Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 53-2 (2020) 5783–5788
Predictor-Based Control for Nonlinear
Predictor-Based
Predictor-Based Control
Control for Nonlinear
Predictor-Based
Mechanical Systems with for
Control for Nonlinear
Nonlinear
Measurement
Mechanical
Predictor-Based
Mechanical Systems
Systems with
Control

with Measurement
for Nonlinear
Measurement
Mechanical Systems with
Delay.  Measurement
Mechanical Systems Delay.
with
Delay.  Measurement
∗
Delay.  Caballero-Barragan ∗∗
Linda P. Osuna-Ibarra ∗Delay.
Humberto ∗∗
Linda P. Osuna-Ibarra ∗ Humberto Caballero-Barragan
Linda
∗∗∗
Alexander ∗∗ Loukianov
Linda P. P. Osuna-Ibarra
Osuna-Ibarra Alexander
∗ Humberto
Humberto
Loukianov
Caballero-Barragan
Caballero-Barragan
∗∗∗
∗∗∗
∗∗

Linda P. Osuna-Ibarra Alexander

Alexander ∗ Loukianov
Humberto
Loukianov Caballero-Barragan ∗∗
∗∗∗
∗
∗
Centro de Investigación Alexander y Estudios
Loukianov Avanzados ∗∗∗ del IPN, Unidad
Centro
∗Guadalajara,
Centro de
de Investigación
Jalisco,
Investigación yy Estudios
México. e-mail:
Estudios Avanzados del IPN,
lposuna@gdl.cinvestav.mx
Avanzados del IPN, Unidad
Unidad
∗
Centro de Investigación
Guadalajara, Jalisco,
∗∗ y
México. Estudios
e-mail:
hcaballero@gdl.cinvestav.mx Avanzados del IPN,
lposuna@gdl.cinvestav.mx Unidad
∗Guadalajara, Jalisco, México. e-mail: lposuna@gdl.cinvestav.mx
Centro de Investigación
Guadalajara, Jalisco,
∗∗ ∗∗∗México. y Estudios
e-mail:
∗∗ hcaballero@gdl.cinvestav.mx Avanzados del IPN,
lposuna@gdl.cinvestav.mx Unidad
louk@gdl.cinvestav.mx
∗∗ hcaballero@gdl.cinvestav.mx
Guadalajara, Jalisco, ∗∗∗México. e-mail: lposuna@gdl.cinvestav.mx
hcaballero@gdl.cinvestav.mx
∗∗∗ louk@gdl.cinvestav.mx
∗∗ ∗∗∗ louk@gdl.cinvestav.mx
hcaballero@gdl.cinvestav.mx
louk@gdl.cinvestav.mx
Abstract: In this paper, a controller ∗∗∗for nonlinear systems with delay in the measurement using
louk@gdl.cinvestav.mx
Abstract:
aAbstract:
predictor-based In
In this
this paper,
strategy
paper, aa controller
is proposed.
controller for
for nonlinear
The nonlinear
nonlinear systems
systems with
systems
with delay in
considered
delay in the
the inmeasurement
this work are
measurement using
usingthe
aAbstract:
predictor-based
class In this
of mechanicalstrategy paper,
strategy a controller
is proposed.
systemsisofproposed. for
triangularThe nonlinear
Theform nonlinearsystems
and which with
systems delay in
considered
mathematical the
model inmeasurement
this work
canwork are
be obtainedusingthe
a predictor-based
Abstract: In this strategy
paper, a controller for The nonlinear
nonlinear systems systems considered
with considered
delay in the in this are the
a
by predictor-based
class of
means mechanical
of the systems
Euler-Lagrange isofproposed.
triangular
formulation. form nonlinear
A and which
predictor systems
mathematical
for the mechanical model inmeasurement
this
can
nonlinear work
be are using
obtained
system the
is
class of mechanical
a predictor-based systems
strategy of triangular
isofproposed. Theform and
nonlinear which mathematical
systems considered model can
innonlinear
this be
work obtained
are the
class
by
applied of
means mechanical
in of the
order to systems
Euler-Lagrange
get a delay-free triangular
formulation.
system. form A and
Then, a which
predictor
tracking mathematical
for the
PD mechanical
controller modelis can
designed be obtained
system
using is
the
by
classmeans
of in of the Euler-Lagrange
mechanical systems formulation.
of triangular form Aand
predictor
which for the mechanical
mathematical model nonlinear system
can be usingobtained is
by means
applied
predictor of
to the
order Euler-Lagrange
to
compensate get a delay-free
the formulation.
system.
delay effect. A
Then,
In Then, predictor
order toa tracking for
demonstrate the
PD mechanical
controller is
the effectiveness nonlinear
designedof the system
control is
the
applied
by means
applied in
in oforder
the
order to
to get
get a
Euler-Lagrange
a delay-free
delay-free system.
formulation.
system. A
Then, aa tracking
predictor
tracking for PD
the
PD controller
mechanical
controller is
is designed
nonlinear
designed using
system
using the
is
the
predictor
scheme
predictor to
to compensate
proposed,
compensate the
two examples
the delay
delay effect.
areeffect. In
presented,
In order to demonstrate
onetotodemonstrate
order show the performancethe
the effectiveness
effectiveness of the
of the predictor
of the control
and
control
applied
predictor
scheme in toorder to
compensate
proposed, twoget a delay-free
the delay system.
effect. In Then,
order toa tracking
demonstrate PD controller
the is
effectiveness designedof theusing
control the
the
schemeother
predictor
one to show
proposed,
to compensate two examples
the use ofare
examples
the delay
such
are presented,
predictorone
presented,
effect. In orderone to
to
in order show to the
todemonstrate
show the performance
control the system.
performance
the effectiveness
of
of the
Then,
the predictor
a tracking
predictor
of the
and
and
control
scheme
the other
controller proposed,
one
for to two examples
show
time-varying the use of are
such
references presented,
predictor
is designedone
in toforshow
order to the performance
control
mechanical the system.
systems ofwith
the
Then,predictor
aa tracking
measurement and
the
schemeother one
proposed, to show the
two examples use of such
are predictor
presented, in
one order
toforshow to control
the the
performance system. Then,
ofwith
the tracking
predictor and
the other
controller
delay, one
and forfor to show
time-varying
the time-varying the use
performance references of such
references
is shown during predictor
is in
designed order
simulation. to control
mechanical
Additionally, the system.
systems Then,
it is considered a tracking
measurement
the case
controller
the other
controller one
for to show the use
time-varying of
references is
such predictor
is designed
in order
designed for
for mechanical
toAdditionally,
control the
mechanical systems
system.
systems with
withThen,measurement
a tracking
measurement
delay,
when
delay, and
external
and the
the performance
disturbances isare
performance is shown
shownpresent during
during simulation.
in the system Additionally,
simulation. and to deal with it
it is considered
them a method the case of
controller
when
estimation and for
delay, external the
of the
time-varying
performance
disturbanceisare
disturbances references
isshown
present
proposed.
is designed
during in simulation.
the
The
for mechanical
system
results areAdditionally,
and
shownto deal
and withit is
systems is considered
discussed
with measurement
considered
them
the
the case
aa method
and finally case of
the
when external
delay, external
and the disturbances
performance are present
is isshown during in the system
simulation. and to deal
Additionally, with them
it is considered method
the case of
when
estimation
