You are on page 1of 6

Proceedings,16th IFAC Symposium on

Proceedings,16th
Information Control
Proceedings,16th IFAC Symposium
Problems
IFAC on
in Manufacturing
Symposium on
Proceedings,16th
Proceedings,16th
Information
Bergamo, Control
Italy, IFAC
IFAC
June Symposium
Symposium
Problems
11-13, in
2018 on
Available online at www.sciencedirect.com
on
Manufacturing
Information
Information Control
Proceedings,16th
Control Problems
IFAC in
Symposium
Problems Manufacturing
on
in Manufacturing
Manufacturing
Information
Bergamo, Control
Italy, JuneProblems
11-13, in
2018
Bergamo, Italy,
Information June
Control 11-13, 2018
Bergamo,
Bergamo, Italy,
Italy, JuneProblems
June 11-13, in Manufacturing
11-13, 2018
2018
Bergamo, Italy, June 11-13, 2018 ScienceDirect
IFAC PapersOnLine 51-11 (2018) 740–745
Robotic manipulators with double encoders: accuracy improvement based on
Robotic
Robotic manipulators
advancedwith double encoders: andaccuracy improvement based on
Robotic manipulators
manipulators with
with double
stiffness encoders:
modeling
double encoders: accuracy
intelligent
accuracy improvement
control
improvement based
based on
on
Robotic manipulators
advanced
advanced with double
stiffness
stiffness encoders:
modeling
modeling and
and accuracy
intelligent
intelligentimprovement
control
control based on
advanced stiffness modeling and intelligent control
advanced stiffness modeling
A. Klimchik*
*
and intelligent
, A. Pashkevich *,**
*,**
control
A.
A. Klimchik**,, A. Pashkevich*,**
A. Klimchik
A. Klimchik*,, A.
Klimchik A. Pashkevich
A. Pashkevich*,**
Pashkevich *,**
*
*Innopolis University, Universitetskaya A. Klimchik 1, 420500 Innopolis, *,**
, A. Pashkevich The Republic of Tatarstan, Russia
*Innopolis
*Innopolis University,
University, Universitetskaya
Universitetskaya(e-mail: 1, 
420500 Innopolis,
a.klimchik@innopolis.ru)
1, 420500 Innopolis, The Republic
The Republic of of Tatarstan,
Tatarstan, Russia Russia
*Innopolis
*Innopolis University, University, Universitetskaya
Universitetskaya 1,
1, 420500
420500 Innopolis,
Innopolis, The The Republic
Republic of of Tatarstan,
Tatarstan, Russia Russia
** IMT Atlantique, 4 rue Alfred-Kastler,
*Innopolis University, Universitetskaya (e-mail:
Nantes
(e-mail: a.klimchik@innopolis.ru)
44307, Le Laboratoire
a.klimchik@innopolis.ru) des Sciences du
1, 420500 Innopolis, The Republic of Tatarstan, Russia Numérique de Nantes (LS2N)
(e-mail:
(e-mail:44307, a.klimchik@innopolis.ru)
a.klimchik@innopolis.ru)
**
** IMT
IMT Atlantique,
Atlantique, 44 rue
rue Alfred-Kastler,
Alfred-Kastler, (e-mail: Nantes
Nantes Le Laboratoire
anatol.pashkevich@imt-atlantique.fr).
44307, Le Laboratoire des
des Sciences
Sciences du
du Numérique
Numérique de
de Nantes (LS2N)
**
** IMT
IMT Atlantique,
Atlantique, 44 rue rue Alfred-Kastler,
Alfred-Kastler, (e-mail:
(e-mail:
Nantes
Nantes a.klimchik@innopolis.ru)
44307,
44307, Le Le Laboratoire
Laboratoire des
anatol.pashkevich@imt-atlantique.fr). des Sciences
Sciences du du Numérique
Numérique de de Nantes
Nantes (LS2N)
Nantes (LS2N)
(LS2N)
** IMT Atlantique, 4 rue Alfred-Kastler, (e-mail:
(e-mail:
(e-mail: Nantes anatol.pashkevich@imt-atlantique.fr).
44307, Le Laboratoire des Sciences du Numérique de Nantes (LS2N)
anatol.pashkevich@imt-atlantique.fr).
anatol.pashkevich@imt-atlantique.fr).
(e-mail: anatol.pashkevich@imt-atlantique.fr).
Abstract: The paper deals with advanced stiffness modeling and intelligent control of industrial robots
Abstract:
with double The paper dealsIn with advanced stiffness modeling
that and intelligent control of industrial robots
Abstract:
Abstract:
Abstract: Theencoders.
The
The paper
paper deals
paper deals
deals
contrast
with
with
with advanced
advanced
advanced
to previous
stiffnessworks
stiffness
stiffness modeling
modeling
modeling
concentrate
and
and
and intelligentoncontrol
intelligent
intelligent
the actuator
control
control of
of industrial
of
compliance
industrial
industrial robots
robots
robots
with
with
Abstract: double
compensation,
double The encoders.
this
encoders.
paper work
dealsIn
In is contrast
aimed
contrast
with at
advanced to
to previous
reduction
previous of
stiffness works
the
works that
that
modeling concentrate
compliance errors
concentrate
and intelligent on
on the
produced
the
control actuator
by
actuator
of all compliance
manipulator
compliance
industrial robots
with double
with double encoders. encoders. In In contrast
contrast to to previous
previous works works that
that concentrate
concentrate on on thethe actuator
actuator compliance
compliance
compensation,
components. this
Particular work is
attention aimed is at
paid reduction
to the of the compliance
identifiability of the errors
stiffnessproduced
model by all
parameters manipulator
compensation,
with double
compensation, this
compensation, this
encoders. work
this work
work is In is aimed
contrast
is aimed
aimed at at to reduction
previous
at reduction
reduction of of the
works
of the compliance
that
the compliance errors
concentrate
compliance errors produced
on
errors produced the
produced by by
actuator
by allall manipulatora
using
manipulator
compliance
all manipulator
components.
combination of
components. Particular
the tool
Particular attention
location
attention is
is paid
measurements
paid to
to the
the identifiability
and information
identifiability of
of the
from
the stiffness
doubleproduced
stiffness model
encoders
model parameters
integrated
parameters using
into
using thea
compensation,
components.
components.
combination
this
Particular
Particular
of the
work
tool
is
attention
attention
location
aimed is at
paid reduction
to
is paid to the and
measurements
the of the compliance
identifiability
identifiability
information
of the
offrom errors
stiffness
the stiffness
double
model by all
parameters
model parameters
encoders integrated usingtheaaa
manipulator
using
into
actuated
combination
components. joints.of the For
Particular tool the
locationconsidered
attention measurements architecture,
is paid to the and and new strategies
information
identifiability from
offrom for
double
the stiffness intelligent
encoders
model parameters compliance
integrated intoerror
the
usingthe
combination
combination
actuated joints.
of
of the
theFortool
toolthe location
locationconsidered
measurements
measurements and
architecture,
information
information
new from
strategies
double
double
for
encoders
encoders integrated
integrated into
into thea
compensation
actuated
combination
actuated joints.
joints.of are
the proposed.
For
tool
For the
location
the The
considered
considered developed
measurements techniques
architecture,
and
architecture, new were
information
new applied
strategies
from
strategies for
double
for tointelligent
the stiffness
intelligent
encoders
intelligent
compliance
modeling
compliance
integrated
compliance intoerror
and
error
the
error
actuated
compensation joints.areFor the considered
proposed. The developed
developed architecture,
techniques new were
strategies
applied fortointelligent
the stiffness
stiffness compliance
modeling error
and
accuracy joints.
compensation
actuated
compensation improvement are
are For of aconsidered
proposed.
the
proposed. typical
The
The industrial
developed robotic
techniques
architecture,
techniques newmanipulator,
were applied
strategies
were applied for
for to
towhich
the
intelligent
the the impact
stiffness complianceof external
modeling
modeling and
error
and
compensation
accuracy improvement are proposed. of The developed
aa typical industrial techniques
robotic were applied
manipulator, for towhich
the stiffness
the impact modeling
of and
external
loading and
accuracy gravity
improvement on the of positioning
aa typical accuracytechniques
industrial was
roboticessentially
werereduced.
manipulator, for
compensation
accuracy
accuracy
loading and
improvement
improvement
gravity
are proposed.
on
of The developed
typical
of positioning
the typical industrial
