Sei sulla pagina 1di 10

EPE-PEMC 2002 Dubrovnik & Cavtat P.

1
Analysis and Optimization of an Electronic Throttle for
Linear Operating Modes
Joko Deur
*
, Danijel Pavkovi
*
, Nedjeljko Peri
**
, and Martin Jansz
***

*
Faculty of Mechanical Engineering and Naval Architecture, University of Zagreb, I. Luia 5,
HR-10000 Zagreb, Croatia; Tel. +385-1-6168-372, Fax. +385-1-6168-351, josko.deur@fsb.hr;
Tel. +385-1-6168-325, Fax. +385-1-6168-351, danijel.pavkovic@fsb.hr

**
Faculty of Electrical Engineering and Computing, University of Zagreb, Unska 3,
HR-10000 Zagreb, Croatia; Tel. +385-1-6129-855, Fax. +385-1-6129-809, nedjeljko.peric@fer.hr

***
Ford Motor Company Ltd., Product Development Europe, Dunton Technical Centre, Laindon,
Basildon, Essex SS15 6EE, UK; Tel. +44-1268-40-4821, Fax. +44-1268-40-4796, mjansz@ford.com

Keywords
Automotive components, automotive applications, control, DC machines, motion control, servo drives.
Abstract
Electronic throttles are increasingly being used in automotive systems in order to improve vehicle
driveability, fuel economy, and emissions. The paper presents an analysis of the electronic throttle
body linear model. The analysis results in simplifications of the process model structure and controller
optimization procedure. A method of identification of the simplified process model is outlined. A
linear feedback/feedforward throttle position controller is algebraically optimized according to the
damping optimum. The linear controller is extended with a gain scheduling algorithm, in order to deal
with different process parameters for the regions below and above the limp-home position. The
proposed controller represents a core of the overall nonlinear electronic throttle control strategy which
also includes friction and limp-home compensators. The designed control system is examined by
computer simulation and experiment.
Introduction
Electronic throttle is a DC servo drive which needs to provide precise positioning of the engine throttle
plate. The advantages of the electronic throttle compared to the traditional "mechanical throttle"
include: (i) possibility of implementation of engine-based vehicle dynamics systems including traction
control, and (ii) improvement of vehicle emissions, fuel economy, and driveability.

In comparison with a traditional DC or AC servo drive (see e.g. [1]), the electronic throttle includes
some specific design characteristics such as the absence of the inner current control loop, the use of a
potentiometer as a position sensor, and the presence of a return spring which constrains the throttle
motion. Therefore, the standard servo-controller cascade structure and related design methods [1]
cannot be directly applied to the electronic throttle.

An electronic throttle control strategy can generally be divided in two modules: (i) a linear feedback
controller extended with a linear feedforward compensator, and (ii) compensators of emphasized
nonlinear effects such as friction or limp-home nonlinearity. Design of linear feedback/feedforward
controller is considered in this paper, as a part of more comprehensive research activities on
development of the overall nonlinear control strategy [2,3]. A PID throttle position controller extended
with a linear feedforward term is utilized rather than some more advanced controllers (e.g. those
presented in [4,5]), because it is widely used and well-understood in the automotive practice, and is
well-fitted to the relatively simple electronic throttle body (process) model. PID controller gain
scheduling is applied, in order to adapt the controller to the change of return spring stiffness while the
throttle passes through the limp-home position.
EPE-PEMC 2002 Dubrovnik & Cavtat P. 2
An analysis of the linear process model is carried out, with the aim to simplify the process model,
process identification procedure, and controller design. An algebraic method of controller design
according to the damping optimum [6] is proposed. The behavior of the designed control system is
examined by computer simulation and experiment.
Process model
A schematic of the electronic throttle control system is shown in Fig. 1. The system consists of the
electronic throttle body, a MOSFET chopper, and a microcontroller system. The electronic throttle
body includes a throttle valve which controls the air mass flow into the engine manifold. The throttle
valve is driven by a DC drive and constrained by a dual return spring. The return spring returns the
throttle into its initial position (so-called limp-home position) in the case of power supply failure, thus
enabling the driver to limp the vehicle home. The throttle position is measured by a potentiomenter.

The block diagram of the electronic throttle body+chopper model (the process model) is shown in Fig.
2. The process model consists of the well-known linear model of DC motor [1], and the nonlinear
friction and return spring models. The process is in the linear operating mode, if the following two
conditions are satisfied:
(i)
C f m
M m t >

(no operation in the zero-speed (stiction) region),

(ii)
l s s LH
K K m > (no operation in the limp-home region).

