Neural Comput & Applic (2014) 24:277285

DOI 10.1007/s00521-012-1224-7

ORIGINAL ARTICLE

Mathematical model with sensor and actuator for a transelevator

Jose de Jesus Rubio Jaime Pacheco
J. Humberto Perez-Cruz Francisco Torres

Received: 9 December 2011 / Accepted: 23 September 2012 / Published online: 21 October 2012
Springer-Verlag London 2012

Abstract In this paper, the structural, sensor, and actuator mathematical models of a transelevator are presented;
after, the mathematical model with sensor and actuator of
the transelevator is obtained by using the combination of
the above mentioned mathematical models. The proposed
mathematical model is validated comparing the simulation
results against the experimental results. Finally, the stability analysis of the aforementioned model is studied.
Keywords Mathematical model  Transelevator 
Robotic systems  Validation

1 Introduction
The structural mathematical model is used for the control
applications, but when a system is controlled, a sensor and
an actuator are required, so this work is interested to obtain
the mathematical model with sensor and actuator for a
transelevator. A transelevator is a crane similar to a
Cartesian robotic arm which is moved inside of a warehouse, this transelevator is used to move loads from one
place to another one inside of the warehouse.

J. de Jesus Rubio (&)  J. Pacheco  F. Torres

Seccion de Estudios de Posgrado e Investigacion,
ESIME Azcapotzalco, Instituto Politecnico Nacional,
Av. de las Granjas no. 682, 02250 Col. Santa Catarina,
Mexico D.F., Mexico
e-mail: [email protected]; [email protected]
J. H. Perez-Cruz
Centro Universitario de Ciencias Exactas e Ingenieras,
Universidad de Guadalajara, Boulevard Marcelino Garca
Barragan No. 1421, 44430 Guadalajara, JAL, Mexico

The research about the robotic arms is classified in three

kinds: the trajectory planning [6, 14], the position estimation [1, 4, 19], and the control [2, 3, 5, 7, 20].
The trajectory planning of a robotic arm has the objective to provide the best trajectory of the robotic arm from
an initial point to a final point. In [6], they address the
challenge of realizing full-body behaviors in scalable
modular robots. The trajectories planning and a collisions
detector for robotic arms are presented by [14].
The position estimation is to estimate the parameters of
the position and the orientation using other parameters as
are the parameters of an image, an identifier or an observer
can be used for the position estimation. In [1], the authors
show that spatial computing language (Proto) can be
applied to multi-mobile-robot systems with an actuator that
turns a vector field into device motion. The identification of
a golf swing robot based on four kinds of soft computing
methods is investigated by [4]. In [19], an innovative
neuro-evolutionary approach for mobile robot ego-motion
estimation with a 3D camera is proposed.
The control has the objective that the robotic system has to
follow a desired trajectory. In [2], they study how a selforganized mobile robot flock can be steered toward a desired
direction through externally guiding some of its members.
An adaptive neural network sensor-less control scheme for
machines is introduced by [3]. In [5], a model-free selftuning output recurrent cerebellar model articulation controller to control an inverted pendulum is investigated. A
network structure develops a biologically plausible GRN
model for robot control is adopted by [7]. In [20], a robust
training algorithm of multi-input multi-output recurrent
neural network and its application in the fault tolerant control
of a robotic system are investigated.
For the trajectory planning, the position estimation, or
the control, the mathematical models help to obtain the

123

278

Neural Comput & Applic (2014) 24:277285

objective, that is, it is necessary to obtain a mathematical

model of the system or to obtain the measure of all the
parameters of the system. It is less expensive to obtain a
mathematical model of the system than to get the measure
of all the parameters of the system.
In most of the aforementioned cases, only the structural
mathematical model is used, but when a system is controlled, a sensor and an actuator are required; therefore, this
work is interested to obtain the mathematical model with
sensor and actuator for a transelevator.
This paper is organized as follows: In Sect. 2, the
method to obtain the structural mathematical model is
detailed. In Sect. 3, the structural, sensor, and actuator
mathematical models are obtained. In Sect. 4, the mathematical model with sensor and actuator for a transelevator
is obtained. In Sect. 5, the validation of the proposed
mathematical model is shown. In Sect. 6, the stability
analysis of the proposed model is presented. In Sect. 7, the
conclusion and the future research are detailed.