industrial
accuracy
robotic
robotic
was
manipulator,
manipulator,
essentially
applied
reduced. for towhich
for which
which the
the impact
the stiffness
the impact
impact of
of external
modeling
of and
external
external
loading
accuracy and gravity
improvement on the
of positioning
a typical accuracy
industrial was
roboticessentially reduced.
manipulator, for which the impact of external
© 2018, and
loading
loading
Keywords: IFAC
and (International
gravity
gravity
Robotic on manipulator,
on the Federationdouble
the positioning
positioning of Automatic
accuracy
accuracy wasControl)
was
encoders, essentially
essentiallyHosting
reduced.
reduced.
stiffness by Elsevier parameter
modeling, Ltd. All rights reserved.
identification,
loading
Keywords: and gravityRobotic on manipulator,
the positioningdouble accuracy was essentially
encoders, reduced.
stiffness modeling, parameter identification,
compliance
Keywords:
Keywords: Robotic error
Robotic compensation.
manipulator,
manipulator, double encoders, stiffness modeling, parameter identification,
Keywords:
compliance Robotic
error manipulator, double
compensation. double encoders,
encoders, stiffness
stiffness modeling,
modeling, parameter
parameter identification,
identification,
compliance
compliance error
Keywords:
compliance Robotic
error
error compensation.
manipulator, double encoders, stiffness modeling, parameter identification,
compensation.
compensation.
compliance 1. error
INTRODUCTIONcompensation. To address this issue, the remainder of the paper is organized
1. INTRODUCTION To
as
To address this
follows.
address this issue,II the
Section
issue, the remainder
defines
remainder of the
the research
of the paper
paper
problem.is organized
is organized
Section
1.
1. INTRODUCTION
INTRODUCTION To
To
as address this
address
follows. this issue,II the
issue,
Section the remainder
remainder
defines the of the
of the paper
research paper
problem.is organized
is organized
Section
Modern trends in robot 1. INTRODUCTIONapplications lead to an essential as III
To
as describes
follows.
address
follows. a
this robotic
Section
issue,
Section II system
defines
the remainder
II defines
defines and
the experimental
research
of
the research the
research paper setup
problem.
problem.is for the
Section
organized
Section
Modern trends in 1. INTRODUCTION
robot applications lead to an essential as
III follows.
describes
stiffness Section
a robotic
parameters II system
identification.the
and experimental
Sectionsproblem.
setup
IV Section
for the
and the
V
revision
Modern of
trends end-user
in robot requirements.
applications Contemporary
lead to an robots
essential III
as
III describes
follows.
describes aa robotic
Section
robotic II system
defines
system and
the
and experimental
research
experimental setup
problem.
setup for
Section
for
Modern
Modern
revision trends
trends
of in robot
in
end-user robot applications Contemporary
applications
requirements. lead to
lead to an an essential
essential
robots III describes
stiffness
highlight a roboticofsystem
parameters
prospective and
identification.
double experimental
encoders Sections
for the setup
IV for the
and
compliance the
V
should
revision
Modern
revision be fast
of
trends
of and
end-user stiff
in robot
end-user enough
requirements. to execute
applications Contemporary
requirements. such technological
lead to an essential
Contemporary robots III
robots stiffness
describes
stiffness parameters
parameters identification.
a robotic system
identification. Sections
and experimental
Sections IV
setup
IV and
for the
and VV
revision
should be of
fast end-user
and stiff requirements.
enough to execute Contemporary
such robots
technological stiffness
highlight
error parameters
prospective
compensation andidentification.
of double
propose encoders Sections
relevant for the
control IV and
compliance
strategy. V
operations
should
revision
should be
be ofas
fast
fast milling
and
end-user
and stiff
stiff(Klimchik
enough
requirements.
enough to
toet al.
execute
execute , 2017,such
Contemporary
such Vosniakos
technological and
robots
technological highlight
stiffness
highlight prospective
parameters
prospective of
of double
identification.
double encoders
encoders for
Sections
for the
the compliance
IV and
compliance V
should
operationsbe fast and
as milling
milling stiff enough
(Klimchik to execute
etexecute
al. ,, (Guillo
2017,such technological
Vosniakos and error highlight
error
Finally, prospective
compensation
Section of
and
VI summaries double
propose encoders
relevant
theencoders
main for the
control
contributions. compliance
strategy.
Matsas,
operations
should
operationsbe2010) as
fast
as and stiff
and
milling friction
(Klimchik
enough
(Klimchik stir welding
toet
et al.
al. , 2017,
2017,such and
Vosniakos Dubourg,
technological
Vosniakos and
and error compensation
highlight prospective
compensation and
and propose
of double
propose relevant
relevant forcontrol
control strategy.
the compliance
strategy.
operations as milling error compensation and propose relevant control strategy.
Matsas,
2016,
Matsas,
operations 2010)
Mendes
2010) and
and
as milling , (Klimchik
friction
et al.friction2014)stir
(Klimchik stir
with etpositioning
welding
welding
al. , (Guillo
2017, accuracy
et al. , (Guillo
Vosniakos
2017, Vosniakos and Dubourg,
and Dubourg, and Finally,
about
and error
Finally,
Finally,
Section VI
Section
compensation
Section
VI summaries
VI
summaries
and propose
summaries
the main
the
the
main
relevant contributions.
contributions.
control strategy.
Matsas,
Matsas,
2016, 2010)
2010)
Mendes and
and
et al. friction
friction
, 2014) stir
stir welding
welding
with positioning (Guillo
(Guillo and Dubourg,
and
accuracy Dubourg,
about Finally, Section 2. VIPROBLEM
summaries the mainmain
STATEMENT contributions.
contributions.
0.1
2016,mm.
Matsas,
2016, Mendes However,
2010)
Mendes et
and al.
et al. in,
friction this
2014)
al. in,, 2014)
2014) applications,
with
stir positioning
welding
with positioning
positioning the
(Guillo robot
accuracy
and
accuracy accuracy
about
Dubourg,
about Finally, Section VI summaries the main contributions.
2016,
0.1 mm.Mendes However, et with
thisprecession
applications, accuracy
thegeometric
robot about
accuracy 2. PROBLEM
PROBLEM STATEMENT STATEMENT
depends
0.1
2016,
0.1
0.1
mm.
mm.
mm.Mendesnot
However,
However,
However,
only
et al.onin
in
in, the this
2014)
this
this
applications,
applications,
applications,
of the
with positioning the
the
the
robot
accuracy
robot
robot about The problem of2.
model
accuracy
accuracy
accuracy
2.stiffness
2. PROBLEM
PROBLEM modelingSTATEMENT
STATEMENT for robotic manipulators
depends
(integrated
depends
0.1 mm. not
not into
However, onlythe
only onin
on the
controller)
the this precession
precession but also
applications, of the
of the on
thegeometric
the
geometric
robot model The problem of2.stiffness
manipulator
model
accuracy PROBLEM modelingSTATEMENT for robotic
robotic manipulators
depends
depends
(integrated not
notinto only
only theon the
oncontroller) precession
the precession of
of the
but compliance
also the geometric
ongeometric
theerrors model
model The
manipulator has
The been in
problem
problem theof
of focus
stiffness
stiffnessof robotic
modeling
modeling community
for
for robotic since the 1980s
manipulators
manipulators
stiffness
(integrated
depends
(integrated that
notinto
into maythe
only provoke
theoncontroller) essential
the precession
controller) but
but also
of the
also on
on the
geometric
the caused
manipulator
model has
manipulator The
has problem
been in theof focus
stiffnessof modeling
robotic for
community robotic sincemanipulators
the 1980s
(integrated
stiffness that into the
maythe controller)
provoke essentialbut also
compliance on the manipulator
errors caused The when
has Salisbury
been in
problem
been in the published
focus
of focus
stiffnessof his
robotic
modeling work on
community the
for robotic active
