www.ietdl.

org
Published in IET Control Theory and Applications
Received on 13th October 2010
Revised on 12th April 2011
doi: 10.1049/iet-cta.2010.0591

ISSN 1751-8644

Brief Paper
Output feedback controller design of a unicycle-type
mobile robot with delayed measurements
E.-H. Guechi J. Lauber M. Dambrine M. Defoort
UVHC-LAMIH, CNRS UMR 8201, Univ Lille Nord de France Le Mont Houy, F59313 Valenciennes, France
E-mail: [email protected]

Abstract: This study deals with the trajectory tracking controller design of a wheeled mobile robot subject to delayed
measurements. Using the atness properties of the unicycle-type mobile robot, an output feedback approach based on a
non-linear predictor to estimate the state variables is proposed. It enables one to guarantee the exponential stability of
the tracking errors. A nite-time observer based on sliding mode theory is applied for the estimation of the delay whose
value is assumed to be constant but unknown. The stability property of the closed-loop system with delayed outputs is
studied using the Lyapunov theory. Numerical simulations in a realistic environment show the efciency of the proposed
approach.

Introduction

This paper deals with the design of an output feedback

controller with delayed measurements to solve the trajectory
tracking problem for a wheeled mobile robot. One difculty
for the control of a car-like robot arises from the
so-called non-holonomic constraints imposed by the rolling
wheels [1].
Obstacles to the trajectory tracking issue for such systems
are the uncontrollability of their linear approximation and
the fact that the Brocketts necessary condition to the
existence of a smooth time-invariant state feedback is
not satised [2]. To overcome those difculties, various
methods have been investigated: homogeneous and timevarying feedbacks [3, 4], dynamic linearisation feedback [5],
sinusoidal and polynomial controls [6], piecewise continuous
controls [7], atness-based controls [8] and sliding mode
control [9, 10], for instance. In [11], the authors present
a parallel distributed compensation (PDC) controller to
solve the trajectory tracking problem. This technique is
based on the TakagiSugeno fuzzy model and linear matrix
inequalities formalism [1214]. In [15, 16], a non-linear
control law combining feedforward and feedback control
actions is proposed.
Besides the non-holonomic constraints of a wheeled
mobile robot, time delay is another important factor that may
pose signicant challenges in designing feedback controller.
Actually, delayed measurements usually occur because of
the presence of a vision system (especially because of the
image processing) such as in robot soccer game, space
exploration and so on. If not properly dealt with, these
delays may produce a signicant degradation in the tracking
726
The Institution of Engineering and Technology 2012

performance or even lead to loss of stability. However, most

of the existing literature does not take such a problem into
account in order to solve the trajectory tracking problem for
a non-holonomic mobile robot.
In this paper, we propose a control design method,
based on the atness properties of the system, for tracking
controllers of mobile robot with delayed measurements.
A non-linear predictor observer is introduced in order to
estimate the state variables using the delayed measurements.
A nite-time observer based on sliding mode [17] is also
added to obtain the estimation of the unknown delay. The
closed-loop stability of the combined observercontroller is
proved using the Lyapunov theory.
The remainder of this paper is organised as follows.
Section 2 presents the unicycle-type model and states the
tracking problem. The output feedback controller is designed
in Section 3. A proof of the exponential stability of the
whole closed-loop system is given. Finally, simulation
results are provided in the last section to demonstrate the
efciency of the proposed approach.

2
2.1

Problem statement
Kinematic model

The mobile robot, shown in Fig. 1, is of unicycle type.

It has two driving wheels mounted on the same axis and
independently controlled by two actuators (DC motors). This
robot is described by a vector of generalised coordinates q
constituted by the coordinates (x, y) of the midpoint between
the two driving wheels and by the orientation angle with
IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733
doi: 10.1049/iet-cta.2010.0591

www.ietdl.org
Let us dene the tracking errors of the robot as

ex (t) = x(t) xr (t)

e (t) = y(t) yr (t)
y
e (t) = (t) r (t)

Fig. 1

The control purpose is to design a controller such that

the vehicle (2) is constrained to asymptotically track the
desired trajectory (4) from some initial tracking errors. In
this paper, the posture of the robot is obtained through a
vision system which introduces a delay . Hence, the goal
is to asymptotically stabilise the zero equilibrium of the
closed-loop system

Unicycle-type mobile robot

e x (t) = v(t) cos( (t)) vr (t) cos(r (t))

e (t) = v(t) sin( (t)) vr (t) sin(r (t))
y
e (t) = w(t) wr (t)

respect to a xed Cartesian frame

q(t) = [x(t), y(t), (t)]