conclusions offorthe disturbances
this work are are
disturbance given present
proposed.
and ideas inThe
the systemare
forresults
future and
workshownto deal
are and with
discussed
proposed. themand a method
finally of
the
estimation
when external of the disturbance
disturbances is
are proposed.
present inThe
the results
system are shown
and to and
deal discussed
with them anda finally
method the
of
estimation offorthe
conclusions thisdisturbance
work are is proposed.
given and ideas Theforresults
future are
workshown
are and discussed and finally the
proposed.
conclusions
Copyright
estimation
conclusions ©offor
2020
forthe this
The
this work
Authors.
disturbance
work are given
are This is and
is proposed.
given and ideas
an open
ideas Thefor
access future
article
forresults
futureare work
under
work are
areCC
the
shown proposed.
andBY-NC-ND
discussed
proposed. license
and finally the
(http://creativecommons.org/
Keywords:
conclusionsDelay for this systems;
work are licenses/by-nc-nd/4.0)
Tracking;
given and Disturbance
ideas for futurerejection. work are proposed.
Keywords:
Keywords: Delay Delay systems;
systems; Tracking;
Tracking; Disturbance
Disturbance rejection.
rejection.
Keywords: Delay systems; Tracking; Disturbance rejection.
Keywords:1. INTRODUCTION
Delay systems; Tracking; Disturbance are caused mainly by the communication delay between
rejection.
1. INTRODUCTION
1. INTRODUCTION are
the caused
caused mainly
are sensors and the by
mainly by the
the communication
controller Fridman (2014).
communication delay
delay between
between
1. INTRODUCTION are sensors
the caused mainly
and the by the communication
controller Fridman (2014). delay between
Industry 4.0 is the1.forth industrial revolution, by integrat- the
INTRODUCTION are sensors and the by controller Fridman (2014).
Industry 4.0 isis the
the forth industrial revolution, by integrat-
integrat- the sensors and the controller Fridman (2014). between
Nonlinearcaused mainly
mechanical the
systems communication
constitute a delay
good case of
ing the Internet
Industry 4.0 of forth
Things into factories
industrial revolution,they will by interact Nonlinear
the
study sensors
since mechanical
and
most theof systems
controller
the constitute
Fridman
industrial robots a
(2014). good
belong case
to thisof
Industry
ing the
withthe
ing other4.0factories,
Internet
Internet
is theof forth
of Things industrial
suppliers,
Things into factories
into
revolution,
factories
consumers, they
they interact Nonlinear
by integrat-
will
transportation
will interact Nonlinear
study since
mechanical
mechanical
most of the
systems
systems
industrial
constitute
constitute
robots
aa good
good
belong
case
case
to
of
of
this
Industry
ing
with the 4.0factories,
Internet
other is theof forth
Things industrial
suppliers, into revolution,
factories
consumers, they by integrat-
will interact
transportation class ofsince
study systems.
most Also,
of the robots
industrialare popular
robots because
belong to they
this
and smart
with other grids. Even
factories, more, inside
suppliers, the factory
consumers, different study
transportation Nonlinear
class since
of mechanical
most
systems. of
Also,thesystems
industrial
robots constitute
are robots
popular a good
belong
because case
to theyof
this
ing
withthe
and Internet
other
smart grids. ofEven
factories, Things
suppliers,
more,intoinside
factories
consumers,
the they interact serve
will different
transportation
factory class
study
class
various
of
since
of
practical
systems.
most
systems. Also,
of
Also,the
purposes.
robots
industrial
robots are
are
Decker
popular
robots
popular
et al. (2017) they
because
belong
because to this
they
cells
and and production stages will be interconnected, making
different serve various practical purposes. Decker et al. (2017)
with smart
and
cells other
smart
and
grids. Even
factories,
grids.
production Evenstages
more, inside
suppliers,
more, will inside
the factory
consumers,
the factory
beproduction
interconnected, transportation
different
making serve various ofpractical purposes. Decker et
et al. (2017)
requests
cells and and giving
production notice
stages of
will thebe interconnected, status, all of class
making The control
serve of systems.
various Also,with
systems
practical robots
purposes. delayare is popular
a current
Decker because
al. (2017) they
challenge for
and smart
cells
requestsand and grids.
production
giving Evenstages
notice more, will
of inside
thebe the factory
interconnected,
production status, different
making
all of The
serve
the control
various
control of systems
practical
community with
purposes.
Richard delay is a
Decker
(2003). current challenge
et al.challenge
Delays (2017)
complicate for
these interactions
requests and giving supported
notice of by production
the wireless sensor-actuator
status, all of The control of systems with delay is a current for
cells
these and
requests production
and giving
interactions stages of
notice
supported will the
bybeproduction
interconnected,
wireless status,
sensor-actuatormaking
all of The
the
the controlimplementation
control
direct of systems with
community Richard delay
of is a current
(2003).
control Delays
techniques, challenge
complicate
becausefor
networks
these (WSAN) supported
interactions Kagermannbyet wireless al. (2013). sensor-actuator The control
control community
of systems Richard
with delay (2003).
is a Delays
current complicate
challenge for
requests
these and
networks (WSAN) giving
interactions
(WSAN) Kagermann notice
supported
Kagermann et of the
by production
wireless
et al.
al. (2013).
(2013). status,
sensor-actuator all of the control
direct
the introduction community
implementation Richard
of a delay inofthe of (2003).
control
output Delays
techniques, complicate
can down-perform because
networks direct implementation
the introduction
control community Richard control
(2003). techniques,
Delays because
complicate
these interactions
Nevertheless,
networks (WSAN) one ofsupported
the main by
Kagermann et wireless
withdraws
al. (2013). sensor-actuator
that holds back the direct
the introduction implementation
controller or of of a
even delay in
destabilizeofthecontrol
output techniques,
the systemcan because
down-perform
when delays
Nevertheless, one ofofKagermann
thethesemainindustrial
withdraws that holdsholdsisback
back aa delay in
inofthe output can
can down-perform
networks
the
Nevertheless, (WSAN)
implementation one of
the main et al. (2013).
withdraws paradigms
that the the the
are
direct implementation
introduction
controller
not taken or
intoof
even delay
destabilize
account. thecontrol
output
the techniques,
system when because
down-perform
delays
Nevertheless,
the implementation one of the
of main
these withdraws
industrial that
paradigms holds is back
the the
the controller
introduction
controller or
or even
of a
even destabilize
