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

net/publication/273187545

Control Design of Mixed Sensitivity Problem for Educational Model of

Helicopter

Article  in  Advances in Electrical and Electronic Engineering · December 2014

DOI: 10.15598/aeee.v12i5.1150

CITATIONS READS

0 168

4 authors:

Stepan Ozana Petr Vojčinák

VŠB-Technical University of Ostrava VŠB-Technical University of Ostrava
88 PUBLICATIONS   268 CITATIONS    8 PUBLICATIONS   57 CITATIONS   

SEE PROFILE SEE PROFILE

Martin Pieš Radovan Hajovsky

VŠB-Technical University of Ostrava VŠB-Technical University of Ostrava
60 PUBLICATIONS   247 CITATIONS    51 PUBLICATIONS   230 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Inverted Pendulums View project

THE CONCEPT OF WIRELESS SENSOR NETWORK FOR EARLY WARNING IN THE CASE OF ACCIDENTAL RELEASES View project

All content following this page was uploaded by Stepan Ozana on 30 September 2015.

The user has requested enhancement of the downloaded file.

CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

Control Design of Mixed Sensitivity Problem for

Educational Model of Helicopter

Stepan OZANA , Petr VOJCINAK , Martin PIES , Radovan HAJOVSKY

Department of Cybernetics and Biomedical Engineering, Faculty of Electrical Engineering and Computer
Science, VSB–Technical University of Ostrava, 17. listopadu 15, 708 33 Ostrava-Poruba, Czech Republic

[email protected], [email protected], [email protected], [email protected]

Abstract. The paper deals with the design of H-∞ ro- In other words, robustness plays a significant role
bust controller, particularly with mixed sensitivity prob- in the design of control systems as real systems are
lem for elevation control. It briefly introduces basic prone to external disturbances and measurement noise.
mathematical background concerning robust control ap- Moreover, there are often differences between the pro-
proach, which is then applied for typical example of posed mathematical models and actual real systems.
MIMO system, that is a helicopter model. The obtained A typical example is the design of a controller that
results are verified on real educational physical model will stabilize the system even if it is originally unsta-
CE 150 by Humusoft, ltd. ble and accept a particular level of performance at the
presence of disturbance signals, noises, hard-to-model
process dynamic characteristics or process parameter
variables. Such tasks are best solved by a feedback
Keywords control mechanism as they bring along a whole range
of problems according to [4]:
Algorithms and software, simulation of dy-
namic systems, robust control of nonlinear sys- • high price (e.g. use of sensors),
tems.
• system complexity (e.g. possibilities of implemen-
tation and reliability),

1. Introduction • system stability (e.g. requirement for internal sta-

bility and stabilizing controllers).
The objective of the robust control is to design a The need and significance of robustness as a part
dynamic control system operating in a real environ- of control system designs have been developing since
ment. The changes to the surrounding conditions can 1980s. Robustness in the standard SISO control is
be caused by the following factors according to [15]: provided by a suitable gain margin and phase mar-
gin. When the first design techniques for multi-variable
• component aging, systems developed in 1960s, emphasis was laid on the
achievement of good performance, not robustness. The
• temperature effect, methods that use multivariables were based on the
linear–quadratic criterion and Gaussian disturbances.
• effect of the working environment. It was demonstrated that they can be successfully used
in a whole range of aviation applications where it is
The control system must not only be resistant to possible to set up precise mathematical models, in-
the aforementioned factors but it also must eliminate cluding the descriptions of external disturbance signals
inaccuracy of the model, i.e. robustness is the relevant or noises. However, the application of these meth-
ability of the control system to accept changes. The ods, called LQG methods (linear-quadratic Gaussian
required output value will be reached even when the control), in other industrial branches evidently showed
changes in the properties of the controlled system are their bad properties from the point of robustness which
limited and constant disturbance signals are operating. led to the effort to develop a theory that would explic-
From the mathematical point of view, the robust con- itly deal with the issue of robustness in the control
troller is not only suitable for one particular system feedback design. The pioneering work on the devel-
but for a set of systems [15]. opment of the theory, today known as the theory of

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 488
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

optimal control in H-∞, was introduced at the begin- We use the rules of block algebra to define the mathe-
ning of 1980s by Georg Zames and Bruce A. Francis. matical relations within the control circuit, i.e. accord-
The H-∞ approach first specified the model of system ing to [15], [9]:
uncertainty, i.e. additive perturbation and/or output
disturbances. In most cases, it is enough to find a suit- • for the open-loop transfer function:
able controller so that the closed loop achieves some
robust stability. The performance is also a part of the
L (s) = K (s) G (s) , (1)
optimization loss (objective) function. The elegant for-
mulations of the solution are based on the solutions of
Riccati equations, e.g. in MATLAB [4]. • for the transfer function of the control error - the
sensitivity function:
If we design a controller for a particular interval of
parameters, then the control circuit is robustly stable.
E (s) 1
Another important parameter of robust controllers is GE (s) = = = S (s) , (2)
their performance, meeting the requirements for pa- W (s) 1 + L (s)
rameters according to [15]:
• for the closed-loop transfer function - the comple-
• control, mentary sensitivity function:

• disturbance, Y (s) L (s)

GW (s) = = = T (s) , (3)
W (s) 1 + L (s)
• speed of response (settling time).

