SICE Journal of Control, Measurement, and System Integration, Vol. 6, No. 1, pp.

038044, January 2013

Fictitious Reference Iterative Tuning of Internal Model Controllers for

Non-Minimum Phase Systems: A Laguerre Expansion Approach

Hien Thi NGUYEN , Osamu KANEKO , and Shigeru YAMAMOTO

Abstract : Fictitious reference iterative tuning (FRIT) is one of the eective data-driven tuning methods for parameters
of a controller with only one-shot experiment. This paper applies FRIT to internal model control (IMC) for linear, time
invariant, stable and non-minimum phase systems, which enables us to simultaneously obtain a desired controller and
a mathematical plant model. Here, the authors consider the case where we do not have any information on the system.
To overcome the diculty of treating the non-minimum phase behaviors of the system, Laguerre expansion is used to
describe the internal model.

Key Words : data-driven approach, fictitious reference iterative tuning (FRIT), internal model control (IMC), non-
minimum phase, Laguerre expansion.

1. Introduction
It is well known that internal model control (IMC), which is
shown in Fig. 1, is useful and eective for the attainment of
a desired tracking property in many practical applications [1].
With an internal model P implemented in parallel to the ac-
tual plant P, the controller C I MC compensates the mismatch be-
tween P and P. Ideally, if the model exactly equals to the plant,
IMC completely yields the desired tracking property. In prac-
tice, there is always a dierence between a model and the actual
plant. The parameters of the controller are therefore tuned to
meet the performance requirement. In these cases, constructing
a mathematical model based on the identification is a rational
approach. However, there are also many cases where it is di-
cult to apply persistently excited signals from the viewpoints of
the safe operation of the plant. In addition, it is also preferable
to reduce time and cost from the viewpoints of the management
for the plant. In such cases, the direct use of the data collected
from experiments is also a useful and eective approach. More-
over, since the data have fruitful information on the dynamics
of the plant, it is expected that data-driven approaches yield a
more desirable IMC controller. From such points of view, there
are a number of studies on data-driven approaches to IMC in
literatures, such as F. De Bruyne [2] with iterative feedback
tuning (IFT, [3]), S. Formentin et al. [4] with virtual reference
feedback tuning (VRFT, [5]), O. Kaneko et al. [6][8] with fic-
titious reference iterative tuning (FRIT, [9]) and so on.
IFT, which was used in [2], is a tuning method of a param-
eterized controller by directly minimizing the performance in-
dex which consists of the data. This minimization can be done
by a nonlinear optimization technique, e.g., the Gauss-Newton
Graduate School of Natural Science and Technology, Kanazawa
University, Kakuma-machi, Kanazawa, Ishikawa 9201192,
Japan
Institute of Science and Engineering, Kanazawa University,
Kakuma-machi, Kanazawa, Ishikawa 9201192, Japan
E-mail: hien@moccos.ec.t.kanazawa-u.ac.jp,
o-kaneko@ec.t.kanazawa-u.ac.jp, shigeru@t.kanazawa-u.ac.jp
(Received March 9, 2012)
(Revised July 2, 2012)
Fig. 1 Internal model control (IMC).
method, in which approximations of the gradient, the Hessian,
and so on, consist of experimental data. This means that IFT re-
quires many experiments to update parameters of a controller.
Thus, it takes considerable expense and time, which are crucial
problems with respect to practical points of view. On this point,
VRFT and FRIT have a great advantage with only a one-shot
experiment, then the time and expense for obtaining optimal
parameters are drastically reduced. In [4], VRFT was applied
to a class of linearly parameterized IMC, and hence, the flexi-
bility and the freedom of the controller are limited. In [6][8],
FRIT was utilized for controller parameter tuning of IMC or
the Smith predictor. The approaches proposed in [6][8] treat
the controller whose denominator and numerator are parame-
terized, which implies that more eective tuning of IMC can be
performed.
On the other hand, one of the features of IMC is that the con-
troller contains a plant model internally, thus the data-driven
approach in IMC simultaneously yields a mathematical model
of the plant [6][8]. From practical points of view, the obtained
model is useful for finding out information on model uncertain-
ties, monitoring the actual status, detecting an aging variation
of the plant, re-designing more advanced controllers, and so
on. From theoretical points of view, it is meaningful to study
this issue since there is a crucial interplay that cannot be sepa-
rated between a mathematical model and a designed controller
as stated in [10]. In fact, the reference [11] provides the itera-
tive design of a controller and a model in the IMC architecture
for time delay systems. However, the proposed method in [11]
requires many experiments.
By the way, it is seen that there are many systems with a non-
minimum phase behavior which cannot be eliminated in various

