Sei sulla pagina 1di 6

Robustness of Discrete-Time Inverse Optimal

Control for Trajectory Tracking


Enrique A. Lastire-Olmedo1 , Edgar N. Sanchez1, Alma Y. Alanis2 and Fernando Ornelas-Tellez3
1

CINVESTAV, Unidad Guadalajara, 2 CUCEI, Universidad de Guadalajara, 3 Universidad Michoacana.


1
Av. del Bosque 1145, Zapopan, Jalisco, C.P. 45019. Mexico ,
2
Apartado Postal 51-71, Col. Las Aguilas, Zapopan, Jalisco, C.P. 45080, Mexico,
3
Gral. Francisco J. Mugica S/N, Ciudad Universitaria, Morelia, Michoacan, C.P. 58030, Mexico.
Email: elastire@gdl.cinvestav.mx, sanchez@gdl.cinvestav.mx, almayalanis@gmail.com, fornelast@gmail.com

yk

uk

AbstractIn this paper, we stablish robustness of discrete-time


inverse optimal control for nonlinear systems trajectory tracking;
we analyze in detail disk margin which includes sector and gain
margins. We illustrate the applicability of this result by means
of a linear induction motor example.

yck
I. I NTRODUCTION
There exist robust techniques for discrete-time linear state
feedback control laws [1], which guarantee that the closedloop system possesses pre-specified sector, gain, and phase
margins. There are also nonlinear optimal state feedback
regulators for continous time, which guarantees sector and disk
margins in presecence of nonlinearities in the sector (1/2, ).
In this paper, we state robustness margins of discrete-time nonlinear inverse optimal control for trajectory tracking [2],[3],
which minimizes a nonlinear performance criterion. We illustrate the proposed scheme with a linear induction motor
(LIM). This kind of motor has been employed widely in
industrial applications such as the steel, textile, nuclear and
space industries [4]. However, the most extensive application
for LIMs is for transportation.
The paper outline is as follows: we present mathematical
preliminaries of disk margin, optimal control, inverse optimal
control and block control form in section II; in section III, the
disk margin of discrete-time inverse optimal control for trajectory tracking is formulated, which is our main contribution.
Section IV describes a discrete-time mathematical model of
the linear induction motor. Then, in section V a discrete-time
trajectory tracking inverse optimal controller is implemented
via simulation to the LIM and a disk margin is calculated.
Finally, conclusions about the obtained results are given in
section VI.
II. M ATHEMATICAL PRELIMINARIES
A. Gain, Sector, and Disk Margins of Discrete-Time Regulators
In this section, we present definitions for gain, sector, and
disk margin of discrete-time nonlinear systems controlled by
nonlinear regulators.

()

u ck

Fig. 1. Feedback interconnection of G and operator ()

Consider the nonlinear system G given as


xk+1
yk

=
=

f (xk ) + g(xk )uk , x(0) = x0 , k Z+ ,


(xk ),
(1)

where uk , yk m , f : n n , g : n nm and
: n m .
We define the robustness margins for G in (1), as follows:
Definition 1: [5]. A system G is zero-state observable if for
all x n , uk 0, yk 0 implies xk 0. A system G is
completely reacheable if for all xk1 , xk2 there exists a finite
integer and square summable control uk defined on [k1 , k2 ]
such that the state xk can driven from xk1 to xk2 .
Definition 2: [6]. Let , be such that 0 < 1
< . Then the nonlinear dynamical system is said to have
a gain margin (, ) if the negative feedback interconnection
of (1) and (uck ) = uck is globally asymptotically stable
for all = diag{k1 , . . . , km } with constants ki (, ),
i = 1, . . . , m.
Definition 3: [6]. Let , be such that 0 <
1 < . Then the nonlinear dynamical system (1) is
said to have a sector margin (, ) if the negative feedback
interconnection of (1) and (uck ) = (uck ) is globally
asymptotically stable for all nonlinearities : n m
T
such that (0) = 0, (uc ) = [1 (uc1 ), . . . , m (ucm )] , and
u2ci < i (uci )uci < u2ci , for all uci 6= 0, i = 1, . . . , m.
Definition 4: [6]. Let , be such that 0 <
1 < . Then the nonlinear dynamical system (1) is
said to have a disk margin (, ) if the negative feedback
interconnection of (1) and () is globally asymptotically
stable for all dynamics operators () such that () is zerostate observable and dissipative with respect to the supply rate

r(uck , yck ) = uTck yck

ycTk yck

uTc uck , where

+ k

= + , = , and , such that 0 < 2 < .