since stiffness
the
manipulators 1980s
by high that
stiffness
(integrated
stiffness external
that into may
may forces/torques
provoke
controller)
provoke essential
essential (Klimchik
but also on and
compliance
compliance theerrorsPashkevich,
caused
manipulator
errors caused has
when been
control in the
Salisburythe
(Salisbury, focus
publishedof
of robotic
1980). robotic
his
In
community
community
work
the on the
early the since
since
activeonly
models,
the
the 1980s
1980s
stiffnessjoint
stiffness
by
2017,high
Rossi that
external may
et al. provoke
forces/torques
, 2016). essential
As known compliance
(Klimchik from and
the errors caused
Pashkevich,
literature and when
has
when Salisbury
been in
Salisburythe published
focus
publishedof his
robotic
his work
work on
community
on the active
since
active stiffness
the 1980s
stiffness
by
by high
stiffness
high external
that
external may forces/torques
provoke
forces/torques essential (Klimchik
compliance
(Klimchik and
and Pashkevich,
errors caused
Pashkevich, when
control Salisbury
(Salisbury,
elasticities were published
1980).
considered his
In
and work
the
the on
early
stiffnessthe active
models,
parameters stiffness
only joint
were
by
2017,high
Rossi
industrial external et
practice,al. forces/torques
, 2016).
the main As known(Klimchik
sources from and
the
of the Pashkevich,
literature
the literature
manipulator and control
when
control (Salisbury,
Salisbury
(Salisbury, 1980).
published
1980). and In
his
In thethe
work
the early
on
early models,
the active
models, only
stiffness
only were joint
joint
2017,
by
2017, Rossi
high
Rossi external et
et al.
al. ,, 2016).
forces/torques
2016). As
As known
known(Klimchik from
from and
the Pashkevich,
literature and
and control (Salisbury,
elasticities
estimated were
almost 1980).
considered
intuitively In the
(Gosselin, early
stiffnessmodels,
1990, parameters
Pigoski only et joint
al. ,
2017, Rossipractice,
industrial
compliance et al.
are , 2016).
the main
concentrated Asinknown
main sources
the from
actuators of thetheand literature
manipulator
gearboxes and elasticities
control
elasticities were
(Salisbury,
were considered
1980). and
considered In the
and the
the stiffness
early
stiffness parameters
models,
parameters only were joint
were
industrial
2017, Rossi
industrial practice,
et
practice,al. , the
2016).
the As
main sources
known
sources from of
of the
the
the manipulator
literature
manipulator and elasticities
estimated
1998). were
almost
Later, considered
intuitively
there were and the
(Gosselin, stiffness
developed 1990, parameters
morePigoski et
advanced were
al. ,,
industrial practice, the main sources of the manipulator estimated
elasticities almost intuitively
were considered (Gosselin,
and 1990,
the stiffness Pigoski
parameters et al.
compliance
(Deutschmann
compliance
industrial
compliance
are etconcentrated
are
practice,
are
concentrated
al.the
concentrated , 2018,mainin inMagrini
in
the actuators
the
sources
the
actuators
actuatorsetof al. and
theand
and
gearboxes
, 2014).
gearboxes
manipulator
gearboxesThis estimated
estimated
1998).
techniques
almost
almost
Later, intuitively
intuitively
there
allowing thewere (Gosselin,
(Gosselin,
user developed
to
1990,
1990,more
estimate
Pigoski
Pigoski
the etwere
et
advanced
stiffness
al.
al. ,,
compliance
(Deutschmann
motivated are
robot concentrated
et al. ,
manufacturers 2018, in the
Magrini
to actuators
integrate et al. and
double , gearboxes
2014). This
encoders 1998).
estimated
1998). Later,
almost
Later, there
intuitively
there were
were developed
(Gosselin,
developed 1990, more
morePigoski advanced
et
advanced al. ,
(Deutschmann
compliance
(Deutschmann are et
et al.
concentrated
al. ,, 2018,
2018, inMagrini
the actuators
Magrini et
et al.
al. ,, 2014).
and gearboxes
2014). This
This 1998).
techniques
parameters Later, there
allowing
form the thewere
dedicated developed
userexperimental
to estimate
estimate more
study advanced
the(Alici stiffness and
(Deutschmann
motivated
in et al.
robotetmanufacturers
the manipulator's ,
manufacturers
joints 2018, andMagrini
touseintegrate
them et al.
double ,
in , the 2014). This
encoders
feedback techniques
1998).
techniques Later,allowing
there
allowing thewere
the user user to
developed
userexperimental
to estimate more
estimatestudy the stiffness
advanced
the(Alici stiffness
motivated
(Deutschmann
motivated robot
robot al. ,
manufacturers 2018, to
Magrini
to integrate
integrate et double
al.
double encoders
2014). This
encoders techniques
parameters
Shirinzadeh, allowing
form
2005,the the
dedicated
Dumas et al. to
, 2011, the
Klimchik stiffness
et and,
al.
motivated
in
control loop robot
the manipulator's
manipulator'sfor manufacturers
joints and
the compensation andto integrate
use them
of themthe joint double
in the the encoders parameters
feedback
compliances techniques
parameters form
form the
allowing
the dedicated
the userexperimental
dedicated to estimatestudy
experimental studythe(Alici stiffness
(Alici and
and
in
in the
motivated
the robot
manipulator's joints
manufacturers
joints and touseintegrate
use them in
double
in the feedback
feedback parameters
encoders Shirinzadeh,
Shirinzadeh,
2013b). form
Recent 2005,the Dumas
dedicated
contributions et inexperimental
al. ,
this 2011,
area study
Klimchik
allow (Alici
the et
user and
al.to,,,
in the
control
(Tsai et manipulator's
loop
al. , for the
2013). joints
compensation
This and
technology use of them
the
is in
joint
currently the feedback
compliances
used by a parameters
Shirinzadeh, 2005,
form
2005,the Dumas
dedicated
Dumas et
et al.
al. ,
, 2011,
experimental
2011, Klimchik
study
Klimchik (Alici et
et al.
and
al.
control
in the
control loop
manipulator's
loop for
for the
the Thiscompensation
joints
compensation and use of the
them
of isthe joint
in
thecurrently compliances
the
joint compliances feedback
compliances Shirinzadeh,
2013b).
take intoRecent 2005, Dumas
contributions
account both joint et in al. ,
this
and 2011,
area Klimchik
allow
linkallow the
flexibility et
user al.to,,
via
control
(Tsai
number
(Tsai
control et loop
et ofal.robot
al.
loop
for
,,, for
the
2013). compensation
manufactures,
2013).
the This
compensationtechnology
including
technology
of
of isthe
joint
Fanuc
currently
joint and used
Kuka.
used
compliances by aa 2013b).
by Shirinzadeh,
2013b).
2013b).
Recent
Recent
Recent
contributions
2005, Dumas
contributions
contributions et in
inal.this
in this
area
, 2011,
this area
area Klimchik
allow
allow
the
the et
the
user
useral.
user
to
to
to
(Tsai
(Tsai et
et of al.
al.robot 2013).
, 2013). This technology
This technology is currently
is Fanuc
currently used
used by a
by a take take
2013b). into
sophisticated
intoRecentaccount
6×6
account both
stiffness
both
contributions joint
matrices
joint
in this and
instead
and link
link
area of
allowflexibility
classical
flexibility
the uservia via
scalarto
number
number
(Tsai et of robot
al.engineering manufactures,
manufactures,
, 2013). This including
including
technology Fanuc and
and Kuka.
Kuka. take
take into
into account
account both
both joint
joint and
and link
link flexibility
flexibility via
via
number
From
number theof
of robot
robot manufactures,
manufactures, point view, isthis
ofincluding
including currently
Fanuc
Fanuc andused
and
technique Kuka.
Kuka.isbynota sophisticated
sophisticated
coefficients
take into
sophisticated
6×6 stiffness
6×6
account
6×6
stiffness
(Klimchik both
stiffness
matrices
et matrices
al.
joint, 2014a,
matrices
instead
instead
and
instead
of flexibility
classicaletscalar
Pashkevich
link of
of classical
classical
scalar
al.via,
scalar
number
From
quite the
new,of engineering
robot
it wasmanufactures,
engineering point of
successfully ofincluding
view,
used before Fanuc
this in and Kuka.
technique
electric is not sophisticated
not coefficients
drives coefficients
2011). 6×6matrix
stiffness
(Klimchik
This(Klimchik etcanmatrices