JCMSI 0001/13/06010038 
c 2012 SICE
 SICE JCMSI, Vol. 6, No. 1, January 2013
applications. In these systems, the performance of the feed-
back system is aected due to the appearance of non-minimum
phase zeros, such as an initial undershoot, zero crossings or an
overshoot in the step response. Moreover, the non-minimum
phase zeros also limit the gain margin, bandwidth which there-
fore exacerbate the tradeo between the robustness and achiev-
able performance of a feedback control system [12]. Thus, it
is meaningful to expand the data-driven approach to these sys-
tems. In [13], an application of IFT for non-minimum phase
systems was proposed. The key point is the use of a flexible
reference model with some free parameters in the numerator.
This idea was developed for VRFT in [14] with a two-step pro-
cedure. However, both of them cannot be applied for the simul-
taneous attainment 1 . In the previous work by the authors [8],
they proposed a simultaneous attainment with FRIT based on
the inner-outer factorization of a system model under the as-
sumption that the structure of the non-minimum phase plant is
known. However, in the cases of no knowledge or only a par-
tial knowledge of the non-minimum phase properties (e.g., the
number of unstable zeros, appearance of time delay), it is di-
cult to apply the results in [8].
From these backgrounds, as an extension of [8], we propose a
data-driven parameter tuning method of IMC for non-minimum
phase systems by utilizing FRIT without any information of
non-minimum phase properties. Here, we propose an em-
bedding of the internal mathematical model approximated by
Laguerre expansion for a treatment of the non-minimum phase
systems. Particularly, it is possible to yield a well-approximated
model even if the actual plant contains a time delay. The pro-
posed approach enables us to simultaneously obtain both a de-
sired controller and a mathematical model of the actual plant.
This paper is organized as follows. The preliminaries are
summarized in Section 2. A brief review of FRIT is also ad-
dressed in this section. Section 3 is the main part of this pa-
per where we express the Laguerre expansion approach to non-
minimum phase systems for a simultaneous attainment with
only one-shot experiment. A numerical simulation in the Sec-
tion 4 shows the validity of the proposed method. Finally, some
conclusions are given in Section 5.
2. Preliminaries
2.1 Notations
Let R and Rn denote the set of real numbers and that of real
vectors of size n, respectively. R(s) is a space of all real-rational
transfer functions. Let u and y denote the input and output data
of a plant in the closed loop system, and we use u(t) and y(t)
to describe the value of u and y at time t. By using these
notations, the data obtained in the finite time with the sam-
pling period are described as {u() u(2) u(N)} and
{y() y(2) y(N)}, where N denotes the number of the
sampled data.
For a transfer function described by G(s) = A(s) B(s) such that
A(s) and B(s) are coprime, the output y of G with respect to
u is the solution of the dierential equation B( dtd )y = A( dtd )u.
However, for the enhancement of the readability, we use the
notation y = Gu. Throughout of this paper, we omit the argu-
ment s or t from transfer functions or time series whenever
there is no danger of confusion. For a time signal w, we denote
1
The method in [14] can identify only zeros.
39
Fig. 2 The IMC with a tunable vector .
a value of w at the time t as w(t). To denote a delayed signal of
w with a time lag , we use a simple notation es w to describe
(es w)(t) = w(t ) for all time t, for the enhancement of the
readability. For a time series w = {w() w(2) w(N)}, we
N
use the notation w2N := N1 k=1 (w(k))2 .
2.2 Model Reference Problem
Let us consider a closed loop system with IMC in Fig. 2. The
system we address in this paper is single-input/single-output
(SISO), linear, time-invariant, stable, strictly proper, and non-
minimum phase. We also assume that the system is with no
zero at the origin. Since we have no knowledge of the plant
P, its model P is parameterized by a tunable vector P whose
elements correspond to the unknown coecients of the denom-
inator/numerator polynomials as
P, s + + P,1 s + P,0
P(P ) = (1)
P,+ s + + P,+1 s + 1
where P := [P,0 P,1 P,+ ]. Similarly, the feedback con-
troller C I MC is parameterized by a tunable vector C as
C,d sd + + C,1 s + C,0
C I MC = (2)
C,d+l sl + + C,d+1 s + 1
where C := [C,0 C,1 P,d+l ]. With the notation :=
[C P ], the input u and output y also depend on , and hence,
we denote them as u() and y(), respectively. We use Gry ()
to denote the transfer function from r to y() of the closed-loop
system.
Let T d denote a reference model from r to y of the closed
loop system. The desired output is denoted with yd := T d r. The
problem here is to find a parameter vector that minimizes the
model-reference criterion
J() := y() T d r2N (3)
with the direct use of experimental data.
On the other hand, since the controller contains the tunable
mathematical model of the plant, it is expected that we can also
simultaneously obtain an appropriate model of the actual plant.
Moreover, it is preferable that the simultaneous attainment can
be performed with as few data as possible in a practical sense.
For this purpose, FRIT, which is briefly explained in the next
subsection, is utilized.
2.3 Fictitious Reference Iterative Tuning: FRIT
The main idea of the FRIT scheme is to construct the model-
reference criterion in the fictitious domain. Consider a conven-
tional closed loop system as Fig. 3 where the controller C is
parameterized by a tunable vector .
First, set an initial parameter vector 0 of the controller and
perform a one-shot experiment on the closed loop system to
obtain the data u(0 ) and y(0 ). The controller C(0 ) is assumed
to stabilize the closed loop system such that u(0 ) and y(0 )
 40 SICE JCMSI, Vol. 6, No. 1, January 2013