• limiting condition applies to the sum of the sen-

The problem of the design of the robust controller
sitivity function and complementary sensitivity
is based on the feedback circuit (closed loop) which
function:
enables working with the sensitivity and elimination of
the disturbance. On one side, the feedback of a non- 1 L (s)
stable system stabilizes it; on the other side, it may S (s) + T (s) = + = 1. (4)
1 + L (s) 1 + L (s)
destabilize a stable system [15]. Consider the standard
control diagram according to Fig. 1.
Consider the requirements for the sensitivity func-
tion and complementary function according to [15]:

• at v1 (t) = v2 (t) = 0, the effect of the control ac-

tion predominates and so the sensitivity function
ĜE (s) = S (s) must be small and the complemen-
tary function GW (s) = T (s) will be large,
Fig. 1: Classical control scheme with the definition of circuit
signals.
• the entire control circuit carries out the elimina-
tion of the low-frequency noise v1 (t), and so again,
Description of the signals in the control circuit: the effect of the control action predominates: the
sensitivity function GE (s) = S (s) must be small
• W (s): Laplace transform of the the reference sig- and the complementary function GW (s) = T (s)
nal, will be large,

• E (s): Laplace transform of the control error sig- • the entire control circuit must not affect the elimi-
nal, nation of the high-frequency noise v2 (t) and so we
eliminate the effects of the control action and con-
• U (s): Laplace transform of the manipulated value trol error, i.e. the sensitivity function GE (s) =
signal, S (s) must be small and the complementary sensi-
tivity function GW (s) = T (s) will also be small.
• V1 (s): Laplace transform of the disturbance: low-
frequency known and unknown disturbances; the
system must eliminate them, As the aforementioned opposing requirements can-
not be met by one controller, it is necessary to find a
• V2 (s): Laplace transform of the disturbance: sen- compromise between the sizes of the sensitivity func-
sors or measurement, high-frequency character tion and the complementary sensitivity function [15],
with insignificant effect. see Fig. 2.

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 489
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

H∞ controllers have their own terminology, notation

and conception. This method leads to a set of suit-
able stable transfer functions that are physically viable.
Similarly to LQR and LQG controllers, we expect op-
timization of the objective function that will compare
different transfer functions and select the most suitable
one from the set. The requirements for the closed loop
are the following, i.e. according to [15]:
Fig. 2: Logarithmic amplitude frequency characteristics for the
sensitivity function (on the left) and complementary
sensitivity function (on the right). • Physical viability: The order of the transfer func-
tion denominator must be higher or equal to the
order of the transfer numerator.
2. Robust Controller Design
• Stability: The transfer poles must lie in the left
half plane of the Gaussian plane or in the area of
2.1. Robust Design Methods the Laplace transform convergence (provided that
the control straight line and imaginary axis are
The methods of the state space in the time domain identical).
allowed to avoid the problems with transfer function
matrices and also provided means of the analysis and The basic prerequisite of the H∞ method is the
design of MIMO systems with more inputs and out- knowledge of the transfer function of the given system,
puts. Approximately at the same time when the meth- evaluating ∞-norm according Fig. 5 according to [15],
ods of optimal control were being developed, research [1]:
focused on the extensions of means of MIMO system kGk∞ = supω {|G (jω)|} . (5)
standard control was conducted. The robust design is
based on the finding of such a controller so that the The norm can be graphically represented as the max-
resulting system in the closed loop is also robust. Ro- imum of the Bode diagram provided that the transfer
bustness became the main standpoint in the field of function is definite and has no imaginary poles, while
control, therefore specifications and methods followed its objective is to minimize it in the ∞-norm. It de-
shortly, i.e. according to [15], [5]: creases the apex of the Bode diagram, which increases
the robust stability margin [15].
• H∞ method,

• H2 method, 2.3. Mixed Sensitivity Problem

• LTR method (loop transfer recovery),
Usually, practical industrial applications do not only
• µ: synthesis, use one objective function but a combination of several
functions like that, e.g. accomplishment of the good
• QFT method (quantitative feedback theory), performance of tracking the reference signal at limited
energy of the reference signal. Then we solve the mixed
• Kharitonov theorem for the examination of robust sensitivity task, or the ’S over KS’ problem defined by
stability, a general relation (for the SISO system) according to
• specification of the small-gain theorem, [8], [12]:

• specification of structured singular values.
S (s)
minKst =
K (s) S (s) ∞
−1
The following presentation will only focus on the H∞ [1 + L (s)]
= minKst
−1 .
(6)
method. K (s) [1 + L (s)] ∞

The Eq. 6 can also be expressed by the requirements

2.2. H∞ Method
of the design concerning the additive perturbation, e.g.
nominal behavior, good performance of tracking the
A control system is robust if it stays stable and meets
reference signal or the elimination of disturbance sig-
particular behavioral criteria at the presence of possi-
nals and robust stability [4].
ble uncertainties. The H∞ optimization method, de-
veloped since 1980s, has proved to be a very efficient Figure 3 shows the standard block diagram of the
and potent design method for robust control in the field H∞ configuration using the linear fractal transforma-
of linear, time-invariant control systems [15], [11], [3]. tion (LFT) with specification of the individual external