It is worth to note that a discrete-time nonlinear system
having a disk margin D(, ) also has gain and sector margins
(, ).

Asummption 1: The full state of system:


xk+1 = f (xk ) + g(xk )uk ,
is measurable.
Definition 5: Let define the control law [3], [11]
V (xk+1 )
1
,
uk = R1 (xk )g T (xk )
2
xk+1

B. Optimal Control
In this section, the discrete-time optimal control is presented. We consider a cost function associated with (1); which
minimize a nonlinear-performance criterion [6], [7]:
J

L(xk , uk ),

(2)

(8)

(9)

to be inverse optimal (globally) stabilizing if:


a) It achieves (global) asymptotic stability of x = 0 for
system (8)
b) V (xk ) is (radially unbounded) positive definite function
such that inequality

k=0

where L(xk , uk ) = l(xk )+uTk R(xk )uk , and l(xk ) is a positive


semidefinite function, to be minimized by the control law uk .
Similar to the continuous-time case, and considering that there
exists a positive definite function V : n , the discretetime Hamiltonian becomes
H(xk , uk )

V := V (xk+1 ) V (xk )
+uTk R(xk )uk

l(xk ) + V (xk+1 ) V (xk )


1 T 1
+ V R (xk )g T (xk )V = 0,
4

(3)
which is used to obtain the control law uk by calculating
(4)

uk

uk

H(xk , uk ),

(5)

A necessary condition, this feedback optimal control law


u(xk ) must satisfy ([8]), is
H(xk , uk ) =

0.

(6)

Therefore, the feedback optimal control law is formulated


as
uk

:=
=

(11)

where

V (xk+1 )
V T (xk+1 )
y V =
.
xk+1
xk+1
As established in definition (5), the inverse optimal control
approach is based on the knowledge of V (xk ). Thus, we
propose a CLF V (xk ), such that (a) and (b) are guaranteed.
That is, instead of solving (11) for V (xk ), we propose a
control Lyapunov function V (xk ) with the form:
V T =

The value of uk which achieves this minimization is a feedback law denoted as uk = u(xk ), then
min H(xk , uk ) =

0,

is satisfied.
When we select l(xk ) := V 0, then V (xk ) is a solution
for the HJB equation

= L(xk , uk ) + V (xk+1 ) V (xk ),

min H(xk , uk ).

(10)

uk
1
V (xk+1 )
g T (xk )
,
2
xk+1

(7)

with the boundary condition V (0) = 0. uk is used to


emphasize that uk is optimal.
C. Inverse optimal control
In this section, the discrete time inverse optimal control
and its solution by using a CLF (Control Lyapunov Function)
is established [3]. For the inverse optimal control approach,
a candidate CLF is used to construct an optimal control
law directly without solving the associated Hamilton-JacobiBellman (HJB) equation [9]. We focus on inverse optimality
because it avoids to solve the HJB partial differential equations
[10].
We establish the following assumptions and definitions,
which allow the inverse optimal control solution via the CLF
approach.

1 T
x P xk ,
(12)
2 k
for control law (9), in order to ensure stability of the equilibrium point xk = 0 of system (8). Moreover, it is established
that control law (9) with (12), which is referred to as the
inverse optimal control law, optimizes a cost functional of the
form (2).
Consequently, by considering V (xk ) as in (12), the control
law takes the following form:
V (xk ) =

(xk ) :=
=

uk
1
(R(xk ) + P2u (xk ))1 P1u (xk ),
2

(13)

where P1u (xk ) = g T (xk )P f (xk ), and P2u (xk ) =


1 T
g (xk )P g(xk ). It is worth to point out that P and R(xk ) are
2
positive definite and symmetric matrices; thus, the existence
of the inverse in (13) is ensured.
D. Nonlinear block control form
The block control method is applied to decompose the
control law synthesis problem into a number of sub-problems
of lower order. The method consists in the conversion of
system (8) to a special state representation, which is referred

as the Nonlinear Block Controllable (NBC) form consisting


of r blocks [12]:
xk+
xk+
xr
k+
xrk+

=
=
..
.
=
=

f (xk ) + B (xk )xk


f (xk , xk ) + B (xk , xk )xk
f r (xk , xk , . . . , xr
) + B r (xk , xk , . . . , xr
)xrk
k
k
f r (xk ) + B r (xk )(xk ),
(14)

III. D ISK

MARGIN OF THE DISCRETE - TIME INVERSE

OPTIMAL CONTROL FOR TRAJECTORY TRACKING

In this section, we extend the results presented in [6] to


trajectory tracking. The new results presented are based on the
discrete-time nonlinear-block control form; therefore stability
margins are obtained for such representation. We assume the
new system representation as in Fig 2:
u zk

yzk
Z



T
rT T
where xk n , xk = xT
, xj nj ;
k , xk , . . . xk ,
j = , . . . , r, nj denotes the order of each rth block;
T

xj = xj , xj , . . . xjn , ; and (xk ) m is the input.
For trajectory tracking, let define the tracking error as
zk

xk x,k ,

(15)

where x,k is the desired trajectory signal.


