-
Target cording to observation of the patten of the covari-
1 0.1 ance matrix.
Constant5 Sample Time Terminator9
After the loop has been closed, the set–point
can be changed. As can be seen from the Fig. 7,
Fig. 5: Control Panel block diagram the transient response using adaptive control is is
poor. When the system settles at the new operat-
c) Multiple Model block diagram
Set−point
Multiple Model block diagram, shown on Fig. 6, 0.13
Output

can be used separately from the Control Panel and


the Adaptive Control block diagram. Usually, the 0.12

model and controller parameters are identified on-


line, for different operating points, using the Adap-
Ca (mol/l)

0.11

tive Control block diagram. Then via the Con-


trol Panel, the parameters are stored to the corre- 0.1

sponding transfer function blocks.


0.09

0.00018z−1
+0.00018z−2
1−1.87z−1
+0.94z−2

Model_0.13 0.08

0.00018z−1
+0.00017z−2
supervisor
1−1.69z−1
+0.78z−2 54 56 58 60 62 64 66 68 70
S−Function
Model_0.105
Time (min)
0.00018z−1
+0.00015z−2
1−1.48z−1
+0.59z−2
691.98
1+0.65z−1
+9.82e−008z−2 Model_0.08
qc_ini Constant1
Filter_0.13

1
Fig. 7: Set–point tracking for the CSTR using adaptive
710.10 PLANT
1
Constant2
0.08

Set Point
1+0.59z−1
+1.17e−007z−2

Filter_0.105 Multiport
1−z−1

Integrator2 CSTR Scope


control: The sampling time was 0.1 minute and forgetting
752.24
Switch1
5728.23−8423.13z−1+3386.88z−2 factor λ was 0.956. The desired closed-loop performance
1+0.65z−1
+9.82e−008z−2
1+0.52z−1
+1.54e−007z−2

Filter_0.08
Feedback_0.13 was defined as second order dominant with a damping
rad
5048.55−7088.15z−1+2749.70z−2
1+0.59z−1
+1.17e−007z−2
ratio of 0.707 and natural frequency of 7.4 min
Multiport Feedback_0.105
Switch2

4344.24−5629.5z−1
+2037.52z−2
1+0.52z−1
+1.54e−007z−2

Feedback_0.08
ing point, the parameters can be stored in to the
Multiple Model block diagram. This is done be
Fig. 6: Multiple Model block diagram switching the Multiple Model Tuning switch in the
Control Panel to on. It is preferable, that adapta-
The Supervisor block calculates the cost, proposed tion (RLS switch) is turned off, while the model
in (10), at each sampling time. The model with the and controller transfer function in the Multiple
lowest cost is found and the corresponding con- Model block diagram function are updated. The
troller is switched in to the closed–loop using the parameters will be stored in to the blocks specified
Multiport Switch. in the Target block. After the parameters have
been stored, the Multiple Model Tuning switch is
V Simulation Example
then switched off, with RLS adaptation switched
The aim of this simulation is to show the difference back on. Then the set–point can be changed again.
of the two different control strategy. The simula- While the simulation is running, all the proper-
tion is separated into two tasks. First, the adaptive ties available in the control Panel can be changed.
control strategy is used. During this period the This is a powerful tool to study adaptive control
model and controller parameters for the Multiple strategy, especially for nonlinear systems.
b) Multiple Model Control [3] K. J. Åström and B. Wittenmark. Self–tuning
controllers based on pole–zero placement. IEE
Once the model parameters for the Multiple Model Proceedings, Part D: Control Theory and Ap-
Control Network have been identified, the cost pa- plications, 127(3):120–130, May 1980.
rameters α, β and λ in the (10) are left to be found.
Parameters α, β and λ are usually chosen empiri- [4] Lennart Ljung. System Identification: Theory
cally, and can be different for each model. Choos- for the User. Prentice Hall PTR, 2nd edition,
ing a different set of parameters for each model Jan. 1999.
can be a difficult task, particularly when the num-
ber of models is large. In this case, parameters α, [5] K. S. Narendra and K. Parthasarathy. Identi-
β were fixed and λ was chosen according to the fication and control of dynamical systems us-
bandwidth of each particular model. As seen from ing neural networks. IEEE Transactions on
Fig. 8 the multiple model approach gives a better Neural Networks, 1(1):4–27, Mar. 1990.
response than self–tuning control. Since an incre- [6] T. Takagi and M. Sugeno. Fuzzy identifica-
mental approach for estimating model parameters tion of systems and its applications to model-
and corresponding controllers was utilized and in- ing and control. IEEE Transactions on Sys-
tegral action was introduced, choosing a different tems, Man and Cybernetics, SMC-15(1):116–
controllers yields a different transient response but 132, Jan.–Feb. 1985.
has no effect on steady–state error.
[7] G. Lightbody M. D. Brown and G. W. Ir-
win. Nonlinear internal model control using
Set−point
0.13
Output local model networks. IEE Proceedings: Con-
0.125

0.12
trol Theory and Applications, 144(6):505–514,
0.115 Nov. 1997.
Ca (mol/l)

0.11

0.105 [8] R. Murray-Smith and T.A. Johansen. Multi-


0.1

0.095
ple Model Approaches to Modelling and Con-
0.09 trol. Tylor & Francis, London, 1997.
0.085

0.08 [9] K. S. Narendra and J. Balakrishnan. Adaptive


Controller Switch Position
control using multiple models. IEEE Trans-
Controller 3
Controller 2
actions on Automatic Control, 42(2):171–187,
Controller 1 Feb. 1997.
64 66 68 70 72 74 76 78 80
Time (min)
[10] K. S. Narendra J. Balakrishnan and M. K.
View publication stats
Fig. 8: Set–point tracking for the CSTR using multiple
Ciliz. Adaptation and learning using multiple
model control: The following set of parameters for Ji was models, switching, and tuning. IEEE Control
chosen: M = 10, α = 0.5, β = 0.7, λ1 = 35, λ2 = 25 and Systems Magazine, 15(3):37–51, Jun. 1995.
λ3 = 2.45.
[11] G. Gregorčič and G. Lightbody. A compar-
ison of multiple model and pole–placement
VI Conclusion self–tuning for the control of highly nonlinear
processes. In Proceedings of the Irish Signals
A Simulink implementation of the Pole–Placement and Systems Conference, pages 303–311, Jun.
Adaptive Control and Multiple Model Control 2000.
Network was presented in this paper. As a part
of the proposed simulation tool, the S–function [12] P. E. Wellstead and M. B. Zarrop. Self–
realization of the PRBS signal, Recursive Least Tuning Systems: Control and Signal Process-
Squares algorithm, Diophantine equation solver ing. John Wiley & Sons, New York, 1991.
and covariance matrix viewer were explained. The
CSTR system was used to show the use of the [13] Writing S–Functions. The MathWorks, Inc.,
Adaptive Control and Multiple Model Network Oct. 1998.
simulation tool. [14] D. Matko. Identifikacije. Založba FE in FRI,
References Ljubljana, 2. izdaja, 1998.

[1] K. J. Åström and B. Wittenmark. Adaptive [15] Using Simulink. The MathWorks, Inc., Jan.
control. Addison–Wesley Publishing Com- 1999.
pany, Reading, 2nd edition, 1995.

[2] K. J. Åström and B. Wittenmark. Computer–


Controlled Systems: Theory and Design.
Prentice Hall, London, 3rd edition, 1997.

You might also like