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 490
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

Fig. 3: Standard block diagram at the H∞ configuration.

inputs, external outputs, inputs into the controller and Fig. 4: Standard block diagram of the mixed sensitivity prob-
its outputs, see [1] (p. 438). The control circuit con- lem: controller and perturbed system containing nomi-
tains a robust controller with transfer function K (s) nal system, control error and manipulated value weight-
ing filters.
and a perturbed (also extended or generalized) system
with transfer function P (s) that has two inputs and
two outputs according to [15]: Figure 4 shows the standard block diagram of the
mixed sensitivity problem and it is basically a more
• w
~ (t) input reference signal vector; external input detailed illustration of Fig. 3, where it is easy to deduce
signals, the following relations, i.e.:

• ~v (t) output manipulated signal vector; output • for external input signals:
control signals from the controller.
w
~ (t) = r (t) , (11)
The main difference between the vectors is that the
• for output control signals from the controller:
controller does not affect the inputs. The input refer-
ence signal vector w
~ (t) includes an external noise, noise ~u (t) = u (t) , (12)
from the sensors and tracking (reference) signals. To
the contrary, the outputs from the system are divided • for measured outputs:
into two groups according to [15]:
~y (t) = e (t) , (13)
• ~y (t) output signal vector; measured outputs,
• for minimized or penalized outputs:
• ~z (t) controlled outputs; minimized or penalized ~z = [z1 (t) , z2 (t)] =
T
outputs. T
= [W1 (s) e (t) , W2 (s) u (t)] . (14)
The task is then defined so that the internally sta- The following applies to the generalized system con-
bilizing controller K (s) is searched for in the control taining weighting filters according to [4]:
circuit of the robust control for the given generalized  
system P (s) that minimizes or penalizes the controlled P11 (s) P12 (s)
P (s) = , (15)
output vector ~z (t). In other words, we minimize the P21 (s) P22 (s)
maximal norm of the transfer function between w ~ (t)
while:
and ~z (t) by the given relation according to [4]:
T T
P11 (s) = W1 (s) [I, 0] = [W1 (s) , 0] , (16)
~z = P11 (s) + P12 (s) K (s) IP−1K P21 (s) w,

~ (7)
T
P12 (s) = [−W1 (s) G (s) , W2 (s) I] , (17)
where: T
IP K = I − P22 (s) K (s) , (8) P21 (s) = [I] = I, (18)
T
P22 (s) = [−G (s)] = −G (s) . (19)
we get a linear fractal transformation after the adjust-
ment: The weighting filters W1 (s) and W2 (s) are com-
~z = F l [P (s) , K (s)] w.
~ (9) monly used in practice; in such case, the Eg. 6 can
be formally adjusted into the form describing a objec-
Then, the H∞ optimization problem can be ex- tive function, i.e. (for the SISO system) according to
pressed by a relation, i.e. according to [4]: [4]:
W1 (s) S (s)
min kFl [P (s) , K (s)]k∞ . (10) min . (20)
Kst Kst W2 (s) K (s) S (s) ∞

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 491
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

3. Controller Design for the algebraic complement. The situation is depicted in

Fig. 5, directly modified for the helicopter model.
Helicopter Elevation
uM mech.part
On the basis of the aforementioned theoretical findings G1(s) M1(emp.)(s)
U1(mot) (elevation)
from the field of the H∞ robust control, we can design
a robust controller according to the selected conception G11(s)
when we deal with the question of the autonomy of the
control and the problem of mixed sensitivity and also
consider the input signals. Derived mathematical mod- Gr (s) Mr(emp.) (s)
U1(reak)
els in elevation and azimuth are crucial for the design
of controllers, when we design the controller for both R21(s)
G21(s)
the mathematical model and real model. However, this
paper only focuses on elevation, due to the extent of uT + + mech.part
the task. Finally, both responses are compared to one + G2(s) M2(emp.)(s) +
(azimuth)
U2(mot)
another.
The mathematical description of helicopter model is G22(s)
taken from official manual, see [6]. The design of eleva-
Fig. 5: Block diagram for the elimination of the elevation-
tion controller particularly comes out from Figs. 2.1, azimuth coupling; the azimuth-elevation coupling is ne-
2.2 stated on pages 8, 15 of this manual. glected.

If the manipulated value expressed by voltage UM

3.1. Autonomy Requirement induces an undesirable response in the output of the az-
imuth mechanical part described by the Laplace trans-
The main problem of this MIMO system namely in- form G21 (s) UM (s), then it can be completely com-
cludes the elimination of the elevation-azimuth cou- pensated by the correction term R21 (s) under the con-
pling. All in all, we naturally want to eliminate both dition according to [2]:
relations but we know that the azimuth-elevation rela-
tion is not as significant. The conception is thus based
on the assumption of designing two independent con- G21 (s) UM (s) + G22 (s) R21 (s ) UM (s) = 0
trollers, i.e. an elevation controller and an azimuth G21 (s) . (22)
⇒ R21 (s) = − G 22 (s)
controller.
The definition of autonomy says that the reference The transfer function of the second correction term
signal w
~ (t) must only affect just one corresponding can be either written directly with the use of the prin-
output signal ~y (t), [7]. ciple of cyclical substitution of indexes or a conditional
Autonomy also requires that the transfer matrix of equation can be set up again Eq. (23), i.e. according
the open loop is diagonal, the elements of the matrix to [2]:
are only located on the main diagonal. A priori, diago-
nality is required in the control matrix. With respect to
the explicitly written sign, the transfer function of the G12 (s) UT (s) + G11 (s) R12 (s) UT (s) = 0
G12 (s) . (23)
correction term is given by the general relation based ⇒ R12 (s) = − Ĝ
11 (s)
on the theory of matrix determinants (Laplace comple-
ment) according to [2]: The relation Eq. (23) describes the azimuth-
 ∗  elevation coupling that, however, is not significant, and
G (s)
(21) thus can be written according to [2]:
j+1 j,i
Ri,j (s) = (−1) Rj,j (s)  ∗ ,
G (s)
j,j