Taking one step ahead in (15), we have

zk+
= f (xk ) + B (xk )xk x,k+ ,

(16)

Equation (16) is viewed as a block with state zk and the


state xk is considered as a pseudo-control input where desired
dynamics can be imposed as follows:

zk+

=f

(xk )

+B

(xk )xk

x,k+

K zk ,

(17)

where K = diag {k , . . . , kn } is a Schur matrix such that


K zk is a stable dynamic. Then, the desired behavior of xk
is calculated as:

(18)
x,k = [B (xk )] x,k+ f (xk ) + K zk .
Proceeding along the same way as for the first block, a second
variable in the new coordinates is defined as
zk = xk x,k ,

yzck

xrk

xr,k .

First, we demonstrate that system (22) is globally asymptotically stable with an inverse optimal feedback controller.
Theorem 1: Consider the block control nonlinear system Z
of the form:

r
zk+
= f r (xk ) + B r (xk )(xk ) xr,k+ .

By means of this change of variables, system (14) can be


represented as

zk+

zk+

r
zk+

K zk + B (xk )zk

=
..
.
=

K zk

+B

f (zk ) + g(zk )uzk ,

yzk

(zk ),

J(z0 , u()) =

(23)

X


L1 (zk ) + uTzk R(zk )uzk ,
k=0

where L1 : and R : n mm are given such


that L1 (z) > 0 and R(z) > 0, zk n .
Assume there exist functions Vz : n , P1u : n
1m , and P2u : n N m , with Vz () continuous such that
Vz (0) = 0,

(24)

V (zk ) > 0, z , z 6= 0,

(25)

V (f (zk ) + g(zk )u(zk ))


= V (f (zk )) + P1u (zk )uz + uTz P2u (zk )uz ,

(20)

(21)

zk+1
with cost functional

(19)

Taking one step ahead yields

uzck

Fig. 2. Feedback interconnection of block Z and operator ()

and the desired behavior for xk can be calculated. Taking these


steps iteratively, the last new variable is defined as
zkr

()

(26)

are satisfied and




1
1 T
V f (zk )) g(zk ) (R(zk ) + P2u (zk )) P1u (zk )
2
V (zk )
< 0,
(27)

(xk , xk )zk

where z n and z 6= 0.
0 =

f r (xk ) xr,k+ + B r (xk )(xk ).


(22)

If the new error variables tend to zero, xk will converge to


x,k as desired. Then, a stabilizing control law (zk ) can be
used to achieve trajectory tracking.

L1 (z) + V (f (zk ))
1
1 T
(zk ),
V (zk ) P1u (zk ) (R(zk ) + P2u (zk )) P1uz
4
(28)

with z n , and
V (zk ) as ||zk || .

(29)

Then the zero solution zk 0 of the closed loop system


zk+1

= V (f (zk )) + V (zk )
1
T
P1uz (zk ) (R(zk ) + P2u (zk ))1 P1u
(zk )
+
4
+ uTzk R(zk )uzk

= f (zk ) + g(zk )(zk ), z(0) = z0 ,


(30)

is globally asymptotically stable with the feedback control law

= V [f (zk ) + g(zk )uzk ] + V (zk )


1
1 T
+
P1u (zk ) (R(zk ) + P2u (zk )) P1u
(zk )
4
+ uTzk R(zk )uzk + P1u (zk )uzk

1
1
(31)
(zk ) = (R(z) + P2u (zk )) P1u (zk ),
2
and the cost functional (24) is minimized in the sense that
J(z0 , (z()))

min

u()S(z0 )

J(z0 n ),

= V [f (zk ) + g(zk )uzk ] + V (zk )


+ [uzk + yzk ]T [R(zk ) + P2u (zk )] [uzk + yzk ]

(32)
Finally,
J((z0 ), (zk ())) = V (z0 ), z0 n .