2 Preliminaries
The Euler Lagragian method is used to obtain the mathematical model of the transelevator. The first step of this
method is to find the kinetic energy Ki, it is the addition of
the speed of the motion and the rotation kinetic energy, it is
given as follows:
n 2
1
1 X
Ki mi v2i Ji hi
2
2 i1

n
X
Ki  Vi

As the fourth step, the following equation is used to obtain

each dynamic equation for the mathematical model:
d oL oL

si
dt oq i oqi

3 Mathematical model of the elements

of the transelevator
3.1 Structural mathematical model
A transelevator is a crane similar to a Cartesian robotic arm
which moves inside of a warehouse. This transelevator is
used to move loads from one place to another one inside of
the warehouse, for example, from the floor to the deposit,
from the deposit to the floor, from one deposit to another
one, or from a trailer to a deposit. Figure 1 shows the
transelevator.
From the Fig. 1, the parameters X1 ; Y1 ; and Z1 of the
Eq. (1) are:

and Zi are the derivatives of Xi ; Yi ; and Zi ; respectively,
Xi ; Yi ; and Zi ; are the coordinates of the linear movement of
the link i.
The second step is to find the potential energy, it is given
as follows:
2

where i 2 1; . . .; n; hi is the height of the center of mass of

the joint and the link i, g is the gravity acceleration.
As the third step, the Lagrangian is obtained as
follows:

123

Remark 1 The method to obtain the mathematical model

of this paper is different to others because they use
Jacobians, inertia tensors, or Christoffel symbols while in
this paper none of these methods is used.

where i 1; . . .; n; n is the number of links, vi is the speed

of motion for the joint i, mi is the center of mass of the joint
i and the link i, if the joint i is rotational Ji is its inertia, hi is
the angular position of the joint i, if the angular positions
2
2
2
P 2
have different directions ni1 hi h1 h2    hn or if
Pn  2
the angular positions have the same direction
i1 hi

2
2
2
 

 2
h1 h2    hn ;v2i Xi Yi Zi where Xi ;Yi ;

Vi mi ghi

i1

Fig. 1 The transelevator

Neural Comput & Applic (2014) 24:277285

X1 x1
Y1 0
Z1 0

279


m13 x1 s1
5

where x1 is the length to the mass center of the link 1. From

the Fig. 1, the parameter x1 is bounded as follows
0 B x1 B xmax. Using (1) and (5), the kinetic energy for
the joint 1 is:
1 2
K1 m1 x1
2

X 2 x1
Y2 0
Z 2 z2

where z2 is the length to the mass center of the link 2. From

the Fig. 1 , the parameter z2 is bounded as follows
0 B z2 B zmax. Using (1) and (7), the kinetic energy for
the joint 2 is:
1 2 1 2
K2 m2 x1 m2 z2
2
2

From the Fig. 1, the parameters X 3 ; Y 3 and Z 3 of the Eq. (1)

are:
X 3 x1
Y 3 y3

Z 3 z2
where y3 is the length to the mass center of the link 3. From
the Fig. 1, the parameter y3 is bounded as follows
0 B y3 B ymax. Using (1) and (9), the kinetic energy for
the joint 3 is:
1 2 1 2 1 2
K3 m3 x1 m3 z2 m3 y3
2
2
2

10

11

V3 m3 gz2
Substituting the Eqs. (6), (8), (10), and (11) in (3) gives the
Lagrangian as follows:
1
1
1 2
2
2
L m13 x1 m23 z2 m3 y3  m23 gz2
2
2
2



m23 z2 m23 g s2

12

where m13 = m1 ? m2 ? m3 and m23 = m2 ? m3.

Using the Euler Lagrange method of Eq. (4) in (12) for
the joint 1 gives:

14

where s2 is the force needed to move the link 2.

Using the Euler Lagrange method of Eq. (4) in (12) for
the joint 3 gives:


15

where s3 is the force needed to move the link 3 from the

home position to another one.
The structural mathematical model of the transelevator
crane is given by Eqs. (13), (14), and (15).
Remark 2 The method of this paper is similar to the
method used by [10], but [10] assumes that the link masses
are concentrated at the end of the links and the inertia is not
considered while in this paper the link masses are concentrated at the half of the link and the inertia is considered. In addition, [10] does not present the mathematical
models of the transelevator.
3.2 Actuator mathematical model
It is necessary to monitor and to act over some variables to
control a transelevator; therefore, an actuator and a sensor
are required.
Figure 2 is used to obtain the following mathematical
model of a motor of each link:
L

dia
Rmi iai Vi  Vbi
dt

16

where:
Vbi Kbi h_ m

From the Fig. 1, the potential energy for the three links is
as follows:
V1 0
V2 m2 gz2

where s1 is the force needed to move the link 1.