delay in
destabilize the the
output
the system
can
system when delays
down-perform
when delays
requirements
the implementationof the wireless
of these networks.
industrial According
paradigms to Kager-
is the are not taken into account.
Nevertheless,
the implementation
requirements one
of theofwireless
thethese
of main withdraws
industrial
networks. that holds
paradigms
According to is back
Kager-the are
the
The
are not taken
controller
notstrategy
taken into
or account.
even
proposed
into destabilize
account. in this the
work system
is a when delays
predictor-based
mann
requirementset al. (2016), the
of the wireless minimum
networks. required
According latency
to Kager- for
the implementation ofthe these industrial paradigms is thefor The strategy proposed in
to this work
requirements
mann
WSANet
mann al. of
etintended
al.
the
towireless
(2016),
(2016), perform
the
networks.
minimum According
required
motion required
minimum control 1toms
latency
islatency Kager-
with
for are
Thenot
approach taken
strategywhich into account.
enables
proposed in work is
compensate
this aa predictor-based
is for the time delay
predictor-based
requirements of the
towireless networks. According tomsKager-
amann
WSAN
WSAN
etintended
reliability al. of
intended
(2016), the minimum
perform
99.9999999%
to perform motion
, however,
motion
required
control
control islatency
current
is 11 mswirelessfor The
with
with
approach
resulting
approach
The
strategy
strategy
which
in the
which
proposed
enables
delay-free
enables
proposed
in
to
to
in
this work is for
compensate
closed-loop
compensate
this work is
a system,
for
a
predictor-based
the
the time
timecan
predictor-based
delay
delaybe
mann
WSAN
a et al.
intended
reliability (2016),
of to the
perform
99.9999999% minimum
motion
, however,required
control is
current latency
1 ms for
with
wireless approach
resulting
applied which
in the
Léchappé enables
delay-free
et al. to compensate
(2015); closed-loop
Loukianov for the
system,
et time
al. candelay
(2017);be
industrial
aWSANreliability networks such
of 99.9999999% as ISA100,
, however, wirelessHART
current and
wireless resulting
approach in
whichthe delay-free
enables to closed-loop
compensate forsystem,
the timecandelaybe
aindustrial intended
reliability of
networks to perform
99.9999999%
such as motion
, however,
ISA100, control is 1 ms
current
wirelessHART with
wireless
and resulting
applied
Caballero-Barragánin the
Léchappé delay-free
et al. (2015);
etal.al.(2015); closed-loop
Loukianov
(2018, Loukianov
2016). system,
et al. can
(2017);be
ZigBee arenetworks
industrial unable tosuch fulfillasthese requirements.
ISA100, wirelessHART and applied
resulting Léchappé
in the et
delay-free closed-loop et
system, al. (2017);
can be
a reliability
industrial
ZigBee are of 99.9999999%
networks
unable to such
fulfill asthese , however,
ISA100, current wireless
wirelessHART
requirements. applied
and Caballero-Barragán Léchappé et etal.
al. (2015);
(2018, Loukianov
2016). et al. (2017);
ZigBee Caballero-Barragán etet al. (2018, 2016).in Léchappé
ZigBee are
industrial
Having that
are unable
networks
in mind,
unable to fulfill
tosuch it isasthese
fulfill requirements.
ISA100,
important
these wirelessHART
to develop control
requirements. applied
Nevertheless,
and Caballero-Barragán Léchappé etal.al.(2015);
the predictors (2018, Loukianov
proposed
2016). et al. (2017);
et al.
Having
ZigBee are
strategies
Having that
that inenables
unable
that in mind, it
to fulfill
mind, is use
theis
it important
these
important to
requirements.
wireless develop control
technologies
to develop to be Nevertheless,
control Caballero-Barragán
(2015);
Nevertheless, Loukianov the
the predictors
etetal.al.(2017);
predictors proposed
(2018, 2016).in
proposed in Léchappé
Caballero-Barragán
Léchappé et
et al.
et al.
Having
strategies that
that in mind,
enables it
the is important
use wireless to develop
technologies control
to be Nevertheless,
(2015);
(2018, 2016) Loukianov the predictors
et
are designedal. (2017); proposed in Léchappé
Caballero-Barragán
linear systems, andet
for Caballero-Barragán et al.
the
implemented.
strategies that Inenables
this paper the it
use is proposed
wireless a strategy totodeal
technologies be (2015);
Nevertheless, Loukianov the et al.
predictors (2017); proposed in Léchappé et
et al.
al.
Having
strategies
implemented. that
that in
In mind,
enables
this it
the
paper itis important
use
it is wireless
is proposed
proposed to develop
technologies
a strategy
strategy control
to to be (2015);
deal (2018,
systems Loukianov
2016) are
considered et al.
designed (2017);
in this work for Caballero-Barragán
linear systems,
are nonlinear, and
since et the
they
with delays in In
implemented. thethis
measurement
paper of the state a process,to which
deal (2018,
(2015); 2016) 2016)
Loukianov are designed
etinal. (2017); for linear systems, and the
strategies
delaysthat
implemented.
with in In enables
thethis the it
paper
measurement use wireless
is proposed
of the
the statetechnologies
statea strategy
process,to todeal
which be systems
(2018,
are mechanical are
considered designed
systems this
modeled for Caballero-Barragán
work linear
are
using systems,
nonlinear,
the andet
since
Euler-Lagrange
al.
the
they
with delays in the measurement of process, which systems
(2018,
systems considered
2016) are
considered in this
designed
in this work
for
work are
linear
are nonlinear,
systems,
nonlinear, since
and
since they
the
they
implemented.
with delays in In
thethis paper
measurement it is proposed
of
 This work was supported by CONACYT, Mexico, under grant the statea strategy
process, to deal
which are mechanical
formulation. systems
So,systems modeled
for the modeled
proposedusing using the
scenario, Euler-Lagrange
the considera-
are
systems
are mechanical
considered
mechanical systemsin this
modeled work using
are the
nonlinear,
the Euler-Lagrange
since
Euler-Lagrange they
 with delays in
wasthe measurement of the state process, which formulation.
tion of a delaySo, infor
the the proposed
measurement scenario,
of a the
nonlinear considera-
system
 This
300959, work
301068 supported
and 252405. byL.CONACYT,
P. Mexico,
Osuna-Ibarra,
This work was supported by CONACYT, Mexico, under grant H. under grant
Caballero- formulation. So, for the proposed scenario, the considera-
 This work are
tion mechanical
formulation.
of a delaySo, systems
infor
the the modeled
proposed
measurement using of the
scenario,
a Euler-Lagrange
the
nonlinear considera-
system
300959, andwas
301068 supported
A.andG. 252405. byL.CONACYT,
P. Osuna-Ibarra, Mexico,H.under del grant
Caballero- complicates
tion of theindesign of a controller Richard (2003).
of aa delay the measurement of aa nonlinear system
Barragán,
300959, 301068 and Loukianov
252405. L. are with CINVESTAV
P. Osuna-Ibarra, H. Caballero- IPN,