where i is matrix line index, j is matrix column R12 (s) = G12 (s) = 0. (24)
j+1  ∗in-
dex, (−1) is Laplace (algebraic) complement,  Gj,i 
 The diagram in Fig. 5 can also be interpreted in the
is matrix determinant G∗j,i (s) and G∗j,j  is matrix de-
following way: the internal physical coupling between
terminant G∗j,j (s).
the elevation and azimuth in the form of transfer func-
The relation Eq. (21) rather has a theoretical char- tion G21 (s) cannot be eliminated without the basic
acter. Directly derived conditions of autonomy pay (constructive) interference into the system. However,
off at a small amount of regulated signals (in our case we can quite easily implement an external connection
there are two signals) rather than having to exactly between the inputs of the helicopter set by the cor-
remember its content and avoid making a mistake in rection term R21 (s) that will ensure the same as the

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 492
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

unfeasible elimination of the cross-coupling inside the e(t) z1(t)

We (s)
MIMO system transfer matrix, decomposition of the
model into two models that seemingly do not influence
u(t) z2(t)
one another. Their control can then be ensured by two Wu (s)
control loops independent of one another [2]. d2(t) d2(t)
Wd (s)
If we substitute in the relation Eq. (23), we will re-
ceive transfer function of the correction term, Eq. (25): d1(t) d1(t) +
Wcmd (s) + K(s) G0(s) +
2
Tr2 s + Tr1 s + 1 ar UM s + br - E(s) U(s) Y(s)
R21 (s) = − A (s) , (25) d2(t)+y(t)
B (s) a2 UT s + b2

2
2 +
where A (s) = T2 s2 + 1 , B (s) = T1 s2 + 1 . The d2(t)+d3(t)+y(t) +
relation (Eq. 25) clearly shows that the transfer func- d3
( t) d ( t)
Wnoise(s) 3
tion in this form does not meet the condition of phys-
ical viability as the numerator order is higher than Fig. 6: Block diagram of the connection of the H robust con-
∞
the denominator order. Thus we define two inertias troller, nominal system and weighting filters for the he-
with conditions ξ1  Tr1 and ξ2  Tr2 . The phys- licopter model.
ically feasible correction transfer function eliminating
the elevation-azimuth coupling is given by the relation
Eq. (26): suboptimal robust controller K (s) with one degree of
freedom (1DOF configuration).
Tr2 s2 + Tr1 s + 1 ar 0.55s + br A (s)
R21 (s) = − . (26) The meaning of the individual blocks is as follows
ξ2 s2 + ξ1 s + 1 a2 0.2s + b2 B (s)
according to [10]:

3.2. Controller Conception • Wcmd (s): this weighting transfer function takes
care of the reference tracking. A normalized signal
The conception of the helicopter model robust con- appears at the input and the signal at the output
trol design was partially explained in the requirements is in relevant physical units,
for the autonomy of the elevation and azimuth con-
trol. The second part concerns the H∞ robust control, • Wd (s): this weighting transfer function adjusts
namely the modification of the mixed sensitivity prob- the frequency and amplitude characteristics of ex-
lem (MSP) where we also penalize the group of external ternal low-frequency disturbance signals affecting
output signals in addition to the control error ~e (t) and the nominal system,
manipulated value ~u (t):
• Wnoise (s): this weighting transfer function rep-
• d~1 (t) = r (t) = w
~ (t) reference or external control resents the models of noises of sensors in the fre-
signal, quency domain. It tries to detect a particular piece
of information in the control derived from labo-
• d~2 (t) low-frequency signal (disturbance), ratory experiments or production measurements.
• d~3 (t) high-frequency disturbance signal (noise). Naturally, the noise shows a high-frequency char-
acter,
We use weighting transfer functions or also called
• We (s): this weighting transfer function penalizes
weighting filters to penalize signals incoming and out-
the control signal from the robust controller and
going from the extended system in the elevation or az-
thus the control error. It determines the inverted
imuth. The whole block diagram of the control circuit
value of the expected form of the output signal.
with extended system is shown in Fig. 6.
The signal that appears at the input of the filter
There are many different ways for the extension of is in relevant physical units and it is normalized
the nominal system. However, the more external in- at the output,
puts and penalized (error) outputs there are, the more
difficult it is to select the weighting filters. The weight- • Wu (s): this weighting transfer function penalizes
ing filters are generally stable transfer functions (not the control signal from the robust controller and
necessarily proper rational functions) of a particular thus the manipulated value signal. It determines
order. Thus, the more we add, the higher the order the inverted value of the expected form of the out-
the resulting system will have. Such an interconnected put signal. The signal that appears at the input
system can be then used to express the state descrip- of the filter is in relevant physical units and it is
tion, or the transfer function of the H∞ optimal, or normalized at the output,