(1)

Under the pure rolling and non-slipping conditions, the ideal

kinematic equations are

x (t) = v(t) cos( (t))

y (t) = v(t) sin( (t))

(t) = w(t)

(6)

using the measured output

q(t ) = [x(t ), y(t ), (t )]T

(7)

(2)

where the control inputs of the mobile robot v(t) and w(t)
are, respectively, the linear and the angular velocities of
the robot.
Remark 1: System (2) is differentially at (see [8] for details
about atness) since all inputs and state variables can be
differentially parameterised by x(t), y(t) as well as a nite
number of their time derivatives, that is



y (t)

(t)
=
arctan

x (t)


2
(3)
v(t) = x (t) + y 2 (t)

w(t) = y (t)x(t) y (t)x(t)

x 2 (t) + y 2 (t)

It should be highlighted that, using the atness property, the

tracking error e converges to zero when ex , ey and their rst
time derivatives are also converging to zero.

Output feedback controller design

In this section, the output feedback controller is described

to achieve the stabilisation of the tracking errors system.
Owing to the presence of the delay because of the vision
system, a non-linear predictor observer is used to compute
the estimated instantaneous state q (t) = [x(t), y (t), (t)]T . A
sliding mode observer is also designed in order to give
an estimate (t) of the actual delay . The whole control
scheme is depicted in Fig. 2. The corresponding blocks will
be described hereafter.
3.1

2.2

(5)

Online delay estimation

Control objective

In order to formulate the trajectory tracking problem, let

us dene the reference trajectory qr (t) = [xr (t), yr (t), r (t)]T
generated by the following virtual robot

x r (t) = vr (t) cos(r (t))

y (t) = vr (t) sin(r (t))
r
r (t) = wr (t)

(4)

where vr (t) and wr (t) are the linear and angular velocities
of the virtual robot, respectively. In this paper, it is assumed
that the reference trajectory, its rst and second time
derivatives are bounded.
Remark 2: A recent algorithm based on a receding horizon
planner enables the real-time generation of the optimal
admissible trajectory which takes into account the physical
capacities of the robot and the environment (see [18, 19] for
further details).
IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733
doi: 10.1049/iet-cta.2010.0591

Since a delay occurs on the measured output given by (7),

a non-linear predictor observer is used in order to estimate
the state q(t). Nevertheless, the delay , which is assumed
to be constant, is not known by advance. That is why,
in this paper, it is proposed to apply a nite-time delay
identication using a variable structure observer.
The sliding mode control methodology is a robust
technique to control non-linear systems operating under
uncertainty conditions [2022]. It allows one to design
sliding mode observers [2325]. The developed algorithms
based on sliding modes have extremely robust behaviour
with respect to the heated material parameter variations and
external disturbances (sliding and skidding for instance).
Moreover, the nite time convergence property of sliding
mode observers is useful in observation problems such
as online parameter identication. The proposed online
delay identication algorithm is based on the sliding mode
observer described in [17]. Using the third equation of
model (2), the unknown delay can be estimated using an
appropriate sliding surface.
727
The Institution of Engineering and Technology 2012

www.ietdl.org

Fig. 2

Whole closed-loop system

Let us dene the sliding surface as follows

(t) = w(t ) w(t (t))

(8)

It is assumed that the robot angular acceleration is bounded

between two positive constants 1 , 2 as follows

2 ,
1 |w(t)|

t with 1 , 2 > 0

(9)

Here, the objective is to stabilise (t) towards zero in nite

time. Assuming that the signal w(t) is not time periodic, this
implies the convergence in nite time of the estimated delay

(t) towards the actual delay .
Lemma 1: Under the previous assumptions, using the
following sliding mode observer
(t) = G w(t
(t))sign((t))
with

if  > 0
1
sign() = 0
if  = 0

1 if  < 0

(10)

(11)

it is ensured the nite time convergence of the estimated

delay (t) towards if the gain satises G > 22 /12 .
Proof: From (8), the derivative of the sliding surface (t)
is given as follows:

(t)
= w(t
) w(t
(t)) + (t)w(t
(t))

(12)

Let us consider the following candidate quadratic Lyapunov

function
1
V (t) = (t)2
(13)
2
From assumption (9) the following relationships hold
1 |w(t
(t))| 2

and |w(t
)| 2

(14)

Hence, the derivative of the Lyapunov function is bounded

by
V (t) (22 G12 )|(t)| V
1/2 (t)
(15)
with > 0 if the gain satises G > 22 /12 . This implies that
 converges towards zero in nite time and consequently,
the convergence in nite time of the estimated delay (t)
towards the actual delay .

