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

net/publication/27557968

Tracking control of a mobile robot based on Taylor Formula

Article · November 2007

Source: OAI

CITATIONS READS
0 243

3 authors:

Andres Rosales Gustavo Juan Eduardo Scaglia

Escuela Politécnica Nacional National Scientific and Technical Research Council
58 PUBLICATIONS   481 CITATIONS    149 PUBLICATIONS   1,187 CITATIONS   

SEE PROFILE SEE PROFILE

Vicente A. Mut
National University of San Juan
155 PUBLICATIONS   2,037 CITATIONS   

SEE PROFILE

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

Mobile Robots Teleoperation View project

A linear algebra controller based on reduced order models applied to trajectory tracking for mobile robots: An experimental validation View project

All content following this page was uploaded by Vicente A. Mut on 27 May 2014.

The user has requested enhancement of the downloaded file.

 TRACKING CONTROL OF A MOBILE ROBOT BASED ON TAYLOR
FORMULA
Scaglia Gustavo
Mut Vicente
Rosales Andres

Universidad Nacional de San Juan(U.N.S.J.),Argentina

Abstract−− In this work, a strategy to calculate velocity is null or not. The work of Shuli (2005) shows
the control actions for a mobile robot following a pre only simulation results.
established trajectory is proposed. For this purpose,
the Taylor series development of controlled variables In Scaglia et al. (2005) a control based on numeric
is used and the control action is calculated to make methods is shown, where the mobile robot linear speed
the system follow the reference trajectory. As main depends on the current and desired orientation. In
result, an easy form to implement the strategy is Scaglia et al. (2006a) the integral trapezoidal method is
obtained. Simulation and experimental results are used to compute the control signal. The design is based
presented, showing the advantages of the control on the model of Kanayama (1990) and here, conditions
strategy proposed. are established over the tracking error so that the
equations system can always have solution. In (Scaglia
Keywords−− Control system design, discrete G., et al., 2006b) the controller is computed solving the
systems, models, tracking, trajectory control, normal equations. In (Scaglia G., et al., 2007) the
numerical methods. system evolution is approximated by means of a linear
I. INTRODUCCIÓN interpolation method, this allows a better system
precision. The strategies previously mentioned can be
Trajectory tracking is one of the main problems of applied to other systems, as can seen in Rosales et al.
control systems for which, in a search for a small (2006), where the controller design for the RTAC
tracking error, various controllers have been proposed in system (also called TORA, a classic nonlinear problem,
the literature, for example, Kim (2003) proposes a which is interesting due to the interaction between the
receding horizon tracking control for time-varying traslational and rotational movements) is proposed in
linear systems with constraints both on the control order that the system tends to the origin from any initial
signal and on the tracking error, based on the position. The proposed design is based on a system
minimization of a functional for finite-time costs. approximation using numerical methods and on the
Besides, Linear Matrix Inequalities (LMI) are used in control signal calculus solving a linear equations
order to synthesize the controller. In Chem et al. (1995) system. Simulation results are presented; these results
a controller is proposed only for linear, non-varying show tha advantages of the proposed control strategy
systems which are exponentially stable and of non- which is obtained by using a simple design procedure.
minimum phase, which needs a set of input/output data.
In this work, a strategy based on Taylor’s
In Fujimoto et al. (2001), a perfect-tracking control development of a variable to control is presented, in
based on multirate feedforward control for linear which the control signal is calculated modifying some
systems is presented. Specifically, the design is made terms of this mathematic development to obtain that the
for a SISO system which, nevertheless, can be extended system output follows the reference signal. The main
for a MIMO system. In Young-Hoon (2000), the case in advantage of this methodology is that depending on the
which the dynamic equations of the tracking error are approximation by Taylor formula considered, different
described by a set of non-linear, time-varying, periodic control signal expressions are obtained. Here, the
differential equations, is considered. proposed methodology is applied to a multivariable
In Shuli (2005) a mobile robot controller that is based non-linear system (considering the problem of tracking
trajectory of a mobile robot). Experimental and
on the error model of Kanayama (1990) is proposed. As
a result, instead of just one controller, two are hended, simulation results for this case show the advantages of
which are used depending on whether the angular the proposed methodology. The main contribution of
this article is that the proposed methodology is based
upon easily understandable concepts and there is no (t − nTo)2
need for complex calculations to get the control signal. y(t) = y(nTo) + yɺ(nTo)(t − nTo) + ɺɺy(nTo) +
2!
(3)
The paper is organized as follows: Section 2 states the (t − nTo)3 (t − nTo)m
+ ɺɺɺ
y (nTo) +⋯+ y (nTo)
(m)
+ Rm(t)
problem and presents the methodology proposed to 3! m!
design the controller. Section 3 describes the application
of the strategy to a typical robotic system described by a Where the complementary term Rm(t ) is calculated
multi-variable nonlinear system as the mobile robot as Eq. (4),
model is. Conclusions and suggested future work are
(t − a)m+1
detailed in Section 4. Rm(t) = y(m+1) [ nTo + (t − nTo)ξ ] , 0 < ξ < 1. (4)
(m +1)!
II. STATEMENT OF THE PROBLEM
A non-linear kinematic model for a mobile robot If the time instant t is close enough to nTo, so the
will be used (see Fig. 1), which is represented by Eq. (3) can be expressed as,
Campion (1996), (t − nTo)2
y(t) ≈ y(nTo) + yɺ(nTo)(t − nTo) + ɺɺy(nTo) +
 xɺ = V cos θ 2!
 (5)
 yɺ = V sin θ (1) (t − nTo)3 (t − nTo)m
y (nTo)
ɺɺɺ +⋯+ y(m) (nTo)
. 3! m!
θ = W
where, V is the linear velocity of the mobile robot, W is Similar expressions can be obtained for the state
the angular velocity of the mobile robot, ( x, y ) is the variables x (t ) and θ (t ) . Eq. (5) shows the different
Cartesian position, θ is the orientation of the mobile order derived influence over the state variable in a later
robot, {R} is the inertial frame and {Rc } is the frame instant of time t. In this paper, the value that this
attached to the robot. The values of derivative must have in each sample time, is calculated
x ( t ) , y ( t ) , θ ( t ) , V ( t ) and W ( t ) at discrete time t =
and therefore, the control signals, too. So that the
mobile robot follows the previous established trajectory,
nTo, where To is the sampling period, and as shown in section III.
n ∈ {0,1, 2,3,⋯} will be denoted as xn , yn , θ n , Vn y Wn .
III. CONTROLLER DESIGN
Then, the aim is to find the values of Vn and Wn so that
A first order Taylor approximation of Eq. (2) will be
the mobile robot may follow a pre-established trajectory
first considered, then,
( xd (t ), yd (t ) ) .  xn +1 ≈ xn + xɺnTo