The process is close to the linear operating mode even if the condition (i) is not satisfied (e.g. for the
pointing control task), provided that the changes of armature voltage u
a
are large enough to efficiently
overcome stiction (the large signal operating mode).

Fig. 3a shows the linear portion of the process model. The relatively low load torque m
L
due to air
mass flow is omitted in Fig. 3a. The gain K
s
represents the return spring stiffness coefficient outside
the limp-home region. It is usually much larger in the region below the limp-home position ( <
LH
)
than in the region above the limp-home position [3].
Process analysis
The DC motor armature dynamics is very fast; i.e. the armature time constant T
a
in Fig. 3a is very
small (typically less than 1 ms [3]). Since such a small value of T
a
is much smaller than the equivalent
time constant of the closed-loop system, the armature dynamics may be neglected and represented by
the pure armature gain K
a
(Fig. 3b). If the back electromotive force u
emf
is referred to the second left-
hand summation point in Fig. 3a, it represents a damping path with the equivalent damping coefficient
(Fig. 3b)


v t a d
K K K K . (1)

Since the damping coefficient K
d
is much larger than the viscous friction damping b (K
d
>> b) [10], the
coefficient b may also be neglected (cf. Fig. 3b and 3a).

M
gearbox
potentiometer
ELECTRONIC THROTTLE BODY
throttle
valve
u
a
i
a
CHOPPER
CONTROLLER
C
R

return
spring



Fig. 1. Functional scheme of electronic throttle.
EPE-PEMC 2002 Dubrovnik & Cavtat P. 3
s T
K
a
a
+ 1 b Js +
1
s
1
l
K
l
K
+
+
v
K
+
-
m

m

m
m
s
m
a
u
emf
u
t
K
a
i
LH

f
m
-
+
-
L
m
ch
K
u
M
S M
C

s


u, u
a
, u
emf
- commanded signal, armature voltage, and back electromotive force; i
a
- armature current;
m
m
, m
s
,m
f
, m
L
- motor, spring, friction, and load torque;
m
, - motor and throttle position;
m
- motor speed;
K
ch
- chopper gain; K
a
, K
t
, K
v
- armature, torque, and voltage gain; J - total moment of inertia referred to motor shaft;
1/K
l
- gear ratio;
LH
- limp-home position; M
C
, M
S
- Coulomb and breakaway friction.

Fig. 2. Block diagram of process model.



s T
K
a
a
+ 1
b Js +
1
s
1
l
K
l
K
s
K
+
-
v
K
+
-
m

m

m
m
s
m
a
u
t
K
a
i
emf
u
ch
K
u
a

Js
1
s
1
l
K
l
K
s
K
+
-
t v a
K K K
m

m

m
m
a
u
t
K
a
K
-
s
m
a
i
ch
K
u
b

Js
1
s
1
l
K
+
t v a
K K K
m

m

m
m
a
u
t
K
a
K
-
a
i
ch
K
u
c

Fig. 3. Illustration of simplification of linear process model.
EPE-PEMC 2002 Dubrovnik & Cavtat P. 4
The transfer function of the process model in Fig. 3b is given by


2 2
) (
) (
) (
l s d
l t a ch
p
K K s K Js
K K K K
s u
s
s G
+ +


. (2)

If the transfer function (2) has conjugate-complex poles, the process response can be characterized by
undesired weakly-damped oscillations caused by the return spring compliance. However, if the
electromotive force-related damping K
d
is large enough, the poles of transfer function (2) assume real
values, and the throttle response oscillations are completely damped. The condition for the aperiodic
response is, thus, given by

J K K K
s l d
2 . (3)

In respect to condition (3), the following relation has been found to be valid for the particular
electronic throttle body [7,3]:

( )

'

< >
> >>

LH
LH
s l d
J K K K


for , 1 4 . 1
for , 1 6 . 5
2 / . (4)

Hence, the process has indeed aperiodic behavior.