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 493
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

The complete calculation of both the transfer func- • the closed-loop transfer functions is calculated
tion of the H∞ robust controller and the key H∞ norm with the use of the linear fractal transformation
is executed in MATLAB. The calculation algorithm CL = F {P (s) , K (s)},
is based on the correct interconnection of the nomi-
nal system with the weighting filters in correspondence • γ0 = kCLk∞ = kF {P (s) , K (s)}k∞ .
with Fig. 6. The program solution in the M-file appears
as follows, i.e.: MATLAB, namely the Robust Control Toolbox, con-
tains other functions that can be used to solve the issue
of the design of a continuous or discrete H∞ robust con-
systemnames = ’G Wcmd Wd Wnoise We Wu’; troller. For completeness, we only give the prototype of
inputvar = ’[d1; d2; d3; u]’; the function focused on the standard mixed sensitivity
outputvar = ’[We; Wu; Wcmd-G-Wnoise]’; problem:
input_to_Wcmd = ’[d1]’;
input_to_Wd = ’[d2]’; [K, CL, gopt, INFO] = mixsyn(G, W1, W2, W3)
input_to_Wnoise = ’[d3]’;
input_to_G = ’[u]’; The problem of the mixsyn function is the number
input_to_We = ’[Wcmd-G-Wnoise]’; of the weighting filters and their character as they only
input_to_Wu = ’[u]’; penalize the control error (W1), manipulated value (W2)
cleanupsysic = ’yes’; and measured output (W3). With regard to the selected
P = sysic design conception, it would not be possible to penalize
NControl = 1; inputs with this function.
NMeasure = 1;
r = [NControl NMeasure];
[K,CL,gopt] = hinfsyn(P,NMeasure,NControl); 3.3. Elevation Controller for
Mathematical Model
First of all, we define what systems we will in-
terconnect, the variable systemnames. Then we de- The transfer function of the dynamics of the mathe-
fine the input signal vector (external control signals matical model in elevation is given by the relation ac-
and control signal from the controller), the variable cording to the relation Eq. (27):
inputvar. The output is represented by the vari- Ψ (s)
able outputvar containing the penalization of the con- Gψ (s) = =
UM (s)
trol error (We), manipulated value (Wu) and the mea- 7.3315s + 1.1883
sured output (Wcmd-G-Wnoise). Subsequently, we con- = 4 . (27)
3s +23s3 +116s2 +519s + 1000
nect all the inputs to the weighting filters. By the
cleanupsysic command with the attribute value set The amplitude and phase frequency characteristics
to yes we confirm that we want to remove the vari- are shown in Fig. 7 which also clearly show that it con-
ables systemnames, inputvar and outputvar from the tains the highest value under the following conditions:
MATLAB work environment (Workspace) immediately
after the creation of the system interconnection. • frequency: ωMAX = 4.9448 rad · s−1 ,
 

The variable P represents the extended system or sys-

• transfer function module:
tem interconnection (sysic, System Interconnection).
To complete the enumeration of parameters for the cal- |Gψ (jω)|M AX =
culation, we have to define the number of control out- = −16.05 [dB] ∼= 0.1576 [−] . (28)
puts from the control (NControl) and the number of
measured outputs (NMeasure). We will obtain the cal- The system is of the fourth order and contains four
culation of the H∞ controller (K), closed loop transfer stable poles, out of which two are complex conjugate
(CL) and maximum closed loop transfer norm (gopt) and one is a double pole:
by activating the hinfsyn function with the following
parameters: P, NMeasure and NControl. p1 = −0.2105 + j4.9448, (29)
The activation of the hinfsyn function can also be p2 = p1 = −0.2105 − j4.9448, (30)
extended by more input and output parameters; in this
p3 = p4 = −4. (31)
actual case according to [10]:
The maximum norm of the given system is (in accor-
• two algebraic Riccati equations are solved, dance with the maximum value of the transfer function
module):
• γ ∈ (0, +∞), kGψ (s)k∞ = 0.1577 [−] . (32)

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 494
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

0
-50
Phase (deg) Magnitude (dB)

-10

Phase (deg) Magnitude (dB)

-100 -20

-30
-150
-40

-200 -50
90
-60
0
0

-90
-45
-180

-270
-2 -1 0 1 2 3
10 10 10 10 10 10 -90
-5 -4 -3 -2 -1 0 1
Frequency (rad/s) 10 10 10 10 10 10 10
Frequency (rad/s)
Fig. 7: Amplitude frequency characteristic and phase frequency -100
characteristic of the given transfer function of the math-
-110

Phase (deg) Magnitude (dB)

ematical model in elevation.
-120

-130
The forms of the weighting filters for the given sys-
tem are as follows, i.e.: -140

-150
90
• weighting transfer function for reference signal:
1
Wcmd (s) = , (33) 45
0.25s + 1

• weighting transfer function for low-frequency dis-

0
turbance signal: -2 -1 0 1 2 3 4
10 10 10 10 10 10 10

0.5 Frequency (rad/s)

Wd (s) = , (34)
0.1s + 1 Fig. 8: Amplitude and frequency characteristics of the low-
pass weighting filter We (s) (upper part) and high-
• weighting transfer function for high-frequency dis- pass weighting filter Wu (s) (lower part): mathematical
turbance signal (noise): model in elevation.