y  yn +1 ≈ yn + yɺ nTo (6)
θ
 n +1 ≈ θ n + θ nTo
ɺ
V W

θ If we desire that the mobile robot goes from its

y Cartesian position ( xn , yn ) to ( xd n +1 , yd n +1 ) , then from
{Rc}
Eq. (6), xɺ and yɺ in time nTo should have the following
values,
x  xd n +1 − xn
x
 xɺn = To
{R}
 (7)
 yɺ = n +1 − yn
yd
Fig. 1 Geometric description of the mobile robot.  n To
The trajectory followed by the mobile robot is
Replacing Eq. (7) in Eq. (1),
described by parametric equations as shown in Eq.(2),
 xd n +1 − xn
 x = x(t )  xɺn = Vd cos θ d =
 To
 y = y (t ) (2)  (8)
θ = θ (t )  yɺ = Vd sin θ d = yd n +1 − yn
  n To
where, Vd, θ d are the linear velocity and orientation
If the function y = y (t ) can be derived until m + 1 angle in instant nTo, respectively, necessaries to make
order, with neighborhood of t = nTo included, so the the mobile robot go from (x n , y n ) to (xd n +1 , yd n +1 ) .
following Taylor’s Formula is valid:
Now, from Eq.(8),
sin θ d yd − yn yd − yn (600, 0)mm and is generated at constant linear and
= tan θ d = n +1 ⇒ θ d = atan n +1 (9)
cos θ d xd n +1 − xn xd n +1 − xn angular velocities respectively known as Vref and
cos θ d  1  xd n +1 − xn  Wref . Fig. 19 shows the system simulation (Koonolige
 sin θ d  Vd = To  yd − y  (10)
K, 1998) on the x-y plane for kv 2 = kw2 = 1 in Eq. (14)
   n +1 n
and Vref = 100mm / sec.
Eq.(10) represents a two equations one unknown
800
(Vd) system, which optimal solution by minimal square
is (Strang G. ,1980), 600

cosθd 1 xdn+1 − xn 
[cosθd sinθd]  Vd = To[ cosθd sinθd] yd − y  (11)
400

y C oordinate [mm.]
sinθ d   n+1 n 
200

D esired Trajectory

xd − x yd − yn
0

Vd = n +1 n cos θ d + n +1 sin θ d (12) -200

To To Simulated Trajectory
-400

In this work, we propose replace θ n+1 in Eq. (6) for -600

θ d given by Eq. (9) and, in this way, calculate the θɺ

-800
-800 -600 -400 -200 0 200 400 600 800
x C oordinate [mm.]