If the condition (3) is largely satisfied, i.e. if

J K K K
s l d
2 >> (5)

(as it is the case for >
LH
, Eq. (4)), the damping path in Fig. 3b is so stiff that the influence of return
spring compliance to the process dynamic behavior may be neglected. In other words, the return
spring feedback path m
s
() in Fig. 3b may be disconnected, which leads to the simplified process
block diagram shown in Fig. 3c. The simplified process has integral+lag behavior described by the
transfer function

s s T
K
s u
s
s G
em
p
p
) 1 ( ) (
) (
) (
+


, (6)

where the gain K
p
and the electromechanical time constant T
em
are given by


v
l ch
p
K
K K
K , (7)

d
em
K
J
T . (8)
Process identification
A detailed experimental identification of the linear and nonlinear process dynamics is presented in [3].
Identification of linear dynamics includes a multi-step identification method based on the physical
model form, and single-step methods based on the continuous-time and discrete-time input-output
model forms. Identification of continuous-time integral+lag type model form (6) is outlined here.

Fig. 4 (dashed lines) shows the experimental response of the throttle speed with respect to step change
U of the commanded signal u, for an operating point above the limp-home position ( >
LH
). The
speed signal is reconstructed by time-differentiation of the measured position signal. According to the
model (6), the speed response is described by the first-order lag transfer function G
p
(s) = (s) / u(s) =
sG
p
(s) = K
p
/ (T
em
s + 1). The process gain is estimated as K
p
=
ss
/ U, where
ss
is the response steady-
state value (Fig. 4). Depending on whether the steady-state speed measurement noise is frequent or
sporadic (Fig. 4), the steady-state value
ss
is determined as a mean or median value of the steady-
state response, respectively. The electromechanical time constant T
em
is estimated numerically to a
value which minimizes the least-squares criterion applied to the measured and predicted speed
responses ([3]; Fig. 4).
EPE-PEMC 2002 Dubrovnik & Cavtat P. 5
0
2
4
0
] / [ s rad

ss
1
3
1 2 3 4 5 6 7
0
3
6
] / [ s rad

ss
1.5
4.5
measurement
prediction
0 1 2 3 4 5 6 7
a b
Normalized time Normalized time

Fig. 4. Measured and predicted speed step responses for cases of frequent (a) and sporadic noise (b).

Controller design
The controller is designed with respect to linear operating modes of the process. It is extended with
friction and limp-home compensators in [2,7], in order to provide desired (linear-like) behavior of the
control system for nonlinear operating modes as well.
Controller structure
The structure of the proposed electronic throttle controller is shown in Fig. 5. The controller consists
of a PID feedback controller, a feedforward controller (FFC), and a gain-scheduling algorithm. The
PID controller is given in a modified form [8,9], where the proportional (P) and derivative (D) terms
act to the measured throttle position signal only. The modified PID controller structure (applied to
the integral+lag type process model (6)) can provide optimal closed-loop system behavior with respect
to both reference and disturbance (e.g. friction or load torque), without the presence of undesired
overshoot of the step response with respect to reference
R
[8,9]. The gain-scheduling algorithm adapts
the controller parameters to the change of return spring stiffness coefficient while the throttle passes
through the limp-home position. The feedforward controller introduces a zero in the closed-loop
transfer function, thus resulting in decrease of the response time with respect to reference.


1 z
z
T
T
K
I
R
z
z
T
T
K
D
R
1
R
K
P
D
I
+
+
I
u
+
-
max
U
D
u
P
u
FFC
R

+
-

u
PID controller
LH

+
-
LH

z
z 1
reset
LH
D I R
T T K
>
] [
LH
D I R
T T K
<
] [
F
ff
ff
F
z z
z z
z
z

1
1
integrator
Gain scheduling

Fig. 5. Block diagram of controller.
EPE-PEMC 2002 Dubrovnik & Cavtat P. 6
Optimization of PID controller
The PID controller parameters are determined according to the damping optimum [6,9]. The control
system is optimized in the continuous-time domain. For the purpose of quasi-continuous optimization,
the discrete-time PID controller in Fig. 5 is replaced with its continuous-time counterpart. Further, the
sampler and the zero-order-hold element, as well as the time-differentiator used in the controller
derivative term, are approximated with the first-order lag term with the time constant T / 2 (T -
sampling time) [9]. These parasitic delays, together with the motor armature delay, are approximately
described by an equivalent first-order unity-gain process lag term with the time constant

T T T
a
+

. (9)

Adding the parasitic lag term to the transfer function (2) yields the overall process transfer function


) 1 )( 1 (
) (
) (
) (
1
2
2
2
+ + +

s a s a s T
K
s u
s
s G
p p
p
p

, (10)

with

l s
t a ch
p
K K
K K K
K
2
, (11)


2
1
l s
d
p
K K
K
a , (12)


2
2
l s
p
K K
J
a . (13)