300959,
Barragán,
This work
Unidad 301068
andwasA.and
Guadalajara, G. 252405.
Loukianov
supported
Av. del byL.Bosque
P. Osuna-Ibarra,
are
CONACYT,with1145CINVESTAV
Mexico,
CP. 45019,H.under
Caballero-
del IPN,
grant
Jalisco,
formulation.
tion
complicates delaySo,indesign
the for
thethe ofproposed
measurement
aa controller scenario,
of Richard the(2003).
nonlinear considera-
system
Barragán,
Barragán,
and A. G. Loukianov
and A.and G. 252405.
Loukianov
are with
are with
CINVESTAV
CINVESTAV
del
del
IPN,
IPN,
complicates
tion
Still, of a
there the
delay
are design
in
in the
the of
measurement controller
literature
complicates the design of a controller Richard (2003). of Richard
a
predictors nonlinear(2003).
system
designed for
Unidad
300959,
Mexico. Guadalajara,
301068 Av. delL.Bosque
lposuna@gdl.cinvestav.mx 1145 CP. 45019,
P.hcaballero@gdl.cinvestav.mx
Osuna-Ibarra, Jalisco,
H. Caballero-
Unidad Guadalajara, Av. del Bosque 1145 CP. 45019, Jalisco, Still,
Unidad
Mexico.
Barragán, Guadalajara,
and A. G. Av. del
Loukianov
lposuna@gdl.cinvestav.mx
louk@gdl.cinvestav.mx. Bosque
are with1145 CP.
CINVESTAV 45019,
hcaballero@gdl.cinvestav.mx
Mexico. lposuna@gdl.cinvestav.mx hcaballero@gdl.cinvestav.mx
Jalisco,
del IPN, Still, there are in the literature predictors designed for
complicates
nonlinear there are
the
systems in the
designsuch literature
of asa controllerpredictors
Bresch-Pietri Richard et designed
(2003).
al. (2015);
for
Still, theresystems
nonlinear are in the such literature
as predictorsetdesigned for
Mexico.
Unidad lposuna@gdl.cinvestav.mx
Guadalajara, Av. del Bosque
louk@gdl.cinvestav.mx.
louk@gdl.cinvestav.mx.
hcaballero@gdl.cinvestav.mx
1145 CP. 45019, Jalisco, nonlinear
Still, there systems
are in as Bresch-Pietri
suchliterature
the Bresch-Pietri
predictors
al.
al. (2015);
etdesigned(2015);
for
louk@gdl.cinvestav.mx.
Mexico. lposuna@gdl.cinvestav.mx hcaballero@gdl.cinvestav.mx nonlinear systems such as Bresch-Pietri et al. (2015);
nonlinear
2405-8963 Copyright © 2020 The Authors. This is an open access article under the CC BY-NC-ND license.
louk@gdl.cinvestav.mx. systems such as Bresch-Pietri et al. (2015);
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2020.12.1615
5784 Linda P. Osuna-Ibarra et al. / IFAC PapersOnLine 53-2 (2020) 5783–5788