(33)

Proof 1: Let consider the Hamiltonian

V (zk+1 ) + V (zk )

[uzk + yzk ] [R(zk ) + P2u (zk )] [uzk + yzk ] .

Then summing over [k1 , k2 ] yields (37).


For the nonlinear system (23), define

H(z, u) = L(z, u) + V [f (zk ) + g(zk )(zk )] V (zk ).

From (26)
H(zk , uzk ) = L(z, uzk ) + [V (f (zk )) + P1u (zk )uzk

(34)
+ uTz P2u (zk )uz V (zk ),

and L(zk , uzk ) = L1 (zk ) + uTzk R(zk )uzk ,


H(zk , uzk ) =

Then
H(zk , uzk ) =

uTzk R(zk )uzk

L1 (zk ) +
+ [V (f (zk )) + P1u (zk )uzk

+ uTzk P2u (zk )uzk V (zk ).

(35)

L1 (zk ) + uTzk [R(zk ) + P2u (zk )] uzk


+V (f (zk )) + P1u (zk )uzk V (zk ).
(36)

The feedback control (31) is obtained by setting


H(zk , uzk )
= 0. Considering (31), it follows that
uzk
(24)-(27) implies V (0) = 0, V (z) > 0, (0) = 0 and
V (F (zk , (zk )) V (zk ) < 0). Then system (23) is globally
asymptotically stable.
The following lemma is proposed in order to determine the
disk margin theorem of the inverse optimal control (31).
Lemma 1: Consider the nonlinear system (23), where (zk )
is a stabilizing feedback control law given by (31) and where
V (zk ), P1u (zk ), P2u (zk ) fulfill (24), (25), (26), (27) and (28).
Then for all u() U and k1 , k2 Z+ , k1 < k2 , the solution
xk , k Z+ , of the closed-loop system (23) satisfies
V (zk2 )

k2 1 h

(uzk + yzk )T (R(zk ) + P2u (zk ))(uzk + yzk ))

sup max (R(z) + P2u (z)),

(38)

(39)

zn

and

inf min (R(z)),

zn

with max (.) the maximum and min (.) the minimum singular
values.
In the next theorem a disk margin for system (23) is presented.
Theorem 2: Consider the nonlinear system (23), where
(zk ) is a stabilizing feedback control law given by (31)
and where V (zk ), P1u (zk ), P2u (zk ) satisfy (24), (25), (27)
and (28). Then 
nonlinear system (23) has a disk margin

q
1
1
, where z , z /z .
,
1 + z 1 z
Proof 3: Note that for all u() U and k1 , k2 Z+ ; it
follows from lemma (1) that the solution zk , k Z+ , of the
closed-loop system (30), satisfies
kX
2 1 h
T
[uzk + yzk ] [R(zk ) + P2u (zk )]
V (zk2 ) V (zk1 )
k=k1


(uzk + yzk ) uTzk R(zk )uzk ,
(40)
which implies
V (zk2 ) V (zk1 )

k2 1 h

z (uzk + yzk )T (uzk + yzk )

k=k1

i
z uTzk uzk .
(41)

1
k=k1
V (zk ), (23) is
Hence, with the storage function Vs (zk ) =
i
2
z
uTzk R(zk )uzk + V (zk1 ).
dissipative with respect to supply rate rz (uzk , yzk ) = uTzk yzk +
(37) (1 2 )
1
z
uTzk uzk + yzTk yzk . Applying corollary (13.17) and
2
2
Proof 2: Note that from (28), it follows that for all uz () U definition (6.3) from [6], then system (23) is globally asympand k Z+
totically stable. Hence, we conclude that (23) has a disk margin
T
T
(z , z ).
uz R(zk )uzk L1 (zk ) + uz R(zk )uzk
k

IV. L INEAR INDUCTION MOTOR


To illustrate the applicability of Theorem 2 results, we use
a linear induction motor, whose discrete-time mathematical
model ([13] ,[14]), is described as follows:
qk+1
vk+1

=
=

k+1

k+1

ik+1

ik+1

qk + vk T
(1 k2 )vk k3 FL T (k1 k 1 + k1 k 2 ) T ik
+ (k1 k 2 k1 k 1 ) T ik
(1 k6 T )k + (k4 vk 1 k4 1 + k5 2 ) T ik
+ (k4 2 k4 vk 2 + k5 1 ) T ik
(1 k6 T )k + (k4 vk 2 k4 2 k5 1 ) T ik
+ (k4 1 + k4 vk 1 + k5 2 ) T ik
(1 + k9 T )ik T k7 k 2 T k8 k vk 1
+T k7 k 1 T k8 k vk 2 T k10 uk
(1 + k9 T )ik + T k8 k vk 2 T k7 k 1
T k7 k 2 T k8 k vk 1 T k10 uk .
(42)