The corresponding transfer function of the equivalent continuous-time system is

1
1 ) ( ) (
1
) (
) (
) (
2
2 2
2
1 2 3
2
1 2 4
2
2
+
+
+
+ +
+
+
+


s T
K K
K K
s
K K
T T a T K K
s
K K
T T a a
s
K K
T T a
s
s
s G
I
p R
p R
p R
I p D p R
p R
I p p
p R
I p
R
c

(14)

The fourth-order characteristic polynomial of the damping optimum is defined as [6,9]:

1 ) (
2 2
2
3 3
3
2
2
4 4
4
2
3
3
2
+ + + + s T s T D s T D D s T D D D s N
e e e e
. (15)

Equating the coefficients of the characteristic polynomial of transfer function (14) with the
corresponding coefficients of characteristic polynomial (15), and rearranging, yields the following
equations for the closed-loop equivalent time constant T
e
, and the controller parameters K
R
, T
I
, and T
D
:


em
e
T T
T
D D D
T
/ 1
1
4 3 2

+
, (16)

,
_


1
1
2
3
2
2
1
2
e
em
p
p
R
T D D
T T
a
K
K , (17)


e
p R
p R
I
T
K K
K K
T
2
2
1+
, (18)


1
1
]
1

,
_

T
T D D
T T
a
K K
T
e
em
p
p R
D
1
1
3 2
1
2
. (19)

EPE-PEMC 2002 Dubrovnik & Cavtat P. 7
If all the characteristic ratios D
2
, D
3
, and D
4
are set to the optimal value of 0.5, the closed-loop system
will have a fast quasi-aperiodic reference step response with overshoot of approx. 6% and risetime
t
1
1.8T
e
[6,9]. The usual requirement on the electronic throttle control system is that the reference
step response has an aperiodic form (i.e. no overshoot allowed). The fastest (boundary) aperiodic
response is obtained by decreasing the dominant damping ratio D
2
to the value D
2
0.37. Eq. (16)
gives the minimum equivalent time constant T
e
for well-damped response. The time constant T
e
can be
arbitrarily increased above this minimum value, e.g. in order to decrease noise in the commanded
signal u.

The above optimization procedure can be repeated for the simplified integral+lag process model (6)
extended with the parasitic process lag term with the time constant T

. The parasitic time constant T


is much less than the electromechanical time constant T
em
[10]. Therefore, the time constant T

may be
lumped to the time constant T
em
, thus preserving the simple integral+lag form of the process model:


[ ]s s T T
K
s u
s
s G
em
p
p
1 ) ( ) (
) (
) (
+ +

. (20)

Applying the optimization procedure results in the following simple expressions for the controller
parameters [10]:


2
3
2
2
*
1
e
em
p
R
T D D
T T
K
K

+
, (21)


e I
T T
*
, (22)

,
_

T T
T D D
T D T
em
e
e D
3 2
2
*
1 , (23)

where the equivalent time constant T
e
is chosen to a value greater or equal to that given by Eq. (16).

The use of Eqs. (21)-(23) should be preferred compared to Eqs. (17)-(19), because they have simpler
forms and are based on the simple integral+lag process model which is simply identified. These
advantages are particularly emphasized for auto-tuning electronic throttle applications. However, Eqs.
(21)-(23) are somewhat inaccurate for throttle operations below the limp-home position ( <
LH
),
because they have been derived based on the assumption of negligible return spring influence. Thus, it
would be convenient to correct Eqs. (21)-(23) so that more accurate results according to Eqs. (17)-(19)
are obtained. The final relations between the two sets of equations can be expressed as:


R R R
K K K
*
, (24)

) 1 (
*
a T T
I I
, (25)


a
aT T
T
D
D


1
*
, (26)

where the coefficients K
R
and a are given by:



u
K K
K K
K
d p
l s
R
2
, (27)



u
K
T T
T D D
T T K
K K T D D
a
p
em
e
em d
l s e
2
3
2
2
2 2
3
2
2
) (
. (28)

The coefficient u / = K
s
K
l
/ (K
ch
K
a
K
t
) in Eqs. (27) and (28) represents the slope of the process static
curve u() [3], which is different for the regions below and above the limp-home position.
EPE-PEMC 2002 Dubrovnik & Cavtat P. 8
Hence, the accurate controller parameters for any operating point can be conveniently calculated by
using the simple equations (21)-(23) and the correction formulae (24)-(26), based on the process
parameters obtained by simple experimental identifications of integral+lag process dynamics and
process static curve.
Optimization of feedforward controller
The feedforward controller introduces a zero (z
ff
in Fig. 5) in the closed-loop transfer function. This
zero should be located near the closed-loop poles, in order to provide compensation of the reference
response delay caused by the poles. The zero can be optimized according to the general design method
based on the magnitude optimum [11]. This method provides optimal quasi-aperiodic reference step
response of the overall closed-loop system, with an overshoot of approximately 5%. The method is
modified here in order to meet the requirement for the aperiodic reference step response.