Bekiaris-Liberis and Krstic (2016, 2017) that can be used

in order to design a controller for a robotic mechanical
system.
The rest of the paper is organized as follows. In Section
2 the main mathematical concepts used in this paper are
reviewed to give the reader a general idea of the applied
techniques. In Section 3 the statement of the problem is
made, lighting out the things aimed to solve in this work.
In Section 4, it is applied a predictor for mechanical non-
linear systems and the tracking PD controller is designed Fig. 1. The control scheme.
using the predictor to compensate the delay effect, two
examples are presented. A tracking controller for time- Assumption 2. The system (4) is fully actuated and the
varying references is designed for mechanical systems with M (x1 ) matrix has full rank. Also, its inverse M (x1 )−1
measurement delay, and it is presented in the Section 5. exists ∀x1 .
Then, the case for disturbed system with time delay is con-
sidered proposing a way to estimate and compensate the The objective of this work is to design a controller using a
external disturbances. Finally, in Section 7 the conclusions predictor for nonlinear systems and reduce the time-delay
for this work are given. effect. Fig. 1 shows the general control scheme proposed in
this work. Having a desired reference for the state of the
2. MATHEMATICAL BACKGROUND system and considering a delay D in the measurement,
the controller has to be designed in order to track said
2.1 Predictor for Nonlinear Systems reference by means of a predictor.

Consider the following nonlinear system: 4. PREDICTOR AND CONTROL DESIGN

ẋ(t) = f (x(t), u(t − D)) (1)
where x ∈ R2n is the state, u ∈ Rn is the control input, D
is a known scalar constant and f : R2n × Rn → R2n is a t
locally Lipschitz vector field that satisfies f (0, 0) = 0. ξ(t) = x̄(t + D) = x̄(t) + f (ξ(θ), τ (θ))dθ (5)
t−D
The predictor ξ(t) = x(t+D) for the system (1) is designed where
as follows Bekiaris-Liberis and Krstic (2017) f (ξ(θ), τ (θ)) =
 
t ξ¯2 (θ)
,
ξ(t) = x(t) + f (ξ(θ), u(θ))dθ (2) M (ξ¯1 (θ))−1 (−C(ξ¯1 (θ), ξ¯2 (θ))ξ¯2 (θ) − G(ξ¯1 (θ)) + τ (θ))
 
t−D ξ¯ (t)
with ξ(t) = ¯1 .
where ξ ∈ R2n is the predictor state. ξ2 (t)
Now, taking the time derivative of (2) along trajectories
of (1), the delay-free system reads as 4.1 Example 1.
˙ = f (ξ(t), u(t)).
ξ(t) (3)
The predictor’s performance is shown using the following
The controller u can be designed using system (3). model of a two-link planar manipulator (see Craig (2005)
and Fig. 2)
3. STATEMENT OF PROBLEM
ẋ1 (t) =x2 (t)
Consider the mechanical model using Euler-Lagrange for- ẋ2 (t) =M (x1 )−1 (−C(x1 , x2 )x2 (t) − G(x1 ) + τ (t)),
     