al. ,,be2014a, instead
2014a, of classical
Pashkevich
obtained using et etscalar
al. ,,
some
From
From the
the engineering point
point of view,
view, this
this technique
technique is
is not sophisticated
coefficients 6×6 stiffness
(Klimchik et
et al.
matrices
al. instead
,,be2014a, Pashkevich
of classical
Pashkevich et al.
scalar
al. ,
From
quite
for the the
new, engineering
backlashit was compensation.point
successfully of view,
used
In beforethis
robotics, technique
in itelectric
proved is not
drives
to be coefficients
2011). This
approximations (Klimchik
matrix
of the etcan al.
manipulator 2014a, Pashkevich
obtained
links using
(Deblaise et
et al.
al. ,,,
some
quite
From
quite new,
the
new, it
it was
engineering
was successfully
point
successfully ofused
view,
used before
beforethis in
in electric
technique
electric drives
is not
drives 2011).
coefficients
2011). This
This matrix
(Klimchik
matrix can
etcan al. ,be
be obtained
2014a, Pashkevich
obtained using
using et some
al.
some
quite
for
very new,
thenew,
backlashit was successfully
compensation. used before
In robotics,
robotics, in electric
itelectric drives
proved drives
to be
be 2011). This
approximations matrix
of the
the can
manipulator be obtained
links et(Deblaiseusing
(Deblaise some
et some
al.or,,
for
quite
for
for
very theefficient
the
the
backlash
backlash
backlash
efficient
for
it wascompensation.
compensation.
for
compensation
successfully
compensation.compensation
In
used
In before
In robotics, of in
robotics,
of
compliances
it
it proved
it proved to
proved
compliances to be
to
in approximations
be
in
2006),
2011). CAD-based
This matrix
approximations
approximations
2006), CAD-based
of
of the
of
modeling
the can (Klimchik
manipulator
manipulator
manipulator
modeling (Klimchik
links
be obtained al.using
links et(Deblaise
links (Deblaise
al.
, 2013a)
, 2013a)
et
et al.
et al.or,,
al.
manipulator's
very
for theefficient joints.
for
backlash compensation. Nevertheless,
compensation In robotics, of there
compliances is
it proved to beanother in identified
2006),
approximationsfrom
CAD-based the experimental
of themodeling
manipulator data
(Klimchik (Klimchik
links et al. et
(Deblaise al.
,, 2013a) , 2015).
et al.or
very
very efficient
efficient
manipulator's for
for
joints. compensation
compensation
Nevertheless,control of
of compliances
compliances
there is another
another in
in 2006),
2006),
identifiedCAD-based
CAD-based
from thethebasedmodeling
modeling
experimental (Klimchik
(Klimchik
datastiffness et
et
(Klimchik al.
al. ,
et, al.2013a)
2013a)
al. or,
or
perspective
manipulator's
very efficient
manipulator's in double
joints.
for
joints. encoder-based
Nevertheless,
compensation
Nevertheless, of there allowing
compliances
there is
is to go
another in Relevant
identified
2006),
identified models
from
CAD-based
from the on the (Klimchik
experimental
modeling
experimental 6×6data
data (Klimchik
et
(Klimchik matrices
al. et
et ,,, proved
2013a)
al.
2015).
2015).
2015). or
manipulator's
perspective
forward in this joints. encoder-based
in double
double
direction Nevertheless,
and to compensate controlthere
theallowing is another
allowing
errors to go
caused identified
go Relevant
Relevant
to from
models
be verymodels thebased
efficient experimental
on
for on the 6×6
off-line 6×6data (Klimchik
stiffness
implementation, et while
matrices al. , proved
2015).
their
perspective
manipulator's
perspective
perspective
in
in
in joints. encoder-based
double
double Nevertheless,
encoder-based
encoder-based
control
control
controlthere is another
allowing
allowing
to
to
to go
go identified
Relevant
Relevant from
models
models thebased
experimental
based
based on
on
the
the
the 6×6
6×6data stiffness
(Klimchik
stiffness
stiffness
matrices
et al. ,proved
matrices
matrices 2015).
proved
proved
forward
by the linkin this
this direction
inelasticity and
(inand to compensate
compensate
addition tocontrol theallowing
the joint errors caused
compliance
to go to to be very
verymodels
application
be efficient
in online for
mode off-line
is
therather implementation,
difficult. while their
forward
perspective
forward
forward
by the
in
in this
in
link this direction
double
direction
direction
elasticity (in and
and
to
encoder-based
to compensate
to compensate
addition to the
the
the errors
the
joint errors
errors
caused
caused
caused
compliance
Relevant
to
to be very efficient
be very
application efficient
efficient
in
based
online
for
for
for
mode
off-line
on
off-line
off-line
is
6×6
rather
implementation,
stiffness matrices
implementation,
implementation,
difficult.
while
while
while
their
proved
their
their
errors).
by
by the
forward
the This
link
in this
link perspective
elasticity
direction(in
elasticity (in andis explored
addition
to compensate
addition to
to in
the
the this
joint paper
the errors
joint that
compliance
compliance is
caused application
to be very in
application online
efficient
in online mode
for
mode is
off-line
is rather
rather difficult.
implementation,
difficult. while their
by the
errors). link
This elasticity
perspective (in addition
isaddition
explored to the
in joint
this compliance
paper that At present,
application inthe most
online mode promising
is application in online mode is rather difficult. is rather technique
difficult. for the online
aimed
errors).
by the
errors). atlink
reduction
This
This elasticity of the
perspective
perspective (in compliance
is
is explored
explored to errors
in
the
in this produced
joint
this paper by all
that
compliance
paper that is
is At present, the most promising technique for the online
errors).
aimed atThis
reduction perspective
of the is explored errors
the compliance
compliance in thisproduced paper that is At
by all
all compensation
At present,
present, of the
the most compliance
promising errors is
technique based on
for the double
online
manipulator
aimed
errors).
aimed at
atThis components.
reduction
reduction of
perspective
of the is explored errors
the compliance
compliance in thisproduced
errors paper that
produced by At
is compensation
by all
all present,In the
compensation
encoders. the most
most
of the
the
contrast
promisingerrors
topromising
compliance
the standard
technique
technique basedfor
isapproach,foron thethe
the
where
online
online
double
the
aimed at
manipulator reduction
components. of errors produced by At present,
compensation of
the
of most
the compliance
promising
compliance errors is
technique
errors is based
based foron
on the
the double
online
double
manipulator
aimed at components.
reduction
manipulator components.
components. of the compliance errors produced by all compensation
encoders.
joint positionIn of the
contrast
feedback compliance
tois the errors
standard
implemented is based
approach,
using on
the the
where double
motor-side the
manipulator encoders.
compensation
encoders. In contrast
of the
In contrast
contrast to the
compliance
tois the
the standard
errors isapproach,
standard approach,based
approach, on where
the
where the the
double
the
manipulator components. encoders.
joint positionIn feedback to standard
implemented using the where
motor-side
joint
joint position
encoders.
position feedback
feedbacktois
In contrast implemented
is the using
using the
standard approach,
implemented the motor-side
where the
motor-side
joint position feedback is implemented using the motor-side
2405-8963 ©
Copyright © 2018,
2018 IFACIFAC (International Federation of Automatic Control) joint position
747Hosting by Elsevierfeedback
Ltd. Allisrights
implemented
reserved. using the motor-side
Copyright
Peer review©
Copyright © 2018 responsibility
under IFAC 747
of International Federation of Automatic Control.
Copyright
Copyright © 2018
© 2018 IFAC
2018 IFAC
IFAC
10.1016/j.ifacol.2018.08.407
747
747
747
Copyright © 2018 IFAC 747
IFAC INCOM 2018
Bergamo, Italy, June 11-13, 2018 A. Klimchik et al. / IFAC PapersOnLine 51-11 (2018) 740–745 741