The non-dominant characteristic ratio D
3
is decreased to 0.4 in order to additionally damp the closed-
loop system (note that the dominant ratio D
2
has been decreased to 0.37). The feedforward controller
zero is then set to the value


e
T T
ff
e z
/ 04 . 2
. (29)

This value corresponds to the dominant closed-loop pole, i.e. the zero-pole canceling approach is
applied. The filter pole z
F
(Fig. 5) is set to zero in this paper (the dead-beat feedforward controller).
Gain-scheduling algorithm
Analysis presented in the previous section has implied that the return spring feedback can be neglected
for operations above the limp-home position
LH
. The same is not valid for operations below the limp-
home position, because the spring is much stiffer in this region. It has been shown in [7] that the
particular electronic throttle has 25% larger response time in the region <
LH
, if it is optimized for
the region >
LH
. In order to provide optimal throttle behavior for the both operating regions, it is
convenient to change the controller parameters while the throttle passes through the limp-home
position. This is achieved by the gain scheduling algorithm illustrated in Fig. 5.

There are two sets of optimal controller parameters, which correspond to the different process
parameters in the regions <
LH
and >
LH
. The actual controller parameters are set to one of
predetermined parameter sets, based on the simple relay logic shown in Fig. 5. The relay function
includes a hysteresis, in order to avoid chattering of the actual parameters and the commanded signal.

When the actual controller parameters are switched from one to the other parameter set, the
commanded signal u changes abruptly. The proportional term u
P
= K
R
dominantly contributes to the
abrupt change of the commanded signal. In order to provide a smooth transition through the limp-
home zone, it is proposed to apply a simple integrator-reset intervention [7]

[ ] ) 1 ( ) 1 ( ) ( ) 1 ( ) 1 ( + k k K k K k u k u
R R I
k
I
, (30)

where the superscript k on the left-hand side of Eq. (30) denotes that the integrator value calculated in
the previous ((k-1)-th) sampling interval is reset in the k-th interval, before it is used to update the
controller output in the k-th interval.
Simulation and experimental results
Fig. 6 shows the experimental step responses of the electronic throttle control system for different
operating points and different step changes of the position reference. The process practically operates
in the linear mode due to relatively large reference steps
R
10
o
(no significant friction influence to
the system transients) and >
LH
(no limp-home effect). Similar aperiodic responses are observed for
all the operating conditions. The settling time t
95%
is approx. 115 ms for
R
= 10
o
. It is somewhat
larger for
R
> 50
o
due to the controller saturation effect. Simulation results in Fig. 7 indicate similar
settling time of the response with respect to the disturbance (the load step at t = 0.25 s).

EPE-PEMC 2002 Dubrovnik & Cavtat P. 9
Comparative simulation responses shown in Fig. 7 illustrate the possibility of improvement of the
control system performance with respect to both reference and disturbance by decreasing the sampling
time T. Decrease of the sampling time implies decrease of the control system equivalent time constant
T
e
(see Eqs. (16) and (9)), which results in decrease of the control system response time. However, this
possibility of improving the control system performance is limited due to the influence of noise in the
measured position signal, which is transferred to the commanded signal u (see experimental results in
Fig. 8) and causes transmission (and potentiometer) wear and "loud" operation. The minimum
equivalent time constant T
e
for "quiet" drive operation has been found to be approx. 35 ms, which
corresponds to the settling time of approx. 80 ms (Figs. 7 and 8). Additional decrease of the settling
time (but only with respect to the reference) to the value of 50 ms is possible by including the
feedforward controller (Figs. 7 and 8). In the large signal operating mode (reference step changes

R
> 10
o
), the feedforward controller could not speed-up the reference response due to the controller
saturation.
Conclusion
Analysis of the electronic throttle body linear model has shown that this model may be approximated
by a simple integral+lag term for operations above the limp-home position. This simplification
facilitates process identification and control system design. The PID electronic throttle position
controller has been optimized according to the damping optimum, thus resulting in simple algebraic
equations for the controller parameters. Simple algebraic corrections of these equations have been
proposed, in order to obtain optimal control system behavior for operations below the limp-home
position as well. A gain-scheduling algorithm changes the controller parameters while the throttle
passes through the limp-home position. The presented controller optimization procedure together with
the related identification technique is suitable for the application in auto-tuning electronic throttle
strategies.