728
The Institution of Engineering and Technology 2012

Remark 3: One should highlight that w(t

) could be
obtained from w(t ) using an exact robust differentiator
via sliding mode techniques [26].
Remark 4: Condition (9) could be relaxed. Indeed, the
proposed observer works well even if w(t)

crosses zero
during a small interval of time. In such a case, the proposed
observer keeps the last estimated value of the delay. Then,
when w(t)

becomes non-singular, the proposed sliding mode

observer (10) is applied.
Remark 5: In the case that w(t)

= 0, on can replace the

sliding surface dened in (8) by the following one (t) =
v(t ) v(t (t)). Hence, one can derive similar results
in the case that the robot drives straight, without change in
angular speed.
3.2

Output feedback controller design

In order to deal with the delayed measurements, the

following non-linear predictor observer is used to compute
the estimated state:

d x (t)

L1 (x(t ) x (t (t)))

= v(t) cos (t)

dt

d y (t)
L2 (y(t ) y (t (t)))
= v(t) sin (t)

dt

d (t) = w(t) L3 ( (t ) (t
(t)))
dt
(16)

where x (t), y (t) and (t)

are, respectively, the estimates
of x(t), y(t) and (t). The observer gains are L1 , L2 and
L3 . The unknown time delay , estimated by the sliding
mode observer (10), is assumed to be bounded by a known
constant, that is, max .
System (6) is not exactly linearisable by means of a static
state feedback. It can be seen that the relative degree of the
tracking errors ex and ey are not well dened (ex and e y
already depend on v but not on w). This fact points out the
need for a rst-order dynamic extension of the control input
v in order to exactly linearise system (6). Differentiating
twice x and y, one obtain

x (t)
cos( (t)) sin( (t))
v (t)
=
(17)
y (t)
sin( (t)) cos( (t))
v(t)w(t)
Since (t) is not known, let us use its estimated value (t)
in order to design the classical feedback linearising control
IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733
doi: 10.1049/iet-cta.2010.0591

www.ietdl.org
laws u1 (t) and u2 (t) as follows

u1 (t) = v (t) cos( (t)) v(t)w(t) sin( (t))
u2 (t) = v (t) sin( (t)) + v(t)w(t) cos( (t))

0
0
0
0
k (t) k (t) k (t) k (t)
2 1
1 2
2 2

1 1
A(t) =

0
0
0

0
k1 2 (t) k2 2 (t) k1 1 (t)
k2 1 (t)

1 (t) = 1 + cos((t))

2 (t) = sin((t))

(t) = (t)x (t) (1 + (t))(k x (t) + k x (t))

2
1
r
1
1
2
(t))

(t)
+
k
y

(t)
+
k
y
+
(t)(
y
r
1
2
2

4 (t) = 1 (t)yr (t) (1 + 1 (t))(k1 y (t) + k2 y (t))

2 (t)(xr (t) + k1 x (t) + k2 x (t))

(18)

The dynamic linearising controller (18) has a potential

singularity at v = 0, that is, when the unicycle is not rolling.
The occurrence of such singularity in the dynamic extension
process is structural for non-holonomic systems and may
only happen during the transient period. This problem is
avoided by suitably choosing the initial state of the dynamic
compensator and designing the control input u1 (t).
The output feedback control laws u1 (t) and u2 (t) are
chosen as follows

u1 = x r + k1 (xr x ) + k2 (xr x )
(19)
u2 = y r + k1 (yr y ) + k2 (yr y )

Hence, the whole dynamics of the closed-loop system, given

by (16), (23), is described by
X (t) = (A + A(t))X (t) + (t)

where k1 and k2 are positive real numbers.