Using the Euler Lagrange method of Eq. (4) in (12) for
the joint 2 gives:

m 3 y 3 s3

From the Fig. 1, the parameters x2, y2, and z2 of the Eq. (1)
are:

13

17

where i = 1, 2, 3, iai is the current in the motor link i, Li is

the inductance of the motor for the link i, Rmi is the
resistance of the motor for link i, Vi is the input voltage of
the motor for link i, kbi is the constant back emf of the
motor for link i; h_ i is the speed of the angle of the motor of
link i.
Figure 3 shows the actuator circuit, this circuit is used to
obtain the current applied for the motors in each link iai as
function of the voltage of the motor Vi of Eq. (16).
From Fig. 3, it is obtained:
Vfi VRci Vmi

18

where Vfi is the input voltage for link i for the actuator
circuit, VRci is the voltage in the resistance for link i for the
actuator circuit, Vmi Vi is the input voltage of the motor
for link i for the actuator circuit, it is known that:

123

280

Neural Comput & Applic (2014) 24:277285

Fig. 4 Voltage divider sensor circuit

h1 max
x1
xmax
h2 max
h2
z2
z2 max
h3 max
h3
y3
y3 max

Fig. 2 Circuit of a motor controlled by DC

h1

24

the derivative of Eq. (24) gives:



VRci Roi ici

19

where i = 1, 2, 3, Rci = Roi is the resistance for the actuator circuit, ici is the current in the resistance Rci.
Substituting the Eq. (19) in the Eq. (18) gives:
Vfi Rci ici Vm1

20

and as the circuit of Fig. (3) is a series circuit, it is known

that:
ici imi ivfi

21

where ici is the current in the resistance Rci for link

i, iai = imi is the current in the motor for link i, and ivfi is
the source current for link i.
Substituting Eq. (21) in Eq. (20) and solving Vmi gives:
Vfi Rci imi Vmi
Vmi Vfi  Rci ici

where himax is the maximum angle of the potentiometer

(radians), x1max, y3max, and z2max are the distance in each link
(m), and x1, z2, and y3 are the displacements of each link.
From Fig. (4), it is obtained:
Voi
hi

Voi max hi max

26

where hi is the angle of the motor of each link, himax is the

maximum angle for the potentiometer of each link, Voi is
the measured voltage in the potentiometer, Voimax is the
maximum measured voltage in the potentiometer.

4 Mathematical model with sensor and actuator

for a transelevator
23

where kmi is the constant torque for link i, imi is the current
in the motor for link i; smi si is the torque in the motor,
and Vmi = Vi.
3.3 Sensor mathematical model
Figure 4 shows the sensor circuit, it is known as a voltage
divider where the output voltage is time-varying depending
of the measured position in each link; that is, this circuit is
used to obtain the displacements of each link in function its
the motor angle h_ i ; i 1; 2; 3: It is important to know that:

123

25

22

From [18], it is known that:

smi kmi imi

h1 max 
x1
xmax

h2 max 
h2
z2
z2 max

h3 max 
h3
y
y3 max 3
h1

Fig. 3 Actuators circuit

Substituting Eq. (23) in the Eq. (13) for the structural

mathematical model for the first link x gives:
m13 x1 km1 im1

27

Substituting Eq. (17) and Eq. (22) into Eq. (16) for i = 1
gives:
L1

d
im1 Rm1 im1 Vf 1  Rc1 im1  kb1 h_ 1
dt

28

Substituting Eq. (25) in the Eq. (28) gives:

L1

d
h1 max 
im1 Rm1 Rc1 im1 kb1
x 1 Vf 1
dt
xmax

29

Neural Comput & Applic (2014) 24:277285

281

From Eq. (26), the output for i = 1 is obtained as follows:

Vo1

Vo1 max
h1
h1 max

30

where i 1; 2; 3; j 1; 9:
Substituting the inputs, the states, and the outputs of
Eq. (39) into Eqs. (27), (29) (30), (31), (33), (34), (35),
(37), and (38) yields:

Equations (27), (29), and (30) describe the dynamic

behavior of the first link.
Substituting Eq. (23) in the Eq. (14) for the second link z
gives:

X_ 1 X2

40

km1
X3
X_ 2
m13

41

m23
z2 m23 g km2 im2

kb1 h1 max
Rm1 Rc1
1
X_ 3 
X2 
X 3 u1
L1
L1
L1 xmax

42

X_ 4 X5

43

km2
X_ 5
X6  g
m23

44

kb2 h2 max
Rm2 Rc2
1
X_ 6 
X5 
X 6 u2
L2
L2
L2 zmax

45

X_ 7 X8

46

km3
X_ 8
X9
m3

47

kb3 h3 max
Rm3 Rc3
1
X_ 9 
X8 
X 9 u3
L3
L3
L2 ymax

48

31

Substituting Eq. (17) and Eq. (22) in the Eq. (16) for i = 2
gives:
L2

d
im2 Rm2 im2 Vf 2  Rc2 im2  kb2 h_ 2
dt

32

Substituting Eq. (25) in the Eq. (32) gives:

d
h2 max 
z 2 Vf 2
L2 im2 Rm2 Rc2 im2 kb2
dt
zmax

33

From Eq. (26), the output for i = 2 is obtained as follows:

Vo2

Vo2 max

h2
h2 max

34

Equations (31), (33), and (34) describe the dynamic

behavior of the second link z.
Substituting Eq. (23) in the Eq. (15) and analyzing the
model mathematical for the third link gives:
m3 y3 km3 im3

35

Substituting Eq. (17) and Eq. (22) in the Eq. (16) for i = 3
gives:
L3

d
im3 Rm3 im3 Vf 3  Rc3 im3  kb3 h_ 3
dt

36

Y1

Vo1 max
X1
h1 max

49

Y2

Vo2 max
X4
h2 max

50

Y3

Vo3 max
X7
h3 max

51

Equations (40)(51) describe the dynamic behavior of the

transelevator crane with sensor and actuator.

Substituting Eq. (25) in the Eq. (36) gives:

L3

d
h3 max 
im3 Rm3 Rc3 im3 kb3
y Vf 3
dt
ymax 3

37

From Eq. (26), the output for i = 3 is obtained as follows:

Vo3

Vo3 max
h3
h3 max

38

Equations (35), (37), and (38) describe the dynamic

behavior of the third link y.
Let us define the parameters Voimax = constant, Rci =
constant, Rmi = constant, himax = constant, xmax = constant, zmax = constant, ymax = constant, the inputs Ui, the
states Xj, and the outputs Yi as follows:
U1 Vf1 ;
X1 x 1 ;
X4 z 2 ;
X7 y 3 ;
Y1 Vo1 ;

U2 Vf2 ; U3 Vf3

X2 x1 ; X3 im1

X5 z2 ; X6 im2

X8 y3 ; X9 im3
Y2 Vo2 ; Y3 Vo3

39
Fig. 5 Circuit of the sensor and the actuator

123

282

Neural Comput & Applic (2014) 24:277285

5 Validation of the mathematical model

To validate the mathematical model, the simulation results
will be compared with the experimental results. The simulations are obtained using the Matlab, the experimental
results are obtained using the prototype of Fig. 1. Figure 5
shows the sensor and actuator which are connected to the
prototype of Fig. 1 to obtain the experimental results.

Table 1 Parameters of links 1, 2, and 3

V1 = 5 V
Rm1 22:8 X

V2 = 5 V
Rm2 125:5 X

Rm3 17 X

Rc2 1 X

Rc3 1 X

L1 = 42.5 mH

L2 = 96.1 mH

L3 = 29.9 mH

km1 = 10 Nm/A

km2 = 1 9 103 Nm/A

km3 = 1 Nm/A

kb1 = 27.50 Vs/rad

kb2 = 0.014 Vs/rad

kb3 = 0.26 Vs/rad

Vf1 = 14 V

Vf2 = 10 V

Vf2 = 4 V

Vo1max = 2.36 V

Vo2max = 2.3 V

Vo3max = 2 V

m13 = 2.9020 kg

m23 = 0.5480 kg

m3 = 0.150 kg

h1max = 12.56 rad

h2max = 42.4115 rad

h3max = 10.9956 rad

xmax = 0.45 m

zmax = 0.29 m

ymax = 0.23 m

where V1, V2, and V3 are the voltages of the circuit of Fig. 2,
Rm1, Rm2, and Rm3 are the values of motor resistance, L1, L2, and L3
are the values of motor inductance, Rc1, Rc2, and Rc3 are the values of
resistance for the circuit of Fig. 3, kb1, kb2, and kb1 are the constants
back emf of the motor, Vf1, Vf2, and Vf3 are the motor supply voltages,
km1, km2, and km3 are the nominal torques of the motors, h1max,
h2max, and h3max are the maximum angles found by the potentiometer,
xmax, zmax, and ymax are the maximum distances that run through the
links, m13 = m1 ? m2 ? m3, m23 = m2 ? m3, and m3 are the link
masses.