M
G(s)  k Lk (s, a) (11)
k=1

with the coecients k , k = 1, , M. The time scale a influ-

ences the convergence rate as well as the truncation error of the
Fig. 3 A conventional closed loop system with a tunable vector . approximation. An appropriate choice of the time scale guar-
antees the minimization of the modeling error. It follows from
are bounded. By using the data u(0 ) and y(0 ), the fictitious the discussions in [18][21] that the optimal time scale depends
reference signal r() [15] is computed as on the characteristics of the systems being represented, e.g., the
dominant time constant, the damping, and the time delay.
r() = C 1 ()u(0 ) + y(0 ). (4) In this paper, we use the Laguerre expansion to approximate
The fictitious reference signal was originally proposed by [15] the non-minimum phase part of the system. With a combination
in the unfalsified control framework. However, we use it for the of the Laguerre expansion, it is expected that we can obtain an
dierent purpose. Notice that the fictitious output with respect appropriate model without information on the system, such as
to r() always equals to the initial output y(0 ). Indeed, together the relative degree, time delay, or number of unstable zeros.
with the trivial relation Pu(0 ) = y(0 ), we see Consider a closed loop system with IMC shown in Fig. 2.
For the non-minimum phase plant P, instead of (1), we use the
PC() model P which is parameterized as
r() = y(0 ). (5)
1 + PC()
For a given reference model T d , we then introduce the cost P() = Pm (m )Pn (n ) (12)
function to be minimized as
where
 2
JF () = y(0 ) T d r()N . (6) 
i si
Pm (m ) =  i=0
, (13)
i=1 +i s + 1
i
Note that the cost function (6) with r() in (4) requires only
the initial data u(0 ) and y(0 ). This means that the minimiza-
tion of (6) can be performed o-line by using only one-shot and
experimental data. 
M
Suppose that JF ( ) = 0 for some . Then, by using (5), we Pn (n ) = k Lk (s, a). (14)
see k=1
  2
 PC( )  Here, Pm (m ) is the minimum phase part, and Pn (n ) is the
 T d r( ) = 0 (7)
1 + PC( ) N non-minimum phase one that is approximated by a truncated
Laguerre expansion with the coecients k . The unknown pa-
holds, which also implies that rameter vector for the plant model is defined as
PC( )