Hereafter, it will be proved that under the controller (19),
the tracking errors system is exponentially stable.
3.3

Closed-loop stability

It is well known that the separation principle does not hold

for non-linear systems. This means that the observer and
the controller cannot be designed separately. Thus, let us
analyse the combined observercontroller stability.
Using (18), system (17) could be written as

x (t)
cos( (t) (t)) sin( (t) (t)) u1 (t)
=
y (t)
sin( (t) (t)) cos( (t) (t)) u2 (t)
(20)
Let us dene the observation error as follows

x (t) = x(t) x (t)

y (t) = y(t) y (t)

(t) = (t) (t)

(21)

Hence, the tracking errors system is as follows

e x (t)
k1 ex (t) + k2 e x (t)
cos( (t)) sin((t))
=
+

e y (t)
k1 ey (t) + k2 e y (t)
sin( (t)) cos((t))


k1 x (t) k2 x (t)

k1 y (t) k2 y (t)


1 + cos( (t))
sin( (t))
+
sin( (t))
1 + cos( (t))

x r (t) + k1 ex (t) + k2 e x (t)

(22)
y r (t) + k1 ey (t) + k2 e y (t)
which can be rewritten on the form of a perturbed linear
system
X (t) = AX (t) + A(t)X (t) + (t)
(23)
with
X (t) = [ex (t), e x (t), ey (t), e y (t)]T ,

0
0 1 0 0
(t)
0 0
k k
2
A= 1 2
, (t) =

0 0 0 1
0
0 0 k1 k 2
(t)
4

IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733

doi: 10.1049/iet-cta.2010.0591

with

(24)

x (t) = e1 (t) + L1 (x(t ) x (t (t)))

y (t) = e2 (t) + L2 (y(t ) y (t (t)))

(26)

= L ( (t ) (t
(t)))
(t)
3

(27)

(25)

e1 (t) = v(t)(cos (t) cos (t))

e2 (t) = v(t)(sin (t) sin (t))

Lemma 2: Under the practical assumption that the linear

velocity is bounded, if the observer gains are chosen such
that Li max < /2 for i {1, 2, 3}, using the online delay
estimator (10), the output feedback controllers (18) and
(19) ensure that the whole closed-loop system (24)(27) is
exponentially stable.
Proof: According to Lemma 1, the estimated delay (t)
converges in nite time towards the actual delay using
the online delay estimator (10). Hence, after a nite time,
one obtains
X (t) = (A + A(t))X (t) + (t)
x (t) = e1 (t) + L1 x (t )
y (t) = e2 (t) + L2 y (t )
= L (t
)
(t)
3

(28)
(29)
(30)
(31)

Since the unknown delay is bounded by a known constant

max , the linear velocity is bounded by a known constant
vmax and the gains Li are chosen such that Li max /2
for i {1, 2, 3}, then the estimation errors asymptotically

converge towards zero. Indeed, from (31), (t)

converges
exponentially to zero as it may be proved easily using
Nyquists stability criterion. Since the sine and cosine
functions are both contractive, the following inequalities
hold


|e1 (t)| |(t)||v(t)|

|(t)|v
max

|e2 (t)| |(t)||v(t)|

|(t)|v
max
Therefore (29) and (30) may be seen as linear time-delay
systems which are exponentially stable with exponentially
vanishing inputs ei (t); this proves the exponential stability
of x (t) and y (t).

towards
According to the exponential convergence of (t)
zero and the denitions of A(t) and (t), one can easily
729
The Institution of Engineering and Technology 2012

www.ietdl.org
derive that these terms converge also exponentially towards
zero. That is to say, there are positive real numbers C1 , C2 ,
such that, for all t t0 , one obtains

from which one can derive the following inequalities

t
X (t) 
(t, t0 )X (t0 ) + 
(t, h)(h) dh
t0

A(t) < C1 et

Ce (tt0 ) X (t0 ) + CC2

(t) C2 et

X (t) Ce (tt0 ) X (t0 ) + CC2 e t

For positive values of k1 and k2 , the matrix A is Hurwitz and

consequently, from the perturbation theory, the solution X (t)
converges exponentially towards zero, that is, there are two
positive constants C and such that 
(t, t0 ) < Ce (tt0 ) .
The general solution of (28) is given by
t
(32)

t0

e( )t e( )t0

(34)

This proves that X (t) converges exponentially towards zero.

Simulation results

In this section, simulations are carried out to evaluate the

performance of the proposed output feedback controller.