0.01s + 1
Wnoise (s) = , (35)
s+1
150
S
• weighting transfer function for control error signal:
Singular Values (dB)

100 T
1/We
1 1/Wu
s + ωbe s + 0.5 50
We (s) = Ke Me = 0.001 , (36)
s + ωbe εe s + 0.0005 0

• weighting transfer function for manipulated value -50

signal: -100
ωbu
s+ Mu s+1
Ŵu (s) = Ku = 10−7 . (37) -150
εu s + ωbu 0.01s + 2
-200

The weighting filters for the control error and manip-

-4 -2 0 2 4
ulated signal have a prescribed transfer function form 10 10 10 10 10

according to [14], the transfer always contains the same Frequency(rad/s)

numerator and denominator order as to ensure the sta- Fig. 9: Amplitude frequency characteristic of the sensitiv-
bility of the inverted transfer functions. The filter for ity function S (s), complementary sensitivity func-
the control error is low-pass and for the manipulated tion T (s), inverted transfer functions 1/We (s) and
1/Wu (s): mathematical model in elevation.
value it is high-pass. The graphic dependences of the
sensitivity functions and inverted transfer functions for

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 495
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

the mathematical model in elevation are as stated in The order of the designed H∞ controller for the
Fig. 8. mathematical elevation model corresponds with the to-
tal of the orders of the individual elements of the ex-
One of the general requirements for the amplitude
tended system, the nominal system is of the fourth or-
frequency characteristics of the sensitivity function
der at the most and all five weighting systems are in
S (s) and the inverted transfer function 1/We (s) de-
the first order at the most. The aforementioned implies
fines robust behavior:
that the controller will be a system of the ninth order
at the most. Its frequency characteristics are shown in
∀ ω ∈ R : |S (jω)| ≤ |1/We (jω)| = Fig. 11.
(38)
= 1/ |We (jω)| ⇔ kWe (s) S (s)k∞ ≤ 1. 150

Phase (deg) Magnitude (dB)

Figure 9 implies that the relation Eq. (38) is fully 100

met. Similarly, it is also possible to define the ampli-

tude frequency characteristics for the inverted transfer 50

function 1/Wu (s) and the product of the controller

transfer and the sensitivity function K (s) S (s): 0
180

∀ ω ∈ R : |K (jω) S (jω)| ≤ |1/Wu (jω)| ⇔ 90

(39)
⇔ kWu (s) K (s) S (s)k∞ ≤ 1. 0

-90
150
KS
-180 -2 0 4
-4 2
1/Wu 10 10 10 10 10
Frequency(rad/s)
Phase (deg) Magnitude (dB)

S
100 1/We
Fig. 11: Amplitude frequency characteristic and phase fre-
quency characteristic of the H∞ controller for mathe-
50 matical model in elevation.

According to Fig. 6, a model of the H∞ controller

0
and the extended system was created in Simulink. The
configuration of the group of external input signals is
-50 as follows:

-100
• reference:
-4 -2 0 2 4
10 10 10 10 10
Frequency(rad/s) d1 (t = 0) = r (t = 0) = −0.25 [−]
for t ∈ (0; 20i , (44)
Fig. 10: Amplitude frequency characteristics of the sensi-
tivity function S (s), product of transfer functions
K (s) S (s), inverted transfer functions 1/We (s) and
1/Wu (s): mathematical model in elevation.
d1 (t) = r (t = 20) = −0.05 [−]
for t ∈ (20; 100i , (45)
Figure 10 implies that the relation Eq. (39) is fully
met. For completeness, we give the values of the key • LF disturbance: not included in the model,
H∞ norms, i.e.:
• HF disturbance: band-limited white noise with
• optimal H∞ norm: power of 0.00001 [W] .

γ = 5.752010−4 , (40) The response of the modeled elevation system to ref-

erence d~1 (t) is as shown in Fig. 12. Thanks to the
• closed loop H∞ norm: balanced ratio of the sensitivity function and the com-
plementary sensitivity function, the elimination of the
kF {P (s) , K (s)}k∞ = 4.783710−4 < γ, (41)
noise and disturbance will be effective.
• sensitivity function H∞ norm:
kS (s)k∞ = 1.5357, (42) 3.4. Elevation Controller for Real
Model
• complementary sensitivity function H∞ norm:
We will use the transfer function from the mathemat-
kT (s)k∞ = 0.9997. (43) ical model for the design of the controller for a real

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 496
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

-20

-30

Phase (deg) Magnitude (dB)

-40

-50

-60

-70

-80
0

-45

-90
-3 -2 -1 0 1 2 3
10 10 10 10 10 10 10
Fig. 12: Dependence of the output elevation angle yψ on time
t in the mathematical elevation model; yψ = f (t). Frequency (rad/s)
Fig. 13: Amplitude and phase frequency characteristics of low-
pass filter We (s), real model in elevation.
model as the controller is created on the basis of the
transfer functions of the nominal system and weighting
filters. The forms of the weighting filters for the given 0