value, this means, Fig. 3: Simulation results: Simulated and Desired

θ d −θn Trajectory, Vref = 100 mm/sec.
θɺ = (13)
To It can also be noticed from this figure that the
mobile robot follows the desired trajectory but in an
From Eqs. (12) and (13) the proposed controller for oscillatory manner. In order to correct this problem, the
mobile robot is as follows, control actions can be calculated by the minimization of
  xd − x yd − y  a quadratic index, in which not only the tracking error
 Vcn = kv 
2 n +1 n
cos θ d + n +1 n sin θ d 
  To To  (14) but also the square of state variables derivatives have
 been considered. Thus, the state variables variation is
Wc = θ d − θn
kw2 minimizing as well as the error between the real and
 n
To desired trajectory,
J = k12 ( xdn +1 − xn +1 )2 + ( ydn+1 − yn +1 )2  + k22 ( xɺn 2 + yɺn2 ) +
where the constants kv2, kw2 allow adjusting the system
behavior and they satisfy 0 < kv 2 ≤ 1 , and 0 < kw2 ≤ 1 . (15)
+k32 (θ dn+1 − θ n+1 )2 + k42θɺn
Simulation and experimental studies were carried ∂J
out with a mobile robot PIONEER 2DX available at the = k12 [ −2To cosθn (xdn+1 − xn −ToVn cosθn ) −
Instituto de Automática (INAUT) to test the proposed ∂V n (16)
controller performance. The simulation software −2To sinθn ( ydn+1 − yn −ToVn sinθn )] + k2 2Vn = 02

SAPHIRA of Active Media was also used (Konoologe

∂J
K, 1998). Fig. 2 shows the Pioneer 2DX and the = −2Tok32 (θ d n +1 − θ n − ToWn ) + 2k42Wn = 0 (17)
laboratory facilities where the experiences were carried ∂W n
out. In the PIONEER 2DX the value of the sample time
To is 0.1 sec. From Eqs (16) y (17),
 k12  ∆x ∆y 
Vn = 2 
cos θ n + sin θ n 
 k  To To 
k12 + 2 2
  To 
 kv 2
 (18)
W = k32 ∆θ
 n k 2 To
 k32 + 4 2
  To
 kw2

Fig. 2. Pioneer 2DX mobile robot and its environment where, ∆x = xd n +1 − xn , ∆y = yd n +1 − yn .

In order to test the performance of the proposed If Eqs. (14) and (18) are compared, then can be seen
controller, a circumference of 600 mm. radius was used that, to minimize the state variables variations, the
as a desired trajectory, with center on the origin of the values of kv2 y kw2 should be chosen less than one. For
coordinate system. The starting point for the robot was this reason, we propose to reduce the values kv2 and kw2
the center of the circumference, and an initial to values kv2 = 0.2 and kw2 = 0.2.
orientation θ = 0º . From this starting point it evolves to
In Fig. 4 the real trajectory of the mobile robot
the desired trajectory. The reference trajectory starts at
PIONEER 2DX in the x-y plane is shown. On the other
hand, Figs. 5 and 6 show the time evolution of the x,y angular velocities, denoted as Vreal and Wreal,
coordinates. respectively, of the mobile robot and the control actions.
As it can be observed, the linear and angular velocities
Fig. 4 shows the mobile robot following the of the mobile robot stay very close to the reference
reference trajectory without making undesirable values generated by the desired trajectory.
oscillations. Figs. 5 and 6 show the time evolution of
800
coordinates x and y of the mobile robot while
navigating, respectively. From Figs. 5 and 6, it can be 600

noted that the mobile robot reaches the desired 400

D esired Trajectory
trajectory very quickly, and it follows this trajectory

y C oordinate [mm.]
200

with an error smaller than 10 mm. In Figs. 4, 5 and 6, it 0

can also be noted that the robot arrives at the end of the
-200
reference trajectory and it remains in that position Real Trajectory

without oscillations. -400

-600
800

-800
600 -800 -600 -400 -200 0 200 400 600 800
x C oordinate [mm.]
400
Fig.7: Experimental results: Real and Desired Trajectory,
y C oordinate [mm.]