encoder, a new technique allows using additional information To use benefits of the double encoders, all experiments were
from the secondary encoder mounted on the manipulator repeated for two feedback modes in the joint position control,
link-side. It is clear that the difference between the motor- i.e. using either the primary or secondary encoders in the
side and link-side encoder readings provide additional closed loop. It is worth mentioning that for a conventional
information allowing essentially improve the robot robot with a single encoder (on the motor-side), each
positioning accuracy. Usually, the double encoder schema is configuration allows to get two sets of experimental data
used to compensate the joint compliance errors (caused by corresponding to the unloaded and loaded modes (Dumas et
the actuator and gear elasticity). In this paper, we address a al. , 2012, Dumas, Caro, 2011, Klimchik, Ambiehl, 2017,
more general problem: to compensate, using the double Klimchik, Wu, 2013b). In contrast, robots with double
encoders, the compliances errors caused by both joint and encoders provide us the additional possibility to carry out the
link elasticities. same experiments with joint feedback on the primary (motor-
On the other side, the double encoders provide with side) and secondary (link-side) encoders. In addition, in all
experiments, it is possible to record both values, obtained
additional information that may be useful for the stiffness
from the primary and secondary encoders. It is clear that this
model identification. This creates a related problem of
identification the stiffness model parameters using the double additional information gives us some benefits for
identification of the manipulator stiffness parameters; it will
encoders, which is also addressed in this paper. In general,
be addressed in the following Section.
the goal of this paper is to investigate and estimate potential
benefits of double encoders in industrial robots for the Thus, for each robot configuration there are 4 types of the
manipulator stiffness modeling and accuracy improvement by measurement modes corresponding to possible combinations
means of intelligent control. of the manipulator loading and feedback control:
U1: unloaded calibration experiment with primary encoder
feedback (without external loading and without joint error
3. EXPERIMENTAL SETUP compensation). It provides with the following data: (i) the
j
end-effector locations p U1, i , (ii) the data of the primary
The robotic system under study is presented in Fig. 1. It encoder q U1,i , (iii) the data of the secondary encoders q (U1,i
(1) 2)
.
consists of a robot, an end-effector location measurement
unit, and equipment for external force/torque generation and U2: unloaded calibration experiment with secondary encoder
estimation. In more details, the system includes the following feedback (without external loading and with joint error
components: compensation). It provides with the following data: (i) the
end-effector locations p Uj 2,i , (ii) the data of the primary
• Robot manipulator with double encoders at each encoder q (1) (2)
U2,i , (iii) the data of the secondary encoders q U 2,i .
actuated joint;
L1: loaded calibration experiment with primary encoder
• Special end-effector equipped with several reference feedback (with external loading Fi and without joint error
markers for position measurement; compensation). It provides with the following data: (i) the
• Additional equipment allowing to generate and end-effector locations p L1, j
i , (ii) the data of the primary
estimate external loadings; encoder q L1,i , (iii) the data of the secondary encoders q(2)
(1)
L1,i .