= Td (8) := m n (15)
1 + PC( )
generically holds. This means that the minimization of JF () is where
deeply related to attaining the desired tracking property.
m = [0 1 . . . . . . + ] R++1 (16)
3. IMC with FRIT for Non-Minimum Phase Systems:
Laguerre Expansion Approach with constraints such that the numerator and the denominator
3.1 Laguerre Expansion for Non-Minimum Phase Sys- of Pm are Hurwitz polynomials, and
tems
Laguerre functions are bases of a complete orthonormal n = [1 . . . M a] R M+1 . (17)
set on the Lebesgue space of squared integrable functions in
The factorization (12) is a specific approach to handle the
[0, ). The continuous Laguerre functions have Laplace trans-
unstable zeros and/or time delay, which has been introduced in
forms [16]
[1] and [11]. In addition, such a factorization is also convenient

2a s a k1 for treating the reference model in the cases of non-minimum
Lk (s, a) = , a > 0, (9) phase systems, as discussed below.
s+a s+a
As stated above, the time scale a was chosen based on the
where the parameter a is called the time scale.
knowledge of the system to guarantee the precision of the ap-
It is shown in [16],[17] that any stable systems can be exactly
proximation. An appropriate choice of a enables us to decrease
represented by an infinite sum of Laguerre functions as
the term M of Laguerre expansion [18][21]. However, since

there is no any information of the system in our setting, we con-
G(s) = k Lk (s, a) (10) sider a as a parameter that is tuned together with other parame-
k=1
ters of the controller to minimize the model-reference criterion.
with suitable coecients k , k = 1, 2, , . However, in Regarding the reference model T d , since the limitation of the
practice, with a given order M of the expansion, a truncated tracking property is deeply related to the non-minimum phase
Laguerre expansion is used for the approximation, e.g. behavior of a system, T d would have the same unstable zeros
 SICE JCMSI, Vol. 6, No. 1, January 2013 41

and/or time delay as the system. However, it is assumed that

we have no information on the plant. Thus, one of the ratio-
nal ways is to include the unknown unstable zeros and/or time
delay as the plant as a parameterized rational transfer function
in the reference model. In [13], a flexible reference model was
proposed with the modified IFT criterion. This reference model
is described by the Laguerre expansion with unknown coe-
Fig. 4 IMC with the controller described as (19).
cients such as it is tuned towards one which is compatible with
the characteristics of the non-minimum phase system. A sim- Notice that, to achieve a desired performance, the designed sys-
ilar idea of a flexible reference model was developed in [14] tem has to produce a zero steady state error. On the other hand,
with the VRFT. In [14], the numerator of the reference model it is easy to see
is linearly parameterized to identify the unknown zeros of the
system. However, after estimating the unstable zeros (if any), T dm (s)
lim C(, s) = lim M Pm (, s)1 .
one has to modify the gain of T d to keep the gain characteristics s0 s0 1 T dm (s) k=1 k Lk (s, a)
of the given reference model. In [22] and [7], only time delay (22)
is treated as a tunable parameter of the reference model.
In the non-minimum phase cases, we set T d (n ) = T dm Pn (n ) If the reference model T dm is given such that T dm (0) = 1, (22)
as the reference model. Here, T dm is a reference model which is can be rewritten as
given by users and to be with minimum phase property. In [8],
1
Pn (n ) is parameterized with the inner function whose numera- lim C(, s) = M Pm (, 0)1 . (23)
s0 1 k=1 k Lk (0, a)
tor consists of the unknown unstable zeros of the plant. Com-
paring with the approach in [14], the approach in [8] enables us Under the assumption that the plant has no zero at the origin, it
to keep the gain characteristics of the given reference model for is natural to set the minimum phase part of the internal math-
any parameters n . And dierently from [13], we do not have ematical model satisfying Pm (, 0)  0 (i.e., 0  0). Thus,
to modify the standard criterion. However, as we stated above, (23) implies that this controller is rarely equipped with the in-
in the cases where we have no information on the plant struc- tegrator. On the other hand, we require the integrator in C() to
ture, the approach in [8] is with some diculties. Utilization of eliminate the steady state error. To overcome this diculty, we
Laguerre expansion can overcome this issue, and it is expected have the following result.
that we also obtain a relevant model of the plant. By using (14),
the reference model T d is expressed as Theorem 1 Assume that a given reference model T dm satisfies