-10
system are as follows, i.e.:
Phase (deg) Magnitude (dB)

-20

-30
• weighting transfer function for reference signal:
-40

1 -50
Wcmd (s) = , (46) -60
0.25s + 1
-70
90
• weighting transfer function for low-frequency dis-
turbance signal:
45
0.5
Wd (s) = , (47)
0.1s + 1
0
• weighting transfer function for high-frequency dis- 10
-2
10
-1 0
10
1
10
2
10
3
10 10
4
10
5

turbance signal (noise): Frequency (rad/s)

0.01s + 1 Fig. 14: Amplitude and phase frequency characteristics of high-
Wnoise (s) = , (48) pass weighting filter 1/W u (s), real model in elevation.
s+1

• weighting transfer function for control controller: 100

S
0.1s + 0.75 T
We (s) = 0.003 , (49) 0
1/We
2.25s + 0.065 1/Wu
Singular Values (dB)

• weight transfer for manipulated value: -100

0.75s + 1 0.75s + 1
Wu (s) = = . (50) -200
s + 2500 s + 4·101 −4
-300
When compared to the mathematical elevation
model, the same penalization of the group of external
input signals is used but the weighting filters for the -400

penalization of error outputs were adjusted to the real

model. However, their character remained unchanged. -500 -2 0 2 4
10 10 10 10
The graphic dependences of the sensitivity functions Frequency(rad/s)
and inverted transfer functions for the real elevation Fig. 15: Amplitude frequency characteristics of the sensitiv-
model are as shown in Fig. 13 and Fig. 14. ity function S (s), complementary sensitivity func-
tion T (s), inverted transfer functions 1/We (s) and
Figure 15 clearly implies that the amplitude fre- 1/Wu (s), real Mmodel in elevation.
quency characteristic of the sensitivity function S (s)

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 497
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

and the inverted transfer function 1/We (s) meet the 100
condition stated in the Eq. (38). KS
1/Wu
S
Figure 16 clearly shows that the amplitude frequency 50

Singular Values (dB)

1/We
characteristic of the product of Laplace transforms
K (s) S (s) and the inverted transfer function 1/Wu (s)
0
meet the condition stated in the Eq. (39). For com-
pleteness, we give the values of the key H∞ norms,
i.e.: -50

• optimal H∞ norm:
-100

γ = 0.0512, (51)
-150 -2 0 2 4
• closed loop H∞ norm: 10 10 10 10
Frequency(rad/s)
kF {P (s) , K (s)}k∞ = 0.0487 < γ, (52) Fig. 16: Amplitude frequency characteristic of the sensitiv-
ity function S (s), product of transfer functions
• matrix H∞ norm: K (s) S (s), inverted transfer functions 1/We (s) and
1/Wu (s), real model in elevation.

We (s) S (s)

Wu (s) K (s) S (s)

= 0.0344 < γ, (53)
∞ 50
Magnitude (dB)

• sensitivity function H∞ norm: 0

-50
kS (s)k∞ = 1.0028, (54)
-100

• complementary sens. fun. H∞ norm:

-150
0
kT (s)k∞ = 0.0073, (55)
Phase (deg)

-45

• control sensitivity H∞ norm: -90

kR (s)k∞ = 6.6069, (56) -135

-180 -2 0 2 4
• open loop H∞ norm: 10 10 10 10
Frequency(rad/s)
kL (s)k∞ = 0.0074, (57) Fig. 17: Amplitude frequency characteristic and phase fre-
quency characteristic of the H-∞ controller for real
• controller H∞ norm: model in elevation.

kK (s)k∞ = 6.6557. (58)

The numerical interpretation of H∞ norms shows

that it is a suboptimal solution. The controller is in
the ninth order at the most, just like in the mathemat-
ical model. Its frequency characteristics are shown in
Fig. 17.
The block diagram in Simulink is formally the same
as in the mathematical model in elevation; the only dif-
ference is thus the nominal system represented by the
real model of a helicopter. The configuration of the in-
puts also does not differ from the mathematical model
as we want to compare the responses. The response
of the modelled elevation system to reference d1 (t) in
Fig. 18. Figure 18 implies that the elimination of noise
Fig. 18: Dependence of the output elevation angle yψ on time
will not be sufficiently effective due to the prevailing t in real elevation model; yψ = f (t).
sensitivity function.

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 498
CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

4. Conclusion [5] LIN, F. Robust control design: an optimal control

approach. Hoboken, NJ: John Wiley, 2007. ISBN
The paper focused on the problem of design of mixed- 04-700-3191-3.
sensitivity H-∞ robust controller, the simulation of [6] CE 150 helicopter model: educational man-
control circuit and verification of the results on edu- ual. In: Department of Information engineer-
cational model CE150, representing physical model of ing and mathematics [online]. 2009. Available
helicopter. It gives an analysis and synthesis of the ro- at: http://www.dii.unisi.it/~control/cmr/
bust controller of elevation of helicopter model, both altro/heli_ce150_manual.pdf.
for its mathematical model and for real system.
The results were verified by implementing the algo- [7] BALATE, J. Automaticke rizeni. Prague: BEN-
rithms on two different platforms: Technicka literatura, 2004. ISBN 978-80-7300-1.