• Laser tracker for measuring the reference-point L2: loaded calibration experiment with secondary encoder
positions (such as Leica). feedback (with external loading Fi and with joint error
compensation). It provides with the following data: (i) the
To identify the stiffness model parameters, two types of
end-effector locations p Lj 2, i , (ii) the data of the primary
experiments were carried out: with and without external
encoder q(1) ( 2)
L2i , (iii) the data of the secondary encoders q L 2,i .
loading, which can be either vertical or non-vertical one
(almost horizontal). It should be stressed that in order to increase the calibration
accuracy, the measurements for all 4 modes should be
executed without moving the robot. It allows us to reduce the
influence of the robot limited repeatability on the
experimental data. Besides, to reduce the impact of the error
compensation, the calibration experiments must be carried
out in the following order: U1, L1, L2, U2. The latter is
motivated by the possible existence of some hysteresis in
on/off switching of the compliance error compensation
algorithm based on the secondary encoder feedback.
In the presented study for the experimental part model of
industrial robot KUKA KR120 R3900 with its true geometric
parameters and joint limits is used. Its stiffness model
parameters (simplified stiffness model) were obtained in our
previous study (Klimchik, Ambiehl, 2017), which will be
used here as an estimate for the joint compliances. Link
stiffness models are approximated by hollow beams with
cross-sections closed to real robot geometry.
Fig. 1. Experiment setup for elastic parameters identification