M lim T dm (s) = 1, (24)
T d (n ) = T dm k Lk (s, a). (18) s0

k=1
and
3.2 Simultaneous Attainment of a Desired Controller and
a Mathematical Plant Model 0  0. (25)
As stated earlier, since IMC involves a mathematical model Then, the steady state error of the closed loop system depicted
internally, it is expected that the achievement of the desired out- in Fig. 4 with respect to the step reference signal r is eliminated
put also yields a mathematical model of the actual plant [6] if and only if
[8],[11]. As shown in [8], if the controller C I MC is parameter-
ized as 
M 
a
(1)k1 k = (26)
2
C I MC () = T dm Pm (m )1 , (19) k=1

the following equivalence holds holds.

Proof The steady state error is eliminated in the closed-loop

P = P() Gry () = T dm Pn (n ). (20)
system in Fig. 4 if and only if
For the detail, refer to [8]. This implies that we also obtain a
lim C(, s) = . (27)
mathematical model of the actual plant by achieving the desired s0
output if we implement (19). The key point here is the structure
Under the assumptions (24) and (25), it follows from (23) that
of the feedback controller (19) illustrated in Fig. 4. Since the
(27) is equivalent to
controller C I MC is parameterized by instead of C , we use
only = [m n ] henceforth. 
M
By rewriting Fig. 4 as Fig. 3, the controller C() can be de- k Lk (0, a) = 1. (28)
scribed as k=1

C I MC () T dm It is easily shown that (28) is also equivalent to (26). This com-

C() = = Pm (m )1
1 C I MC ()P() 1 T d (n ) pletes the proof. 
T dm From the constraint (26), we can express the last coecient
= M Pm (m )1 . (21)
1 T dm k=1 k Lk (a) in (17) as
 42 SICE JCMSI, Vol. 6, No. 1, January 2013


a 
M1
desired output and the attainment of the plant model. For some
M = (1)k1 k (29) parameters  and  , the following three statements are equiv-
2 k=1
alent
for the elimination of the steady state error. Thus, we use the
notation (i) JF ( ) > JF ( ), (36)

n = [1 . . . M1 a] R M (30)  2  2

 T d ( )  0   T d ( )  0 
(ii)  1 y( ) >  1 y( ) ,
henceforth.  Gry ( ) N
Gry ( ) N
(37)
3.3 FRIT for Internal Model Controllers
We apply FRIT for the purpose of obtaining the optimal pa-  2
rameter vector of the controller in Fig. 4. The fictitious refer-  P( ) 

0 
(iii)  1 1 T d ( ) y( )

ence r in (4) can be rewritten as P N
 2
1 T dm Pn (n )  P( ) 

0 
r() = Pm (m )u(0 ) + y(0 ) (31) > 1 1 T d ( ) y( ) . (38)

T dm
where Pm (m ) and Pn (n ) are described as (13) and (14), re-
spectively. Substituting r() into (6), we obtain
JF () = y(0 ) T dm Pn (n )r()2N (32)
together with (26).
As a result of the minimization of JF (), we obtain the opti-
mal vector of the controller
= arg min JF ().
If the minimized value of JF is regarded as suciently small
(JF ( )  0), it is regarded that the optimal parameters yield
a controller C( ) for a desired specification, i.e., Gry ( ) =
T d ( ) generically holds. Moreover, it follows from (20) that
P( ) = P also generically holds.
3.4 Remarks
Remark 1 The results deduced above for the continuous sys-
tem case are easily adapted to the discrete system case. Based
on the discrete Laguerre functions with z-transforms [21]
 k1
1 a2 1 az
Lk (z, a) = , |a| < 1, (34)
z a z a
the condition (26) is modified as