[8] ZHOU, K., J. C. DOYLE and K. GLOVER. Ro-

• Matlab&Simulink + Real Time Toolbox +
bust and optimal control. Upper Saddle River:
MF624: Real Time Toolbox together with mea-
Prentice Hall, 1995. ISBN 01-345-6567-3.
suring card MF624 provide a very elegant way of
fast rapid prototyping for real physical educational [9] SEBEK, M. Robustni rizeni systemu s paramet-
models, using the powerful computational environ- rickymi neurcitostmi: Ucebni texty k seminari.
ment of Matlab&Simulink. In: Czech Technical University in Prague [online].
• REX Control System + WinPAC-8000: The REX Prague, 2012. Available at: http://www.crr.
control system is an advanced tool for design and vutbr.cz/system/files/brozura_07_1210.pdf
implementation of complex algorithms for auto- [10] SODERBERG, B. and T. BUSSARAKONS.
matic control. The algorithms are composed from Compatibility Analysis of Space Qual-
individual function blocks, which are available in ified Intermediate Bus Converter and
extensive function block libraries [13]. It supports Point of Load Regulators for Digital
a wide variety of hardware platforms, including Loads. In: IRF [online]. 2014. Available
programmable automation controllers (PAC) by at: http://www.irf.com/technical-info/
ICPDAS, particularly WinPAC-8000 which was whitepaper/sbbzacompatibility.pdf.
used for implementation of design control algo-
rithms. [11] IOANNOU, P and J. SUN. Robust adaptive con-
trol. New Jersey: Prentice-Hall, 1996. ISBN 04-
864-9817-4.
Acknowledgment
[12] OGATA, K. Modern control engineering. Boston:
Prentice Hall, 2010. ISBN 978-0-13-615673-4.
This work was supported by project SP2014/156, “Mi-
croprocessor based systems for control and measure- [13] REX CONTROLS. Control System for Advanced
ment applications” of Student Grant System, VSB– Process and Machine Control. In: REX Control
Technical University of Ostrava. [online]. London, 2009. Available at: http://www.
rexcontrols.com.

[14] SENAME, O. Robust control analysis and de-

References sign. In: Gipsa-LAB [online]. 2008. Avail-
able at: http://www.gipsa-lab.grenoble-inp.
[1] TEWARI, A. Modern control design: with MAT- fr/~o.sename/docs/schedule.pdf.
LAB and SIMULINK. Chichester: John Wiley,
2005. ISBN 04-714-9679-0. [15] OZANA, S. Navrhovani a realizace regulatoru.
Ostrava: VSB–Technical University of Ostrava,
[2] SULC, B. Teorie a praxe navrhu regulacnich ob-
2012. ISBN 978-80-248-2605-9.
vodu. Prague: Vydavatelstvi CVUT, 2004. ISBN
80-010-3007-5.
[3] TSUI, C.-C. Robust control system design: ad- About Authors
vanced state space techniques. New York: Marcel
Dekker, 2004. ISBN 08-247-4869-7.
Stepan OZANA was born in Bilovec, Czech Re-
[4] DU, D.-W., P. H. PETKO and M. M. KON- public, 1977. He studied electrical engineering at
STANTINOV. Robust control design with MAT- VSB–Technical university of Ostrava where he has
LAB. London: Springer, 2013. ISBN 978-144- got Masters degree in Control and Measurement En-
7146-827. gineering and Ph.D. degree in Technical Cybernetics.

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 499
 CONTROL ENGINEERING VOLUME: 12 | NUMBER: 5 | 2014 | DECEMBER

He works as an Assistant Professor at the Department Bachelor and Masters degree in Control and Measure-
of Cybernetics and Biomedical Engineering, Faculty ment Engineering and also Ph.D. degree in Technical
of Electrical Engineering and Computer Science. At Cybernetics. He works as an Assistant Professor on
present he gives lectures on Cybernetics and Control Department of Cybernetics and Biomedical Engineer-
systems. ing, Faculty of Electrical Engineering and Computer
Science. At present he runs courses of Signals and
Petr VOJCINAK was born in Ostrava, Czech Systems and Cybernetics at the department.
Republic, 1984. He studied electrical engineering at
VSB–Technical University where he has got Bachelor Radovan HAJOVSKY was born in Bilovec,
and Masters degree in Control and Measurement Czech Republic, 1974. He studied electrical engineer-
Engineering. Now he is a Ph.D. student in Tech- ing at VSB–Technical university of Ostrava where he
nical Cybernetics at Department of Cybernetics has got Masters degree in Control and Measurement
and Biomedical Engineering, Faculty of Electrical Engineering and Ph.D. degree in Technical Cybernet-
Engineering and Computer Science. At present he ics. He works as an Assistant Professor on Department
runs practical tutorials of Electrical Measurement at of Cybernetics and Biomedical Engineering, Faculty
the department. of Electrical Engineering and Computer Science. At
present he gives lectures on Measurement systems,
Martin PIES was born in Novy Jicin, Czech Electromagnetic compatibility, Electronic equipment
Republic, 1983. He studied electrical engineer- and Measurement.
ing at VSB–Technical University where he has got

c 2014 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 500

View publication stats