200 D esired Trajectory Vref = 200 mm/sec., Wref = 19.1 deg/sec., kv2 = 0.2, kw2 =
0 0.2.
1200
-200
Real Trajectory
1000

Vreal - Control Action Vcn [mm/sec]

-400

-600 800

-800 Control Action Vcn

-800 -600 -400 -200 0 200 400 600 800 600
x C oordinate [mm.]

Fig. 4: Experimental result: Real and Desired Trajectory, 400 Vreal

Vref = 100 mm/sec.
800
200

600
xd (t)
0

400
xd(t) [mm.] - x(t)[mm.]

x(t) -200
0 20 40 60 80 100 120
200 Time [sec.]

0 Fig. 8: Experimental results: Real Linear velocity and

Control Action Vcn , Vref = 200mm /sec.
-200

25
-400 Control Action Wcn
Wreal - Control Action Wcn [deg/sec]

20
-600
0 10 20 30 40 50 60 70 80 90 100
Time [sec.]

Fig. 5: Experimental results: Evolution of x(t), Vref = 100 15 Wreal

mm/sec.
800 10

600 yd (t)
5
yd(t) [mm.] - y(t) [mm.]

400
y(t) 0
200

0 -5
0 20 40 60 80 100 120
Time [sec.]
-200

Fig. 9: Experimental results: Real angular velocity and

Wcn , Wref = 19.1deg/ sec.
-400

-600
0 10 20 30 40 50 60 70 80 90 100
Time [sec.]
If a second order Taylor approximation is used, so
Fig. 6: Experimental results: Evolution of y(t), Vref = 100
mm/sec.  To 2
 xn +1 ≈ xn + xɺnTo + ɺɺxn
 2
Fig. 7 shows experimental results obtained with the
 To 2
mobile robot PIONEER 2DX when the navigation  yn +1 ≈ yn + yɺ nTo + ɺɺ
yn (19)
velocity is increased to Vref = 200 mm/sec. and Wref =  2
19.1 deg/sec. It can be seen from Fig. 7 that the mobile  To 2
robot follows the desired trajectory with a maximum θ n +1 ≈ θ n + θɺnTo + θɺɺn
 2
error of 10 mm, which is very small when compared
with the distance between wheel axes (330 mm). Figs. 8 By derivation of Eq. (19), we obtain,
and 9 show the time evolution of the real linear and
  xɺn +1 ≈ xɺn + ɺɺ
xnTo In this way, the linear velocity variation in the
 interval  nTo ( n + 1) To  will approximately have the
 yɺ n +1 ≈ yɺ n + ɺɺ
ynTo (20)
θɺ ≈ θɺn + θɺɺnTo shape given in Fig. 10.
 n +1

Then, an equations system can be built with two As it is posible to use only one control signal so that
equations and two unknown for each variable x,y, it stays constant between each sampling period, we
propose to define a new control signal whose value is
To To 2   xɺ   xd − x  defined by,
 2   n  =  n +1 n  (21)
 1 To   ɺɺ
xn   xd
ɺ n +1  Vnew = C1Vn + (1 − C1)(Vn + VɺnTo) ; 0 ≤ C1 ≤ 1 (35)

To To 2   yɺ n   yd n +1 − yn  In a similar way, for the Wn we have,

 2  =  (22)
 1 To   ɺɺ
yn   yd ɺ n +1  2
(θ d n +1 − θ n ) − θɺd n +1
Wn = (36)
obtaining, To
2 2
2
xɺn =
( xd n +1 − xn ) − xd
ɺ n +1 (23) Wɺn = θɺd n +1 − 2 (θ d n +1 − θ n ) (37)
To To To
2
yɺ n = ( yd n +1 − yn ) − yd
ɺ n +1 (24) And as Eq. (35), the expression obtained is
To Wnew = C 2Wn + (1 − C 2 ) (Wn + WɺnTo ) ; 0 ≤ C 2 ≤ 1 (38)
2 2
xn =
ɺɺ ɺ n +1 − 2 ( xd n +1 − xn )
xd (25) V(t)
To To
2 2 V (t ) = Vn + Vɺn ( t − nTo )
yn =
ɺɺ ɺ n +1 − 2 ( yd n +1 − yn )
yd (26)
To To Vn α
where, Eqs. (23) to (26) represent the values that first
and second derivatives of x and y variables should have, tan α = Vɺn
to make the mobile robot go from (x n , y n ) to
(xd n +1 , yd n +1 ) . From Eq. (1), t
2 (n + 1)To
xɺn = Vn cos θ n = ( xd n +1 − xn ) − xd
ɺ n +1 (27) nTo
To Fig. 10 Interpretación grafica de la variación de V(t)
2
yɺ n = Vn sin θ n = ( yd n +1 − yn ) − yd
ɺ n +1 (28) In Fig. 11 x-y plane robot trajectory is shown when a
To
reference trajectory is a 600 mm radius circumference,
then, 800