Consequently the discrete-time flux module is:


m,k+1

(1 2k6 T )m,k
+2T ik [(k4 vk 1 k4 1 + k5 2 ) k
+ (k4 vk 2 k4 2 k5 1 ) k ]
+2T ik [(k4 1 + k4 vk 1 + k5 2 ) k
+ (k4 2 k4 vk 2 + k5 1 ) k ] ,

Hence, the control goal is to track the trajectory for the


velocity and the flux module. For the tracking error Zk1 ,
Zk1
1
Zk+1

1
Xk1 X,k
,

1
1
Xk+1
X,k+1
= K1 Zk1 ,

(45)
where
1
X,k

0.5sin(t)
0.1

(46)

Then,
Zk2
2
Zk+1

2
Xk2 X,k
,

=
=

2
2
Xk+1
X,k+1
.

(47)
The tracking error block controlable form is represented as
follows:

fz

k11
0

0
k12

 
A2
Zk11
+
B2
Zk12
21
C1 X,k+1
22
D1 X,k+1



A3
B3

(48)

(43)
where

and
1 = sin(np qk );
k1 =

np Lsr
;
Dm Lr

k4 = np Lsr ;

Rm
;
Dm
Rr Lsr
;
k5 =
Ls r
k2 =

Lsr Rr
;
Lr (L2sr Ls Lr )
L2r Rs + L2sr Rr
k9 =
;
Lr (L2sr Ls Lr )

k7 =

2 = cos(np qk );
1
;
Dm
Rr
;
k6 =
Lr

Ls rnp
;
L2sr Ls Lr
Lr
=
;
Lr (L2sr Ls Lr )

k8 =
k10

gz

k3 =

qk is the position, vk is the linear velocity, ,k and ,k are


the magnetic flows, i,k and i,k are the currents, u,k and
u,k are the input voltages, Rs is the primary sector resistance,
Rr is the secondary sector resistance, Lsr is magnetizing
inductance, Ls is the primary sector inductance, Lr is the
secondary sector inductance, FL is the load perturbation, Rm
is the viscous friction coefficient, Dm is the secondary sector
mass, np is the number of poles pairs and T is the sampling
time. This model does not take into account the end effects
[15], [16], which can be despised depending on the dimensions
of LIM.
V. L INEAR I NDUCTION M OTOR S IMULATION R ESULTS
To implement an inverse optimal control for the linear
induction model, the block control form is considered [12],
[14]; then the variables are proposed as follows:




ik
vk
2
1
;
Xk =
.
(44)
Xk =
ik
m,k

0
0

C2
0

0
0
,
0
D2

(49)

where
A1
A2
A3
B1
B2

=
=
=
=
=

B3

C1

C2
D1

=
=

D2

(1 k2 )vk k3 FL T,
(k1 k 1 + k1 k 2 ) ,
(k1 k 2 k1 k 1 ) ,
(1 2k6 T )m,k ,
2T [(k4 vk 1 k4 1 + k5 2 ) k
+ (k4 vk 2 k4 2 k5 1 ) k ] ,
2T [(k4 1 + k4 vk 1 + k5 2 ) k
+ (k4 2 k4 vk 2 + k5 1 ) k ] ,
(1 + k9 T )ik T k7 k 2 1
T k8 k vk 1 + T k7 k T k8 k vk 2 ,
T k10 ,
(1 + k9 T )ik + T k8 k vk 2 1
T k7 k 1 T k7 k 2 T k8 k vk 1 ,
T k10 .
(50)

Then for trajectory tracking, we apply the inverse opti1


mal control law (31), with P2u (zk ) = g T (zk )P g(zk ) and
2
P1u (zk ) = g T (zk )P f (zk ). The results for the discrete-time
lineal induction motor model are depicted in the Fig.3 and

Fig 4; the solid line corresponds to measure of velocity and


flux module; the dashed one to the desired references.
0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0.1

0.1

0.2

0.2

0.3

0.3

0.4

0.4

Velocity (m/s)

reference
measure

0.5

0.02
0.04
time (sec)

0.5

5
6
time (sec)

10

Fig. 3. Trajectory tracking for velocity vk .