123

V3 = 5 V

Rc1 1 X

Fig. 6 Comparison results from

simulation and experiment

The mathematical model of the transelevator with sensor

and actuator is described by Eqs. (40)(51). Table 1 shows
the parameters used in the simulation results and the
experimental results.
In this section, the root mean square error (RMSE)
for the inputs and outputs [15] is used to find the effectiveness of the proposed mathematical model, it is given
as:
0 T
112
Z
1
RMSE @
e2 dsA
52
T
where e = e2u = (U1r - U1)2 ? (U2r - U2)2 ? (U3r U3)2, or e2 = e2y = (Y1r - Y1)2 ? (Y2r - Y2)2 ? (Y3r Y3)2, Uir and Yir are the real data of Ui and Yi, respectively.
Figure 6 shows the simulation and the experiment of the
input voltage and the output voltage denoted as mathematical model and real data, respectively.
From Fig. 6, it can be seen that both results are almost
equal, for the inputs are 14 V for U1 and U1r, 10 V for U2
and U2r, and 4 V for U3 and U3r, respectively, and for the
output are 2.0216 V for Y1 and Y1r, 2.2796 V for Y2 and
Y2r, and 1.4921 V for Y3 and Y3r, respectively.
The error of the input Uir - Ui and the output Yir - Yi
denoted as Uei and Yei, respectively, are shown in Fig. 7.
From Fig. 7, it can be observed that both results are
almost equal.
The simulation results for the states Xj are shown in
Fig. 8.
Figure 8 shows the states X1, X4, and X7 as the displacements x1, z2, and y3, the states X2, X5, and X8 as the
2

Neural Comput & Applic (2014) 24:277285

283

Fig. 7 Error results from

simulation and experiment

Fig. 8 Simulation results for

the states

velocities of the displacements x1 ; z2 , and y3 , and the states

X3, X5, and X9 as the armature current in the motors i1, i2,
and i3.
The Table 2 shows the RMSE using (52) for the errors
e2 = e2u, or e2 = e2y , using the mathematical model
described by Eqs. (40)(51).

Table 2 The RMSE for the errors

RMSE
e2u
e2y

0.6553
0.0080

123

284

Neural Comput & Applic (2014) 24:277285

167.10i,
s = 103.34 - 167.10i, s = - 1279.4, s =
- 730.88, s = 0, s = 0, s = 0, the above elements are the
system eigenvalues. Since not all the real parts of the
eigenvalues are negative, the proposed mathematical
model is not guaranteed to be stable.

From Table 2, it is shown that the mathematical model

is a good approximation of the real process because the
RMSE is near to zero.

6 Stability analysis
7 Conclusion

This section presents the stability of the proposed mathematical model.

The mathematical model of Eqs. (40)(51) can be
rewritten as follows:

In this paper, the mathematical model with sensor and

actuator for a transelevator was obtained. The proposed
mathematical model was validated with real measures
where the proposed approach gave a good approximation
of the real process. In addition, the stability analysis was
presented where the aforementioned model was not guaranteed to be stable. As a future research, an observer, a

X AX BU F
Y CX

53

where:

60
km1
0
6
m13
6
6
k h max
c1
 Rm1LR
6 0  Lb11 x1max
1
6
60
0
0
6
6
0
0
0
A6
6
6
60
0
0
6
60
0
0
6
6
60
0
0
4
0
0
0
h
B 0 0 L11 0 0 L121 0
F 0 0
2V

o1 max

h1 max

6
C6
4 0
0

0 g

0
0

1
0

0
0

0
0

0
0

max
 kLb22hz2max

km2
m23
c2
 Rm2LR
2

0
0

0
0

0
0

0
0

1
0

max
0  kLb32hy3max

1
L3

iT

0 0

0 T

Vo2 max
h2 max

Vo3 max
h3 max

0 0

7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5

54

7
0 07
5
0 0

Since F is only a constant element, the system (53) is

considered as a linear system. Using the values of Table 1,
the stability of the proposed mathematical model is
obtained as follows:
detsI  A 0
) 7056  1012 s3 1:7440  1012 s4 6:7166  1010 s5
6:1353  108 s6 1:9706  106 s7 2478:3s8 s9 0
The roots of the polynomial are s = - 337.37 ?
285.77i, s = - 337.37
- 285.77i, s = 103.34 ?