748
IFAC INCOM 2018
742
Bergamo, Italy, June 11-13, 2018 A. Klimchik et al. / IFAC PapersOnLine 51-11 (2018) 740–745

4. IDENTIFICATION OF STIFFNESS MODEL on the primary encoders (L1) and secondary encoders (L2)
PARAMETERS USING DOUBLE ENCODERS depends on the joint compliances, the gravity loading, and
external loading. Hence, these data allow us to identify the
Stiffness model parameters identification from end- joint compliances only if the robot link masses and their mass
effector measurements centers are known.
In this case, the desired stiffness parameters are identified by The experimental results for the robot with double encoders
approximating the linear functions describing force- allow the user to estimate the joint k J and link k L
deflection relations for each manipulator configurations. The compliance as well as their sum k J  L using the following
identification technique remains the same for robots with expressions
both single and double encoders. The relevant procedure 1
includes two steps (Klimchik et al. , 2014b). At the first step,  m n 
k J    Bij ( p ) Bij ( p )  
T

the base pbase , φbase and tool u1tool , utool


2 3
, utool parameters are
 i 1 j 1 
identified using the difference p g ,i between the coordinates (4)
extracted from the classical direct geometric transformations  m n 
   Bij ( p ) (p L2
T

i  p U2, i  p L1i  p U1, i ) 


j j j j
and the experiment results in the unloaded mode:
 i 1 j 1 
pbase ; φbase ; u1tool ; uto2 ol ; u3tool   1
 m n   m n 
k L    Bij ( p ) Bij ( p )    Bij ( p ) (pL2,
T T

i  p U2,i ) 
j j
1 (1) (5)
 m   m   i 1 j 1   i 1 j 1 
   Aij Aij    Aij p g ,i 
T T

 i 1   i 1  1
 m n   m n 
k J  L    Bij ( p ) Bij ( p )    Bij ( p ) (pL1,
T T

i  p U1,i ) 
j j
At the second step, the stiffness parameter vector k is (6)
identified using measured end-effector deflections p ej,i  i 1 j 1   i 1 j 1 
caused by the external loading (for each measurement point) It should be mentioned that here only one component of link
1 compliance is taken into account that corresponds to bending
 m n   m n 
k    Bij ( p ) Bij ( p )    Bij ( p ) pej,i  around the adjacent axis joint. It should be also stressed that
T T
(2)
 i 1 j 1   i 1 j 1  for all identification equations (4)-(6) the observation
matrices are the same and depend on the external loading
Here A ij and Bij ( p ) are the observation matrices, that are applied to the manipulator end-effector.
described in details in our previous work (Klimchik, Wu,
2014b). It is worth mentioning that the matrix A ij depends Hence, the above-presented methodology allows us to
on the geometric parameters only, while the matrix Bij ( p ) estimate separately the joint and link compliances and
depends also on the applied external loading Fi evaluate their impact on the end effector positioning
accuracy. It is worth mentioning that the identification
B j ( p )  procedure developed for conventional industrial robots with
Bij   J1ji J1jiT Fi ,..., J nij J nij T Fi  ; Bij   ij ( )  (3)
 Bi  single encoders (Dumas, Caro, 2011, Klimchik, Ambiehl,
2017) allowed the user to identify the sum of joint and link
where the vectors J1ji ,..., J nji are the columns of kinematic
compliances only, i.e. k J  L . This important contribution
Jacobian.
(separate identification of the link and joint compliances)
will be effectively used below for the on-line error
compensation.
Stiffness model parameters identification from joint
encoders measurements
For the robots with double encoders, there are additional data
for the identification of the stiffness model parameters:
primary and secondary encoders readings. In this case, the
basic equation for the identification can be written as follows
k J qi  JTi  Fi (7)
where qi is a vector of angular deflections due to the
external force Fi applied to the end-effector, and J i is
kinematic Jacobian for the ith measurement configuration. It
is clear that here only the joint compliance can be identified
Fig. 2. A two-link serial system with a passive joint using the usual least square expression
1
It should be mentioned that for robots with double encoders  m  m 
the above relations can be used in several ways in order to k J    J Ti  Fi  qTi   qTi  qi  (8)
produce different sets of the stiffness parameters. In fact, here  i 1  i 1 
the force-deflection relation) essentially depends on the It should be stressed that if link masses are not negligible
measurement mode and allows to identify different (and provoke essential compliance errors) definition of qi
parameters. All possible cases of the compliance errors and should be clarified. In this case, it is prudent to apply
their physical meaning for four measurement modes are equations based on differences of the secondary encoder
presented in Fig. 2 For example, the difference between the measurements for experiments L1 and U1, i.e.
measurements in the loaded mode using the feedback based qi  q(2)
L1  q U1 .
(2)