0.2

0.2

0.18

0.18

0.16

0.16

0.14

0.14

0.12

0.12

0.1

0.1

0.08

0.08

0.06

0.06

0.04

0.04

0.02

0.02

Flux module (wb2)

reference
measure

0.02
0.04
time (sec)

5
6
time (sec)

10

Fig. 4. Trajectory tracking for flux module m,k .

Based on the results obtained in section (III) and in order


to to probe the robustness of the designed controller, we
realize a parametric variation of rs , rr . The obtained results
are presented in the table (I); where MSE is the mean squared
error.
TABLE I
PARAMETRIC VARIATION OF LIM
Variation of
rs and rr
20%
10%
nominal
+10%
+20%

MSE
vk
6.0894e-06
6.4085e-06
5.4873e-06
8.0743e-06
8.8954e-06

MSE
m,k
1.2786e-06
1.5436e-06
2.7618e-06
1.6200e-05
1.9965e-05

Disk
Margin
(0.6801,1.888)
(0.6801,1.888)
(0.6801,1.888)
(0.6801,1.888)
(0.6801,1.888)

VI. CONCLUSION
In this paper, robustness of discrete-time inverse optimal
control for trajectory tracking is analyzed by means of the
disk margin concept. This result is illustrated using a lineal
induction motor. We are able to determine a disk margin for
this kind of motor as we claim in Theorem 2.

R EFERENCES
[1] T. T. Lee and S. H. Lee, Gain and phase margins for discrete-time
systems, International Journal of Control, vol. 44, pp. 14151426, Nov
1986.
[2] F. Ornelas-Tellez, E. N. Sanchez, and A. G. Loukianov, Discrete-time
inverse optimal control for nonlinear systems trajectory tracking, in The
49th IEEE Conference on Decision and Control (CDC), Atlanta, GA,
U.S.A., pp. 48134818, December 2010.
[3] E. N. Sanchez and F. Ornelas-Tellez, Discrete-Time Inverse Optimal
Control for Nonlinear Systems. Boca Raton, FL, U.S.A.: CRC Press,
2013.
[4] H. Toliyat and G. B. Kliman, Handbook of Electric Motors. Boca Raton,
FL, USA: CRC Press, 2 ed., 2004.
[5] D. Hill and P. Moylan, Stability results for nonlinear feedback systems,
Automatica, vol. 13, pp. 377382, Jul 1977.
[6] W. M. Haddad and V. Chellaboina, Nonlinear Dynamical Systems
and Control : a Lyapunov-based Approach. Princeton, N. J., U.S.A.:
Princeton University Press, 2008.
[7] F. L. Lewis and V. L. Syrmos, Optimal Control. New York, N. Y.,
U.S.A.: Wiley, 1995.
[8] D. E. Kirk, Optimal Control Theory: An Introduction. Englewood Cliffs,
N. J., U.S.A.: Prentice-Hall, 1970.
[9] R. A. Freeman and P. V. Kokotovic, Robust Nonlinear Control Design: State-Space and Lyapunov Techniques. Cambridge, MA, U.S.A.:
Birkhauser, 1996.
[10] M. Krstic and H. Deng, Stabilization of Nonlinear Uncertain Systems.
Berlin, Germany: Springer-Verlag., 1998.
[11] F. Ornelas-Tellez, E. N. Sanchez, and A. G. Loukianov, Discrete-time
neural inverse optimal control for nonlinear systems via passivation,
IEEE Transactions on Neural Networks and Learning Systems, vol. 23,
pp. 13271339, 2012.
[12] A. G. Loukianov, Robust block decomposition sliding mode control
design, Mathematical Problems in Engineering, vol. 8, 2002.
[13] M. Hernandez, Discrete-Time Neural Control for a Linear Induction
Motor, in Spanish. M. Sc Thesis, Centro de Investigacion y Estudios
Avanzados del I.P.N. Unidad Guadalajara, 2008.
[14] V. G. Lopez, E. N. Sanchez, and A. Y. Alanis, PSO neural inverse
optimal control for a linear induction motor, in IEEE Congress on
Evolutionary Computation (CEC), Cancun, QR, Mexico, pp. 19761982,
June 2013.
[15] H. A. Hairik and M. H. Hassan, Dynamic model of linear induction
motor considering the end effects, Iraqi Journal for Electrical &
Electronic Engineering, vol. 5, no. 1, pp. 3850, 2009.
[16] L. Radzevicius and E. Matkevicius, The generalized model of the
linear induction motor, Electronics and Electrical Engineering, vol. T,
no. 190, 2006.

Potrebbero piacerti anche