2 2
2  2  600

Vn =  (xdn+1 − xn ) − xd ɺ n+1  +  ( ydn+1 − yn ) − yd ɺ n+1  (29)

 To   To 
400
y C oordinate [mm.]

200
2
( yd n +1 − yn ) − yd
D esired Trajectory
ɺ n +1
sin θ n 0

= tan θ n = To (30)
cos θ n 2
( xd n +1 − xn ) − xd ɺ n +1
-200
Real Trajectory

To -400

2 2 -600

xn = Vɺn cosθn −Vnθɺn sinθn = xd

ɺɺ ɺ n+1 − 2 ( xdn+1 − xn ) (31)
To To -800
-800 -600 -400 -200 0 200 400 600 800
x C oordinate [mm.]
2 2 Fig. 11: Experimental results: Real and Desired
ɺɺyn = Vɺn sinθ n + Vnθɺn cosθ n = yd ɺ n +1 − 2 ( ydn+1 − yn ) (32)
To To Trajectory, Vref = 200 mm/sec., Wref = 19.1 deg/sec., C1 = 1,
C2 = 1.
 2 2 ɺ sin θ 
xd
ɺ − ( xd − x ) + V θ 1400

cos θ n  ɺ  To n +1 To 2 n +1 n n n n 

 sin θ  Vn =  2  1200

 n   yd ɺ n +1 − 2 ( yd n +1 − yn ) − Vnθɺn cos θ n 
2 1000
D esired Trajectory

 To To  800
y C oordinate

(33) 600

 2 2 
Vɺn =  xd ɺ n +1 − 2 ( xd n +1 − xn ) + Vnθɺn sin θ n  cos θ n + 400
Real Trajectory

 To To  200

2 2 
+  yd ɺ n +1 − 2 ( yd n +1 − yn ) − Vnθɺn cos θ n  sin θ n
0

 To To  -200
-200 0 200 400 600 800 1000 1200 1400
x C oordinate
(34) Fig. 12: Experimental results: Real and Desired
Trajectory, Vref = 200 mm/sec.
 If Figs. 7 and 11 are compared, it can be seen that
the performance of the controllers is very similar, and to Future work will entail the generalization of this
remark the advantage of the use of Eqs. (35) and (38), a methodology to cases where the states cannot be
trajectory of a 1200 mm side square generated by a measured and, consequently, observers are needed.
constant linear velocity reference (Vref = 200 mm/sec.), ACKNOWLEDGMENT
will be used. This reference represents a very hard and
demanding trajectory. In Fig. 12 the trajectory followed This work was partially funded by the Consejo Nacional de
by the mobile robot using Eq. (18) for the controller Investigaciones Científicas y Técnicas (CONICET - National
Council for Scientific Research), Argentina.
design, is shown. On the other hand, from Fig. 13, it can
be seen the trajectory followed by the mobile robot REFERENCES
when the control action is calculate using Eqs. (35) and
Campion G., Bastin G., d’Andrea-Novel B. (1996), “Structural
(38), with C1 = 0.5, C2 = 0.8. In Fig. 14, the control Properties and Clasification of Kinematic and Dynamic
signals and the evolution of the linear and angular Models of Wheeled Mobile Robots”, IEEE Transact.
velocities depending on the time instant are shown. It Robotics and Automation, vol. 12, No 1, pages 47-62.
can be observed how the robot linear velocity decreases Chen M. and Chiou F. (1995), “Data-Based Tracking Control
when it arrives to the square corner. for Nonminimun-Phase Systems”, Proc. American
Control Conference, Seatle, Washington.
1400
Fujimoto H., Hori Y. and Kawamura A. (2001), “Perfect
1200 Tracking Control Based on Multirate Feedforward with
1000 Generalizad Sampling Periods”, IEEE Transact. Industrial
D e sired Trajectory Electronics, vol. 48, No 3, pp. 636-644.
y C oordinate [mm.]