mulation x1 (t) q1 (t) q̇1 (t)
ẋ1 (t) =x2 (t) where x(t) = , x1 (t) = , x2 (t) = ,
x2 (t) q2 (t) q̇2 (t)
ẋ2 (t) =M (x1 )−1 (−C(x1 , x2 )x2 (t) − G(x1 ) + τ (t)), x(0) = [pi/2 0 0 0]T , τ (t) = 0.1[sin(0.5t) sin(0.7t)]T .
  (4)
x1 (t − D) M (x1 ) =
y(t) =x(t − D) = , 2 
x2 (t − D) l2 m2 + 2l1 l2 m2 c2 + l12 (m1 + m2 ) l22 m2 + l1 l2 m2 c2
,
where x1 (t) ∈ Rn is the angular position vector, x2 (t) ∈ Rn l22 m2 + l1 l2 m2 c2 l22 m2
is the angular velocity vector, M (x1 ) ∈ Rn×n is the inertia  
−2m2 l1 l2 s2 q̇1 −m2 l1 l2 s2 q̇2
matrix, C(x1 , x2 ) ∈ Rn×n is the matrix of Coriolis and C(x1 , x2 ) =
m2 l1 l2 s2 q̇1 0
and
centrifugal forces, G(x1 ) ∈ Rn is the vector of gravitational  
forces, τ (t) ∈ Rn is the vector of control input, y(t) is the G(x1 ) =
m2 l2 gc12 + (m1 + m2 )l1 gc1
,
available measurement state and D is the time-delay. m2 l2 gc12
  with c1 = cos(q1 ), c2 = cos(q2 ), s2 = sin(q2 ), c12 =
x1 (t − D) cos(q1 + q2 ), g = −9.81 m/s2 , l1 = 0.9 m, l2 = 0.7 m,
Assumption 1. The state vector x(t − D) =
x2 (t − D) m1 = 0.4 kg, m2 = 0.3 kg and D = 100 ms.
is available.
 Linda P. Osuna-Ibarra et al. / IFAC PapersOnLine 53-2 (2020) 5783–5788 5785

−5
x 10
1
e1(t)
e2(t)
0.5
e (t)
3
e4(t)
0

−0.5

−1
Fig. 2. Two-link planar manipulator. 0 0.5 1 1.5 2 2.5 3
Time(s)

1.5
Fig. 5. The predictive error ep (t) = x̄(t) − ξ(t − D).
or
1
ξ¯˙1 (t) =ξ¯2 (t)
(6)
ξ¯˙2 (t) =M (ξ¯1 )−1 (−C(ξ¯1 , ξ¯2 )ξ¯2 (t) − G(ξ¯1 ) + τ (t))
Rad

To make the tracking of the reference xref by the system

0.5 (4) output x1 the control law τ is chosen of the form
τ (t) = −Kp e(t) − Kv ξ¯2 (t) + G(ξ¯1 ), (7)
where e(t) ∈ R is the vector of angular position error,
n
0 and it is defined as
0 0.5 1 1.5 2 2.5 3
Time(s)
e(t) = ξ¯1 (t) − xref ,
Fig. 3. Behavior of x̄(t) and ξ(t − D). with the positive defined matrices Kp ∈ Rn×n and Kv ∈
Rn×n .
1.572 To prove the convergence of the closed-loop system (4)
and (7), the following positive definite candidate Lyapunov
function is proposed:
1.57 1 1
V = eT Kp e + ξ¯2T M (x1 )ξ¯2 . (8)
2 2
Rad

1.568 Taking the time derivative of (8) along the trajectories of

the closed-loop system yields
1
V̇ = ξ¯2T Kp e + ξ¯2T M (x1 )ξ¯˙2 + ξ¯2T Ṁ ξ¯2 ,
1.566 2
= ξ¯2T Kp e + ξ¯2T M (ξ¯1 )(M (ξ¯1 )−1 (−C(ξ¯1 , ξ¯2 )ξ¯2 − G(ξ¯1 )
1
1.564 − Kp e − Kv ξ¯2 + G(ξ¯1 ))) + ξ¯2T Ṁ ξ¯2 ,
0 0.5 1 1.5 2 2.5 3 2
Time(s) 1
= −ξ¯2T C(ξ¯1 , ξ¯2 )ξ¯2 + ξ¯2T Ṁ ξ¯2 − ξ¯2T Kv ξ¯2 .
2
Fig. 4. Behavior of q̄1 (t), ξ1 (t) and ξ1 (t − D). (9)
Fig. 3 shows the behavior of the complete state x̄(t) = Making use of the property for articular robots stated
x(t − D) and the retarded state of the predictor in order in Kelly and Santibañez (2003) q̇ T ( 12 Ṁ − C)q̇ = 0, the
to compared the predictor and the measured state of the derivative (9) becomes
system. In Fig. 4 presents the comparison between q̄1 (t), V̇ = −ξ¯T Kv ξ¯2 .
2 (10)
ξ1 (t) and ξ1 (t − D). The prediction error is presented in
Fig. 5 and the error is defined as ep (t) = x̄(t) − ξ(t − D). Since the function (10) is semi negative definite the LaSalle
invariance principle can be easily applied to show that the
4.2 Linear Feedback Control equilibrium point e = 0 and ξ¯2 = 0 of the closed-loop
system is asymptotically stable.
In order to design the controller τ (t) it is necessary to
4.3 Example 2.
obtain a delay-free system using the predictor 5. Taking
the time derivative of (5), results
To illustrate the performance of the proposed controller
˙ = f (ξ(t), τ (t)),
ξ(t) (7), the closed-loop system is simulated using MATLAB
5786 Linda P. Osuna-Ibarra et al. / IFAC PapersOnLine 53-2 (2020) 5783–5788

x(t) τ(t)
2 4

1.5 2
0
1

Torque(Nm)
−2

0.5 −4
−6
0
−8
−0.5 −10
−12
−1
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Time(s) Time(s)

Fig. 6. Response of x̄(t). Fig. 8. Torque control τ (t).