M
1 a
k = . (35)
1 + a
k=1
Namely, the extension to the discrete-time case can be trivially
done.
Remark 2 For a fixed M, the number of the terms of Laguerre
expansion, the optimal a might not yield convergent coecients
for the approximation. In this case, it is expected that increasing
M leads to the improvement of precision of the identification.
However, increasing this term makes calculation more compli-
cated. How to decide the term M as well as the degree and
of Pm is a significant issue that should be addressed as impor-
tant future researches.
Remark 3 In a practical sense, it might be dicult to achieve
JF ( ) = 0. In [8], the authors analyzed the relationship among
the minimization of the cost function, the achievability of the
 P N
See [8] for the more details.
The above equivalence also holds in this paper. In the state-
ment (ii), the relative error of the closed loop Gry () and the
reference model T d () is evaluated. In the statement (iii), the
relative error of the actual plant P and the internal mathematical
model P() is evaluated. Thus, the decrease of JF () leads to
the decreases of the involved quantities on the achievement of
the desired output and the accuracy of the mathematical model
under the influence of y(0 ) and (1 T d ()) y(0 ), respectively.
(33)
3.5 Algorithm
The algorithm of the proposed method is summarized as fol-
lows.
1. Parameterize the minimum phase and the non-minimum
phase parts of a plant model as (13) and (14).
2. The parameter vector is determined by m and n as (16)
and (30), respectively.
3. Set an initial parameter vector 0 and perform a one-shot
experiment to obtain the data u(0 ) and y(0 ).
4. Calculate the fictitious reference signal r() by using (31),
construct the cost function JF () as (32) and minimize it
by an o-line nonlinear optimization.
5. Obtain the optimal parameter vector = arg min JF ()
which yields both a desired controller and a mathematical
model of the plant.
Note that, at each step in the optimization process, m is
checked with the constraints such that the numerator and the
denominator of Pm (m ) are Hurwitz polynomials. The process
will continue if the constraints are satisfied. Otherwise, it stops,
we either take the parameter vector of the previous step or ad-
just the initial parameter and then repeat the procedure.
4. Numerical Example
We apply the proposed approach to an unknown plant
4(2s 3)(s 2)
P= es (39)
(s2 + 5s + 2)(s2 + 4s + 5)
with a time-delay and unstable zeros.
The reference model is given
 SICE JCMSI, Vol. 6, No. 1, January 2013 43

Fig. 5 The reference signal r (the dot-and-dash line), the initial output Fig. 6 The reference signal r (the dot-and-dash line), the optimal output
y(0 ) (the solid line) and the desired output T dm Pn (0n )r (the dotted y( ) (the solid line) and the desired output T dm Pn (n )r (the dotted
line). line).

1
T dm = . (40)
2s + 1
Without information on the plant P, we parameterize the
plant model P with a first-order minimum phase part Pm for
simplicity
0
Pm (m ) = (41)
1 s + 1

and a Laguerre approximation for the non-minimum phase part


M
Pn (n ) = k (a)Lk (s, a). (42) Fig. 7 Step responses: P (the solid line), P( ) (the dotted line).
k=1