123

0
km3
m3
c3
 Rm3LR
3

control, or an identifier will be applied for the proposed

mathematical model [8, 9, 11, 12, 13, 16, 17].
Acknowledgments The authors are grateful to the editors and
the reviewers for their valuable comments and insightful suggestions, which helped to improve this research significantly. The
authors thank the Secretara de Investigacion y Posgrado and the
Comision de Operacion y Fomento de Actividades Academicas
del IPN and the Consejo Nacional de Ciencia y Tecnologa
for their help in this research. The third author would like to
thank the financial support through a postdoctoral fellowship
from Mexican National Council for Science and Technology
(CONACYT).

Neural Comput & Applic (2014) 24:277285

References
1. Bachrach J, Beal J, McLurkin J (2010) Composable continuousspace programs for robotic swarms. Neural Comput Appl
19:825847
2. Celikkanat H, Sahin E (2010) Steering self-organized robot flocks
through externally guided individuals. Neural Comput Appl
19:849865
3. Chaoui H, Sicard P (2011) Adaptive Lyapunov-based neural
network sensorless control of permanent magnet synchronous
machines. Neural Comput Appl 20:717727
4. Chen C, Inoue Y, Shibata K (2011) Identification of a golf swing
robot using soft computing approach. Neural Comput Appl
20:729740
5. Chiu C-H (2010) Self-tuning output recurrent cerebellar model
articulation controller for a wheeled inverted pendulum control.
Neural Comput Appl 19:11531164
6. Christensen DJ, Campbell J, Stoy K (2010) Anatomy-based
organization of morphology and control in self-reconfigurable
modular robots. Neural Comput Appl 19:787805
7. Lee W-P, Yang T-H (2011) Combining GRN modeling and
demonstration-based programming for robot control. Neural
Comput Appl 20:909921
8. Leite D, Ballini R, Costa P, Gomide F (2012) Evolving fuzzy
granular modeling from nonstationary fuzzy data streams. Evolv
Syst 3(2):6579
9. Lemos A, Caminhas W, Gomide F (2011) Fuzzy evolving linear
regression trees. Evolv Syst 2(1):114
10. Lewis FL, Dawson DM, Abdallah CT (2004) Control of robot
manipulators. Theory Pract. ISBN: 0-8247-4072-6

285
11. Lughofer E, Bouchot JL (2011) On-line elimination of local
redundancies in evolving fuzzy systems. Evolv Syst 2(3):
165187
12. Perez-Cruz JH, Alanis AY, Rubio JJ, Pacheco J (2012) System
identification using multilayer differential neural networks: a new
result. J Appl Math 2012:120
13. Perez-Cruz JH, Ruiz-Velazquez E, Rubio JJ, de Alba-Padilla CA
(2012) Robust adaptive neurocontrol of SISO nonlinear systems
preceded by unknow deadzone. Math Probl Eng 2012:123
14. Rubio JJ, Garca E, Pacheco J (2011) Trajectory planning and
collisions detector for robotic arms. Neural Comput App, online
15. Rubio JJ, Torres C, Rivera R, Hernandez CA (2011) Comparison
of four mathematical models for braking of a motorcycle. IEEE
Latin Am Trans 9(5):630637
16. Rubio JJ, Ortiz F, Mariaca CR, Tovar JC (2011) A method for
online pattern recognition for abnormal eye movements. Neural
Comput Appl. doi: 10.1007/s00521-011-0784-2
17. Rubio JJ, Salazar M, Gomez AD, Lugo R (2012) Modeling of the
relative humidity via functional networks and control of the
temperature via classic controllers for a bird incubator. Neural
Comput Appl 21(7):14911500
18. Spong MW, Hutchinson S, Vidyasagar M (2006) Robot modeling
and control. Wiley, London. ISBN-13: 978-0-471-64990-8
19. Villaverde I, Grana M (2011) Neuro-evolutionary mobile robot
egomotion estimation with a 3D ToF camera. Neural Comput
Appl 20:345354
20. Wu Y, Sun F, Zheng J, Song Q (2010) A robust training
algorithm of discrete-time MIMO RNN and application in fault
tolerant control of robotic system. Neural Comput Appl 19:
10131027

123