5. TRACKING OF TIME-VARYING REFERENCE
2 Consider the system (4) under Assumption 1 and using
the predictor (5), the delay-free system is obtained as (6)
1.5 ξ¯˙1 (t) =ξ¯2 (t)
(11)
ξ¯˙2 (t) =M (ξ¯1 )−1 (−C(ξ¯1 , ξ¯2 )ξ¯2 (t) − G(ξ¯1 ) + τ (t)).
1
To track time-varying references the controller τ (t) is de-
Rad

signed as follows. The time-varying references are obtained

0.5 by following exosystem:
xref (t) = p(g), ġ = d(g). (12)
0 Now, using the reference xref (t), the tracking error can be
defined as follows
−0.5
0 0.5 1 1.5 2 2.5 3
e(t) = ξ¯1 (t) − xref (t). (13)
Time(s) Using the backstepping technique Krstic et al. (1995) the
tracking control is designed in two steps.
Fig. 7. Reference tracking.
Step 1.
Simulink. The matrices for the system (4) are defined Taking the time derivative of (13), the dynamic of e(t) is
accordingly to the robot structure depicted in Fig. 2. obtained
Consider the parameters described in example 1 4.1. ė(t) = ξ¯2 (t) − ẋref (t). (14)
Now, using the virtual control ξ¯2 , the following change
The control gains are described by variable is proposed
    z1 (t) = ξ¯2 (t) + K1 e(t) − ẋref (t), (15)
15 0 90
Kp = and Kv = . where K1 > 0 is a positive and symmetric matrix.
0 13 07
A candidate Lyapunov function is proposed and it reads
The constant reference is described as follows as
  1
pi/4 V1 (t) = eT (t)e(t). (16)
xref = . 2
0
Taking the time derivative of (16) and using (14) and (15),
The control is defined as (7), reads as results
   
15 0 90 ¯ V̇1 (t) =eT (t)ė(t) = eT (t)(−K1 e(t) + z1 (t))
τ (t) = − e(t) − ξ (t) + G(ξ¯1 ), (17)
0 13 07 2 = − eT (t)K1 e(t) + eT (t)z1 (t).
where e(t) = ξ¯1 (t) − xref . Step 2.
Now, taking the time derivative of (15) along the trajec-
Simulations results are presented in Figs. 6-8. Fig. 6 tories (11) and (14), it read as
presents the behavior of the system state x(t). Fig. 7 shows ż1 (t) =M −1 (ξ¯1 )(−C(ξ¯1 , ξ¯2 )ξ¯2 (t) − G(ξ¯1 ) + τ (t))+
the tracking of the constant reference. The torque control (18)
+ K1 ξ¯2 (t) − K1 ẋref (t) − ẍref (t).
is presented in Fig. 8. The results without predictor are not
presented because the result is foreseeable and well-known The control τ (t) is designed as follows
(instability or best-case scenario a bad transient behavior τ (t) =C(ξ¯1 , ξ¯2 )ξ¯2 (t) + G(ξ¯1 ) + M (ξ¯1 )(−K1 ξ¯2 (t)+
(19)
in close-loop). + K1 ẋref (t) + ẍref (t) − K2 z1 (t) − e(t)),
 Linda P. Osuna-Ibarra et al. / IFAC PapersOnLine 53-2 (2020) 5783–5788 5787

where K2 > 0 is a positive and symmetric matrix.

2
The closed-loop system by the control 19 results

ė = − k1 e(t) + z1 (t) 1.5

(20)
z˙1 = − k2 z1 (t) − e(t).
A candidate Lyapunov function is proposed as follows 1
1
V2 (t) = V1 (t) + z1T (t)z1 (t). (21)
2 0.5
Now, taking the time derivative of (21) and using (17) and
(20), it reads as
V̇2 (t) = − eT (t)K1 e(t) − z1T (t)K2 z1 (t) 0
(22) 0 0.5 1 1.5 2 2.5 3
V̇2 (t) ≤ − λmin (K1 ) e(t) − λmin (K2 ) z1 (t) . Time(s)
The time derivative of (21) is negative defined then the
tracking control lim e(t) → 0. Fig. 9. Tracking of time-varying reference, x̄(t) vs xref (t −
t→∞ D).
5.1 Example 3.
3
To test the effectiveness of the controller described in this
section the following model of two-link planar manipulator 2
(see Craig (2005) and Fig. 2) described in section 4.1 is
used: 1
ẋ1 (t) =x2 (t)
ẋ2 (t) =M (x1 )−1 (−C(x1 , x2 )x2 (t) − G(x1 ) + τ (t)), 0

with the same matrices description and parameters.

−1
The exosystem that describe the references to track is
described by following system: −2
 