Then, the unknown parameter vectors are m = [0 1 ] and

n = [1 M1 a]. We set M = 5 and the initial param-
eter vectors as 0m = [2 2]T and 0n = [0.6 0.3 0.3 0.2 3.0]T .
Then, we perform one-shot experiment on the closed loop sys-
tem to obtain the initial input u(0 ) and output y(0 ), with the
sampling period = 0.01 s. In Fig. 5, the initial output y(0 ),
the reference signal r and the reference output T dm Pn (0n )r are
drawn as the solid line, the dot-and-dash line, and the dotted
line, respectively. By applying the proposed algorithm, the op-
timal parameters are obtained as m = [2.3942 2.1744] and
n = [0.1582 0.4298 0.2394 0.2789 2.9957] 2 . We implement
these parameters to the closed loop system in Fig. 4 and per- Fig. 8 Gain characteristics: P (the bold solid line), P( ) (the dotted line),
form the final experiment. The result is shown in Fig. 6. In and T dm Pn (n ) (the thin solid line).
this figure, the optimal output y( ), the reference signal r and
the desired output T dm Pn (n )r are described by the solid line,
the dot-and-dash line and the dotted line, respectively. From
Fig. 6, we see that the actual output y( ) and the desired out-
put T dm Pn (n )r are almost the same, which implies that a de-
sired controller is achieved by using .
In addition, with the obtained parameters, we can compare
the actual plant P and its model P( ). Figure 7 draws their
step responses in the open-loop with the solid line and the dot-
ted line, respectively. As shown in Fig. 7, the outputs of the
actual plant and its model are well matched. Moreover, their
frequency characteristics are also compared in Figs. 8 and 9. In
these two figures, characteristics of P, P( ) and T dm Pn (n ) are
Fig. 9 Phase characteristics: P (the bold solid line), P( ) (the dotted
illustrated by the bold solid line, the dotted line, and the thin line), and T dm Pn (n ) (the thin solid line).
solid line, respectively. It is seen that the frequency characteris-
tics of P and those of P( ) are almost the same in the frequency
range of the reference model T d . These illustrations show that the model P( ) appropriately reflects the dynamics of the ac-
2
In this example, we perform Gauss-Newton method for the o- tual plant in the time domain as well as frequency domain even
line optimization. if the process contains a time delay.
 44 SICE JCMSI, Vol. 6, No. 1, January 2013
5. Conclusion
This paper has expressed a data-based controller tuning
method for non-minimum phase systems by utilizing fictitious
reference iterative tuning (FRIT) to the IMC architecture. With
the combination of the Laguerre expansion, the approach has
shown its validity for the case of non-minimum phase systems
without any information on the non-minimum phase property.
The proposed approach enables us to obtain not only a desired
controller but also a mathematical model of the actual plant.
As the future works, some theoretical issues, such as the de-
cision on the degree of the minimum phase term or the term of
Laguerre expansion, should be clarified.
This work was partially supported by the JSPS Grant
in Aid for Scientific Research (B) No. 23360183 and (C)
No. 22560442.
References
[1] M. Morari and E. Zafiriou: Robust Process Control, Prentice
Hall, 1989.
[2] F. De Bruyne: Iterative feedback tuning for internal model
controllers, Control Engineering Practice, Vol. 11, No. 9,
pp. 10431048, 2003.
[3] H. Hjalmarsson, M. Gevers, S. Gunnarsson, and O. Lequin:
Iterative feedback tuning: Theory and applications, IEEE Con-
trol Systems Magazine, Vol. 18, No. 4, pp. 2641, 1998.
[4] S. Formentin, M. Corno, S.M. Savaresi, and L. Del Re: Virtual
reference feedback tuning of internal model controllers, Pro-
ceedings of the 49th IEEE Conference on Decision and Con-
trol, pp. 55425547, 2010.
[5] M.C. Campi, A. Lecchini, and S.M. Savaresi: Virtual reference
feedback tuning: A direct method for design of feedback con-
trollers, Automatica, Vol. 38, No. 8, pp. 13371346, 2002.
[6] O. Kaneko, Y. Wadagaki, and S. Yamamoto: Fictitious refer-
ence iterative tuning for internal model controller, Proceedings
of the 10th IFAC Workshop on Adaption and Learning Control
and Signal Processing, Fr-2A-1, CDROM, 2010.
[7] O. Kaneko, S. Yamamoto, and Y. Wadagaki: Simultaneous at-
tainment of model and controller for linear time delay systems