The behavior of the proposed electronic throttle control system has been verified by computer
simulation and experiment. It has been shown that the decrease of sampling time results in faster
control system response (higher bandwidth) with respect to both disturbance and reference. However,
this possibility of the control system performance improvement is limited by the influence of noise.
The response time with respect to reference can additionally be decreased in the small signal operating
mode by introducing a simple, lead-lag feedforward controller.

The optimized controller represents a core of an overall nonlinear electronic throttle control strategy
[2], which also includes compensation of friction and limp-home nonlinear effects.
Acknowledgment
The support from Ford Motor Company, Dearborn, MI, USA, and the Ministry of Science and
Technology of the Republic of Croatia is gratefully acknowledged.


0 0.5 1 1.5 2 2.5 3 3.5 4
30
40
50
60
70
] [ ,
R

t [s]
R

a
0 0.5 1 1.5 2 2.5 3 3.5 4
30
40
50
60
70
80
t [s]
] [ ,
R

b
Fig. 6. Experimental step responses of control system (T = 5 ms; no feedforward controller (FFC)).
EPE-PEMC 2002 Dubrovnik & Cavtat P. 10




0
1
2
3
4
5
6
0
0.5
1.0
1.5
2.0
0 0.1 0.2 0.3 0.4
0 0.1 0.2 0.3 0.4
t [s]
t [s]
] [
] V [ u
T = 5ms, w/o FFC
T = 2ms, w/o FFC
T = 2ms, w/ FFC
5
R

L
m

0 0.05 0.1 0.15 0.2 0.25
30
32
34
36
] [
T = 5ms, w/o FFC
T = 2ms, w/o FFC
T = 2ms, w/ FFC
t [s]
R

0
1
2
3
4
0 0.05 0.1 0.15 0.2 0.25
t [s]
T = 5ms, w/o FFC
] V [ u
0
1
2
3
4
0 0.05 0.1 0.15 0.2 0.25
t [s]
T = 2ms, w/o FFC
] V [ u
0
1
2
3
4
0 0.05 0.1 0.15 0.2 0.25
] V [ u
t [s]
T = 2ms, w/ FFC

Fig. 7. Comparative simulation responses for
different controller structures and parameters.

Fig. 8. Comparative experimental responses for
different controller structures and parameters.


References
[1] Leonhard, W.: Control of Electrical Drives, Springer Verlag, Berlin, 1985.
[2] Deur, J., Pavkovi, D.: Design and Experimental Verification of an Electronic Throttle Control Strategy,
Internal memorandum 12/15/01, University of Zagreb, Croatia, 2001.
[3] Pavkovi, D., Deur, J.: Experimental Identification of an Electronic Throttle Body, Internal memorandum
10/27/01, University of Zagreb, Croatia, 2001.
[4] Song, J.-B., Byun, K.-S.: Throttle Actuator Control System for Vehicle Traction Control, Mechatronics,
Vol. 9, pp. 477-495, 1999.
[5] Yokoyama, M., Shimizu, K., Okamato, N.: Application of Sliding-Mode Servo Controllers to Electronic
Throttle Control, Proc. of the 37th IEEE Conference on Decision and Control, pp. 1541-1545, 1998.
[6] Naslin, P.: Essentials of Optimal Control, Chap. 2, Iliffe Books Ltd, London, 1968.
[7] Deur, J., Pavkovi, D.: Design and Experimental Verification of an Electronic Throttle Control Strategy,
Internal memorandum 12/15/01, University of Zagreb, Croatia, 2001.
[8] Astrm, K. J., Wittenmark, B.: Computer Controlled System, Prentice-Hall, London, 1984.
[9] Deur, J.: Design of linear servosystems using practical optima, Internal memorandum 04/19/2001
(translation of Chap. 3 of Ph. D. Thesis by J. Deur), University of Zagreb, Croatia, 2001.
[10] Deur, J., Pavkovi, D.: Analysis and Optimization of Electronic Throttle Control System, Internal
memorandum 03/20/01, University of Zagreb, Croatia, 2001.
[11] Deur, J., Peri, N., Staji, D.: Design of Reduced-Order Feedforward Controller, UKACC International
Conference on CONTROL '98, pp. 207-212, Swansea, UK, 1998.

Potrebbero piacerti anche