800
Kanayama Y, Kimura Y., Miyazaki F., Noguchi T. (1990), “A
600
Stable Tracking Control Method for an Autonomous
400 Mobile Robot”, Proc. IEEE Int. Conf. Robotics and
200
Real Trajectory Automation, 1990, 384-389.
Kim K. (2003), “Receding horizon tracking control for
0
constrained linear continuous time-varying systems”, IEE
-200
-200 0 200 400 600 800 1000 1200 1400
Proc. Control Theory and Applications, vol. 150, Issue: 5,
x C orrdinate [mm.]
ISSN: 1350-2379. Onpage(s): 534-538.
Fig. 13: Experimental results: Real and Desired Konolige K. (1998), SAPHIRA MANUAL VERSION 6.1.
Trajectory, Vref = 200 mm/sec., C1 = 0.5, C2 = 0.8. Morari M. and Zafiriou E. (1989), Robust Process Control,
120

Prentice Hall Englewood Cliffs, New Jersey.

Wreal - Control Action Wnew [deg/sec]

350
Vreal - Control Action Vnew [mm/sec]

100
300

Vnew Eq. 35
80 Rosales A., Scaglia G., Mut V., di Sciascio V. (2006),
250 60
“Controller Designed by Means of Numeric Methods for
a Benchmark Problem: RTAC (Rotational Translational
200 40

20
150

0
Actuator)”. CERMA2006, IEEE Computer Society,
100
Vreal -20
Wreal
Cuernavaca, Morelos, Mexico.
50
-40
Control Action Wnew Eq. 38 Scaglia G., Postigo J.F., Mut V., Kuchen B. (2005), “A
0
0 10 20 30
Time [sec.]
40 50 60
-60
0 10 20 30
Time [sec]
40 50 60
method for paths tracking in mobile robots”, XI Reunión
Fig. 14:Experimental results: a) Real Linear velocity and de Trabajo en Procesamiento de la Información y Control
Control Action Vnew , b) Real angular velocity and Wnew . (RPIC), 21 al 23 de septiembre de 2005.
Scaglia G., Quintero O., Mut V., Di Sciascio F. (2006a),
It can be observed here, that when a higher order “Paths tracking in mobile robots using a trapezoidal
integration method”, XX Congreso Argentino de Control
approximation is used, a better system response is Automatico – AADECA 2006,Buenos Aires - Argentina.
obtained, when an abrupt change in the trajectory Scaglia G., Mut V., Quintero O., Di Sciascio F. (2006b),
orientation exists. “Diseño de un controlador basado en métodos
algebraicos: Aplicación a robótica móvil”. XII Latin-
IV. CONCLUSIONS
American Congress on Automatic Control, CLCA2006,
In this work, a strategy to find the control actions to Salvador, Bahia, Brazil.
make the system follow a pre-established reference Scaglia G., Mut V., Rosales A., Quintero O., "Tracking
trajectory is proposed. For this purpose, the Taylor’s Control of a Mobile Robot using Linear Interpolation",
Proc. 3rd Inter. Conf. on Integrated Modeling and
development of the state of the system is made. This
Analysis in Applied Control and Automation, IMAACA
methodology is simple, and can be applied to different 2007, vol. 1, pp. 11-15, ISBN: 978-2-9520712-7-7
kinds of systems, and it is not necessary to make Shuli S., (2005), “Designing approach on trajectory-tracking
complex calculus to compute the control actions. As the control of mobile robot”, Robotics and Computer-
system approximation order using Taylor’s series is Integrated Manufacturing, vol. 21, Issue 1, pages 81-85.
greater, different characteristics of the control signal can Strang G. (1980), Linear Algebra and Its Applications,
be obtained. In this way, it can be obtained, not only the Academic Press, New York, 1980.
amplitude but also the first derivative (slope), second Young-Hoon Kim, In-Joongha Ha. (2000), “Asymptotic state
derivative and so on, of the control action in a time tracking in a class of nonlinear systems via learning-
based inversion”, IEEE Transact. Automatic Control,
instant nTo. This can be used to improve the system
On page(s): 2011 – 2027, vol. 45.
behaviour when the trajectories have abrupt variations.
e_mail : [email protected]
[email protected]

View publication stats