Fig. 3

Trajectory tracking with delayed and noise measurements: actual and reference trajectories of the robot

Fig. 4

Evolution of the tracking errors in position and orientation

730
The Institution of Engineering and Technology 2012

(33)

Assuming (without loss of generality) that  = , the

following bound is obtained

X (t) = (A + A(t))X (t)

(t, h)(h) dh

e (th) eh dh

t0

Let
(t, t0 ) be the transition matrix of the system

X (t) =
(t, t0 )X (t0 ) +

t

IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733

doi: 10.1049/iet-cta.2010.0591

www.ietdl.org

Fig. 5

Evolution of the actual and reference linear velocities

Fig. 6

Evolution of the tracking errors in linear velocity

Fig. 7

Evolution of the actual and reference angular velocities

IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733

doi: 10.1049/iet-cta.2010.0591

731
The Institution of Engineering and Technology 2012

www.ietdl.org

Fig. 8

Evolution of the tracking errors in angular velocity

Fig. 9

Convergence of the estimated delay towards the actual delay using the sliding mode observer

Note that in order to obtain realistic results, a zero-mean

Gaussian noise with a variance of 0.001 and an upper bound
of 0.02 m (0.02 rad respectively) has been added in each
output measurement (i.e. x, y and ).
The reference trajectory is generated using the reference
model (2). Here, it is a combination of Bezier curves and is
depicted in Fig. 3. Owing to the presence of a vision system
in many practical applications, an unknown constant delay
is introduced in the measurements (here = 0.1 s). To test
the robustness of the proposed strategy, some initial errors
of position and heading direction are assumed to exist, that
is, ex = 0.2 m, ey = 0 m and e = 45 .
By means of a pole placement, one can design the
parameters of the proposed controller, that is, k1 = 150 and
k2 = 25. The initial value of the online delay estimation
algorithm is arbitrarily set to (0) = 0.04 s. The gain G
is tuned according to Lemma 1 and Remark 4 in order to
obtain a fast enough estimation of delay . Assuming that a
bound of the delay is a priori known (i.e. max = 0.15 s), one
can apply Lemma 2 to derive the observer gains using the
732
The Institution of Engineering and Technology 2012

condition Li max < /2. Here, they are chosen as follows:

L1 = 10, L2 = 10 and L3 = 10.
In the simulations, the control objective is the online delay
estimation and the stabilisation of the tracking errors in spite
of the presence of noise and delayed measurements.
Fig. 3 shows that the actual trajectory of the robot (in
dashed line) converges towards the reference trajectory (in
solid line). As we can see in Fig. 4, the tracking errors
(ex , ey and e ) tend to zero in spite of measurement noise.
Therefore one can conclude that the proposed controller
efciently corrects the undesirable deviation because of the
delay measurement. It manages to accomplish the trajectory
tracking task with satisfactory accuracy. The applied linear
and angular velocities of the robot are depicted in Figs. 5
and 7. One can see that the tracking errors in velocities are
low in Figs. 6 and 8. In Fig. 9, it is highlighted that the
estimated delay converges towards the actual delay. Fig. 10
depicts the convergence phase of the online delay estimator.
One can see that the sliding mode observer enables a fast
and robust estimation of the delay, that is, in about 0.25 s.
IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733
doi: 10.1049/iet-cta.2010.0591

www.ietdl.org

Fig. 10

Convergence phase of the online delay estimator

Conclusion

This paper proposes a new approach for the trajectory

tracking of a non-holonomic mobile robot subject to delayed
measurements. The delay, because of the vision system is
unknown and assumed to be constant. A nite time observer
based on sliding mode enables its estimation. Using the
atness properties of the unicycle-type mobile robot, an
output feedback approach that uses a non-linear predictor
to estimate the state variables is derived. The exponential
stability of the tracking errors is proved. Numerical
simulations show the efciency of the proposed approach.

Acknowledgments

This work is supported by the International Campus on

Safety and Intermodality in Transportation, the European
Community, the Ministre de lEnseignement suprieur et
de la Recherche, the Rgion Nord - Pas de Calais and the
Centre National de la Recherche Scientique. The authors
gratefully acknowledge the support of these institutions.

7
1
2

3
4
5
6
7
8

References
Laumond, J.-P.: Robot motion planning and control (SpringerVerlag, 1998)
Brockett, R.W.: Asymptotic stability and feedback stabilization, in
Brockett, R.W., Millman, R.S., Sussmann, H.J. (Eds.): Differential
geometric control theory (Birkhauser, Boston, MA, 1983), pp. 181
195
Pomet, J.: Explicit design of time-varying stabilizing control laws for
a class of controllable systems without drift, Syst. Control Lett., 1992,
18, (2), pp. 147158
Samson, C.: Control of chained systems: application to path following
and time-varying point-stabilization of mobile robots, IEEE Trans.
Autom. Control, 1995, 40, pp. 6477
Oriol, G., De Luca, A., Vendittelli, M.: WMR control dynamic
feedback linearization: design, implementation, and experimental
validation, IEEE Trans. Control Technol., 2002, 10, (6), pp. 835852
Murray, R., Sastry, S.: Nonholonomic motion planning: steering using
sinusoids, IEEE Trans. Autom. Control, 1993, 38, (5), pp. 700716
Hespanha, J.P., Morse, A.S.: Stabilization of nonholonomic
integrators via logic-based switching, Automatica, 1999, 35, (3),
pp. 385393
Fliess, M., Levine, J., Martin, P., Rouchon, P.: Flatness and defect of
nonlinear systems: introductory theory and examples, Int. J. Control,
1995, 61, (6), pp. 13271361