with the data-driven Smith compensator, Preprints of the 18th
IFAC World Congress, pp. 76847689, 2011.
[8] O. Kaneko, H.T. Nguyen, Y. Wadagaki, and S. Yamamoto: Fic-
titious reference iterative tuning for non-minimum phase sys-
tems in the IMC architecture: Simultaneous attainment of con-
trollers and models, SICE Journal of Control, Measurement,
and System Integration, Vol. 5, No. 2, pp. 101108, 2012.
[9] S. Souma, O. Kaneko, and T. Fujii: A new method of controller
parameter tuning based on input-output data Fictitious Ref-
erence Iterative Tuning (FRIT), Proceedings of the 8th IFAC
Workshop on Adaption and Learning Control and Signal Pro-
cessing, pp. 788794, 2004.
[10] R.E. Skelton: Model error concepts in control design, Interna-
tional Journal of Control, Vol. 49, No. 5, pp. 17251753, 1989.
[11] N. Abe and H. Ichihara: An iterative design of closed-loop
identification and control using IMC structure for time-delay
systems, Transactions of the SICE, Vol. 36, No. 7, pp. 563
568, 2000 (in Japanese).
[12] J.B. Hoagg and D.S. Bernstein: Non-minimum phase zeros:
Much to do about nothing: Classical control revisited, Part II,
IEEE Control Systems Magazine, Vol. 27, No. 3, pp. 4557,
2007.
[13] A. Lecchini and M. Gevers: On iterative feedback tuning for
non-minimum phase plants, Proceedings of the 41st IEEE Con-
ference on Decision and Control, pp. 46584663, 2002.
[14] L. Campestrini, D. Eckhard, M. Gevers, and A.S. Bazanella:
Virtual reference feedback tuning for non-minimum phase
plants, Automatica, Vol. 47, No. 8, pp. 17781784, 2011.
[15] M.G. Safonov and T.C. Tsao: The unfalsified control con-
cept and learning, IEEE Transactions on Automatic Control,
Vol. 42, No. 6, pp. 843847, 1997.
[16] B. Wahlberg: System identification using Laguerre models,
IEEE Transactions on Automatic Control, Vol. 36, No. 5,
pp. 551562, 1991.
[17] B. Wahlberg and P.M. Makila: On approximation of stable lin-
ear dynamical systems using Laguerre and Kautz functions,
Automatica, Vol. 32, No. 5, pp. 693708, 1996.
[18] G.J. Clowes: Choice of the time-scaling factor for linear system
approximations using orthonormal Laguerre functions, IEEE
Transactions on Automatic Control, Vol. 10, No. 4, pp. 487
489, 1965.
[19] L. Wang and W.R. Cluett: Optimal choice of time-scaling
factor for linear system approximations using Laguerre mod-
els, IEEE Transactions on Automatic Control, Vol. 39, No. 7,
pp. 14631467, 1994.
[20] T. Oliveira e Silva: Optimality conditions for truncated
Laguerre networks, IEEE Transactions on Signal Processing,
Vol. 42, No. 9, pp. 25282530, 1994.
[21] Y. Fu and G.A. Dumont: An optimum time scale for discrete
Laguerre network, IEEE Transactions on Automatic Control,
Vol. 38, No. 6, pp. 934938, 1993.
[22] S. Masuda, M. Kano, Y. Yasuda, and G.D. Li: A fictitious refer-
ence iterative method with simultaneous delay parameter tun-
ing of the reference model, International Journal of Innovation
Computing, Information and Control, Vol. 6, No. 7, pp. 2927
2939, 2010.
Hien Thi NGUYEN
She received her B.E. and M.E. degrees from the Hanoi
University of Agriculture, Vietnam in 2000 and 2002,
respectively. In 2001, she joined Faculty of Engineer-
ing, Hanoi University of Agriculture, as a Lecturer. She
is currently a doctoral student at Kanazawa University,
Japan. Her research interests include control systems and
its applications.
Osamu KANEKO (Member)
He received his B.E. and M.E. degrees from the Na-
gaoka University of Technology, Japan, and Ph.D. degree
from Osaka University, Japan, in 1992, 1994, and 2005,
respectively. In 1999, he joined Osaka University as an
Assistant Professor. In 2009, he moved to Kanazawa Uni-
versity as an Associate Professor. His research interests
include control system design and system theory.
Shigeru YAMAMOTO (Member)
He received his B.E., M.E., and Ph.D. degrees from
Osaka University, Japan in 1987, 1989, and 1996, respec-
tively. In 1989, he joined Osaka University as an Assis-
tant Professor of the Faculty of Engineering. In 1994, he
moved to the Faculty of Engineering Science, Osaka Uni-
versity, and served there as a Lecturer and an Associate
Professor. In 2007, he moved to Kanazawa University
as a Professor. His research interests include control system design and
system theory.