0 −α
ġ(t) = g(t) −3
α 0 0 1 2 3 4 5
  (23) Time(s)
π/3
xref (t) =g(t) + ,
π/7 Fig. 10. Tracking performance under the effect of external
where g(0) = [0.2 0]T and α = 2. The state available
 is disturbances.
8 0
x(t − D) = x̄(t). The control gains are K1 = K2 = . the controller described in (7) the tracking performance
0 7
can be seen in Fig 10, with the references set at 2π/3 and
The simulation results are presented in Fig 9. The pre- −2π/3 for each angle. The blue and brown lines are the
dictor tracks the reference xref (t) and the measurement tracking performance under disturbance.
state x̄(t) tracks the retarded reference xref (t − D) as Fig. The error in the tracking is present because predictors are
9 shows. This is a natural situation when working with highly sensitive to external disturbances, as is reflected in
systems with delay in the input and/or in the output. the predictor performance under disturbance ∆(t).
6. TRACKING OF REFERENCE UNDER In order to deal with these undesired effects, it is necessary
DISTURBANCES to mitigate the effect of the disturbances in the predictor
so it does not affect the tracking performance. To do so,
Consider now the system (1) under external disturbances, an approximation of the disturbance ∆(t) is proposed as:
such that the description of the dynamics result as
ẋ(t) = f (x(t), u(t − D), ∆(t)) (24)  
x1 (t − D)
where ∆(t) is a disturbance vector of appropriate dimen- x̄ = ,
x2 (t − D)
sions and x is the state vector under disturbance.
x̄(t) − x̄(t − λ) (25)
Using as example the model described in section 4.1 under ˆ
∆(t) = − f (x̄, u(t − D)),
λ
disturbance:
ẋ1 (t) =x2 (t)
with λ a design constant. Using this ∆(t)ˆ approximation
ẋ2 (t) =M (x1 )−1 (−C(x1 , x2 )x2 (t) − G(x1 ) + τ (t))+
+ ∆(t), in the predictor design
with the same matrices description and parameters and t
  ˆ
2.5 ξ(t) = x̄(t) + [f (ξ(θ), u(θ)) + ∆(θ)]dθ. (26)
setting the disturbance as ∆(t) = M (x1 )−1 . Using
1 t−D
5788 Linda P. Osuna-Ibarra et al. / IFAC PapersOnLine 53-2 (2020) 5783–5788
3 in the systems and to consider time-varying delays. This
2 error and as a result the analysis is more challenging and
1
0
−1
−2
−3
0 1 2 3 4 5
Time(s)
Fig. 11. Tracking performance using estimation of distur-
bance in the predictor.
6
5 Loukianov, A.G. (2016).
4 Conference on Decision and Control, CDC 2016, Las
3 Caballero-Barragán, H., Osuna-Ibarra, L.P., Loukianov,
2
1
0 Control: Chapter 6. Manipulator Dynamics. Pearson
0 1 2 3 4 5
Time(s)
Fig. 12. Tracking mean squared error with (blue) and
without (red) disturbance estimation used in the
prediction.
And, since the prediction improves, the tracking error also
diminishes as shown in Fig 11, where the purple and green
lines are the tracking performance of each joint and the
black lines are the references to be reached. In order to
compare the effect of the use of disturbance estimation on
the prediction, Fig 12 shows the mean squared error of
the tracking performance with and without the use of the
disturbance estimation.
7. CONCLUSIONS
In this work a predictor-based tracking control for me-
chanical systems with measurement delay is presented.
The predictor is applied in simulation to show the be-
havior in open-loop and near to an equilibrium point.
The predictor and tracking PD controller is designed to
control the mechanical system with measurement delay
and to track constant references. Simulation results are
presented. A tracking control for time-varying references
and a predictor for mechanical systems with measurement
delay is applied using an exosystem to get the time-varying
reference and the simulations results are shown as well.
Also, it is considered the case of the tracking performance
under disturbances. In order to diminish the effect of these
disturbances on the prediction error (and consequently
on the tracking error) an estimation of the disturbances
is proposed and the simulation results are shown. As a
future work is proposed to considered parametric variation
is interesting because in this case there exist prediction
complex.
REFERENCES
Bekiaris-Liberis, N. and Krstic, M. (2016). Stability of
predictor-based feedback for nonlinear systems with
distributed input delay. Automatica, 70, 195–203.
Bekiaris-Liberis, N. and Krstic, M. (2017). Predictor-
feedback stabilization of multi-input nonlinear systems.
IEEE Transactions on Automatic Control, 62(2), 516–
531.
Bresch-Pietri, D., Petit, N., and Krstic, M. (2015).
Prediction-based control for nonlinear state- and input-
delay systems with the aim of delay-robustness analysis.
2015 54th IEEE Conference on Decision and Control
(CDC), 6403–6409.
Caballero-Barragán, H., Osuna-Ibarra, L.P., and
Dynamic predictor for
linear time-delay systems with disturbances. 55th IEEE
Vegas, NV, USA, December 12-14, 2016, 2290–2295.
A.G., and Plestan, F. (2018). Sliding mode predictive
control of linear uncertain systems with delays. Auto-
matica, 94, 409–415.
Craig, J.J. (2005). Introduction to Robotics Mechanics and
Prentice Hall, New Jersey, USA.
Decker, M., Fischer, M., and Ott, I. (2017). Service
robotics and human labor: A first technology assess-
ment of substitution and cooperation. Robotics and
Autonomous Systems, 87, 348 – 354.
Fridman, E. (2014). Introduction to Time-Delay Sys-
tems:Analysis and Control. Birkhäuser, London, UK.
Kagermann, H., Wahlster, W., and Helbig, J. (2013). Rec-
ommendations for implementing the strategic initiative
industrie 4.0: Final report of the industrie 4.0 working
group. National Academy of Science and Engineering.
Kagermann, H., Wahlster, W., and Helbig, J. (2016).
Network-based communication for industrie 4.0. Federal
Ministry for Economy Affairs and Energy.
Kelly, R. and Santibañez, V. (2003). Control de
Movimiento de Robots Manipuladores. Pearson Prentice
Hall.
Krstic, M., Kanellakopoulos, I., and Kokotovic, P.V.
(1995). Nonlinear and Adaptive Control Design. Wiley,
New York, USA.
Léchappé, V., Moulay, E., Plestan, F., and Glumineau,
A. (2015). New predictive scheme for the control of lti
systems with input delay and unknown disturbances.
Automatica, 52, 179–184.
Loukianov, A.G., Caballero-Barragán, H., Osuna-Ibarra,
L., Espinosa-Guerra, O., and Castillo-Toledo, B. (2017).
Robust control for uncertain linear delay systems via
sliding mode control. Int. J. Robust. Nonlinear Control,
27(18), 4825—-4845, doi: 10.1002/rnc.3834.
Richard, J.P. (2003). Time-delay systems: an overview of
some recent advances and open problems. Automatica,
39(10), 1667–1694.