IET Control Theory Appl., 2012, Vol. 6, Iss. 5, pp. 726733

doi: 10.1049/iet-cta.2010.0591

9
10
11

12
13
14
15

16
17
18
19
20
21
22
23
24
25
26

Defoort, M., Floquet, T., Kokosy, A., Perruquetti, W.: Sliding-mode

formation control for cooperative autonomous mobile robots, IEEE
Trans. Ind. Electron., 2008, 55, (11), pp. 39443953
Defoort, M., Floquet, T., Perruquetti, W., Drakunov, S.: Integral
sliding mode control of extended Heisenberg system, IET Control
Theory Appl., 2009, 3, (10), pp. 14091424
Guechi, E.-H., Lauber, J., Dambrine, M., Blaic, S., Klanar, G.:
Tracking-error model-based PDC control for mobile robots with
acceleration limits. Proc. IEEE Int. Conf. on Fuzzy Systems, FUZZIEEE2009, 2009, pp. 197202
Takagi, T., Sugeno, M.: Fuzzy identication of systems and its
application to modeling and control, IEEE Trans. Syst. Man Cybern.,
1985, 15, (1), pp. 116132
Taniguchi, T., Tanaka, K., Wang, H.O.: Model construction, rule
reduction and robust compensation for generalized from of Takagi
Sugeno systems, IEEE Trans. Fuzzy Syst., 2001, 9, (4), pp 525537
Tuan, H.D., Apkarian, P., Narikiyo, T., Yamamoto, Y.: Parameterized
linear matrix inequality techniques in fuzzy control system design,
IEEE Trans. Fuzzy Syst., 2001, 9, pp. 324332
Kanayama, Y., Kimura, Y., Miyazaki, F., Noguchi, T.: A stable
tracking control method for a non-holonomic mobile robot. Proc.
IEEE/RSJ Int. Workshop on Intelligent Robots and Systems IROS,
Osaka, Japan, 1991, pp. 12361241
Klancar, G., Skrjanc, I.: Tracking-error model-based predictive
control for mobile robots in real time. Robot. Auton. Syst., 2007, 55,
(3), pp. 460469
Drakunov, S.V., Perruquetti, W., Richard, J.-P., Belkoura, L.: Delay
identication in time-delay systems using variable structure observers,
Annu. Rev. Control, 2006, 3, pp. 143158
Defoort, M., Palos, J., Kokosy, A., Floquet, T., Perruquetti, W.:
Performance-based reactive navigation for non-holonomic mobile
robots, Robotica, 2009, 27, (2), pp. 281290
Defoort, M.: Distributed receding horizon planning for multi-robots
systems. IEEE Multi-Conf. on Systems and Control, Yokohama,
Japan, 2010
Utkin, V., Guldner, J., Shi, J.: Sliding modes control in
electromechanical systems (Systems and Control Book Series, Taylor
and Francis, 1999)
Perruquetti, W., Barbot, J.P.: Sliding mode control in engineering
(Control Eng. Series, Marcel Dekker, 2002)
Edwards, C., Spurgeon, S.K.: Sliding mode control: theory and
applications (Systems and Control Book Series, Taylor and Francis,
1998)
Davila, J., Fridman, L., Levant, A.: Second-order sliding-mode
observer for mechanical systems, IEEE Trans. Autom. Control, 2005,
50, (11), pp. 17851789
Floquet, T., Edwards, C., Spurgeon, S.K.: On sliding mode observers
for systems with unknown inputs, Int. J. Adapt. Control Signal
Process, 2007, 21, pp. 638656
Levant, A.: Higher-order sliding modes, differentiation and output
feedback control, Int. J. Control, 2003, 76, pp. 924941
Levant, A.: Robust exact differentiation via sliding mode technique,
Automatica, 1998, 34, (3), pp. 379384
733
The Institution of Engineering and Technology 2012