749
IFAC INCOM 2018
Bergamo, Italy, June 11-13, 2018 A. Klimchik et al. / IFAC PapersOnLine 51-11 (2018) 740–745 743

5. CONTROL STRATEGIES AND COMPLIANCE ERROR a combination of the primary and secondary encoders that
COMPENSATION FOR ROBOTS WITH DOUBLE provides the difference between the angles on the motor-side
ENCODERS and on the link-side. If the joint stiffness is known, this
difference provides another way of the external force
Robotic manipulators with double encoders provide us with estimation.
two options for control of the actuated axes: using feedback
either from primary or secondary encoders. It is clear that the
second approach is more attractive because it allows us to
compensate directly the errors caused by the joint
compliances. However, the link compliances are outside of
such error compensation strategy. The latter is confirmed by
experimental results presented in Fig. 3, where the secondary
encoder feedback compensates less than 2/3 of the end-
effector compliance errors. Moreover, in the case of analysis
in the entire robot workspace, the efficiency of the
compliance error compensation based on the second encoder
feedback for the worst case reduced down to 50% (Fig. 3).
Results presented in Table III show that accuracy estimation
based on the experimental set (configuration used for the
parameters identification) gives slightly better results than
analysis within entire robot workspace, while this difference
is important for the maximum compliance errors. (a) Deflections without compensation

Deflections with
compensation based
on stiffness model

Deflections with
compensation based
on second encoder

Deflections without
compensation

Positioning errors, mm
(a) Deflection histograms for the experimental set

Deflections with
compensation based (b) Deflections with compensation based on second encoder
on stiffness model

Deflections with
compensation based
on second encoder
Deflections without
compensation

Positioning errors, mm
(b) Deflection histograms the entire robot workspace

Fig. 3. Efficiency of embedded and proposed compliance


error compensation algorithms for the experimental set (a)
and for entire robot workspace (b) (results corresponds to the
model of industrial robot Kuka KR120 R3900 under external
loading 100kg) (c) Deflections with compensation based on stiffness model

It is worth mentioning that similar results (i.e. compensation Fig. 4. End-effector deflections caused by external loading
the joint compliances) can be achieved using the primary 100kg for the robot Kuka KR120 R3900 with primary (a) and
encoder only. In this case, the manipulator should be
secondary (b) encoder feedback and (c) compliance error
equipped with a force/torque sensor allowing to estimate
compensation based on stiffness model
external loading. Using this data and assuming that the
manipulator stiffness model is known, it is possible to modify To compensate both joint and link compliances, the
the actuator inputs to compensate the above-mentioned following strategy can be applied. First, using the difference
compliance errors. On the other side, the desired external between the primary and secondary encoder as well as the
force/torque measurement can be also done indirectly, using joint stiffness coefficients, it is estimated the external

750
IFAC INCOM 2018
744
Bergamo, Italy, June 11-13, 2018 A. Klimchik et al. / IFAC PapersOnLine 51-11 (2018) 740–745

loading. Further, using the complete stiffness model (that twice better than the error compensation algorithm used by
includes both joint and link compliances), the actuator inputs the robot manufacturers (see Table I and Fig. 3 and 4c)
are modified to compensate both types of the compliance
errors. It should be mentioned that here the feedback relies on Table 1 Comparison of error compensation algorithms based
the primary encoder, while the secondary encoder is used to on experimental set and within entire robot workspace
measure the elastic deflections. As follows the dedicated
study, this control strategy allowed to compensate about 4/5 Compensation End-effector errors, mm
of the end-effector deflections, which is more than twice algorithm
better compared to the embedded error compensation mean std max
algorithm based on the secondary encoder feedback (see Experimental set
Figure 3c). More details concerning the efficiency of the
4.03 2.17 9.52
proposed error compensation technique and comparison with Primary encoder
the methods that are currently used in industry with the entire (100%) (100%) (100%)
robot workspace are given in Fig. 4 and Table 1. 1.74 0.93 3.36
Secondary encoders
Taking into account that enhanced stiffness model is rather (43.1%) (42.7%) (35.2%)
difficult for industrial implementation, an alternative control
strategy can be used. Its main idea is based on the 0.64 0.33 1.52
Stiffness model
proportionality of the errors caused by the joint and link (15.7%) (15.3%) (15.9%)
compliances. This allows us to replace the conventional Entire robot workspace
feedback coordinate q (2) from the secondary encoder by the
linear combination q (2)    (q (2)  q (1) ) computed using both 4.03 2.28 9.17
Primary encoder
encoder readings. In this expression, the coefficient (100%) (100%) (100%)
  kL / k J considers the ratio between the link to joint 1.81 0.79 4.63
compliances, which differs from joint to joint. As follows Secondary encoders
from the simulation study, this industry-oriented strategy is (44.7%) (44.8%) (50.4%)
slightly worth compared to the above presented one, but the
compensation degree is quite acceptable for practice: it is

Table 2 Comparison of error compensation algorithms based on single and double encoder schemes
End-effector errors,
Scheme Compensation algorithm mm Comment
mean max
Build-in control algorithms (compensation based on joint feedback)
Controller built-in compensation 4.03 9.17 Manipulator control using conventional
Single encoder
OFF (100%) (100%) feedback (1)
Controller built-in compensation Built-in algorithm based on the second
Double 1.81 4.63
ON encoder feedback (2) compensates 1/2 of
encoders (44.7%) (50.4%)
compliance errors
Control algorithms based on reduced stiffness model (compensation of joints deflections)
Controller input adjustment by Degree of compensation corresponds to the
Double   2  1 ; 1.81 4.63 built-in compensation algorithm based on
encoders Evaluation of the end-effector (44.7%) (50.4%) the second encoder feedback
deflections using geometric model
Controller input adjustment by Degree of compensation corresponds to the
Single encode
  k J  ; 1.81 4.63 built-in compensation algorithm based on
and force
Evaluation of the end-effector (44.7%) (50.4%) the second encoder feedback
sensor
deflections using stiffness model
Control algorithms based on enhanced stiffness model (compensation of joints and links deflections)
Controller input adjustment by Degree of compensation is twice better
Double   k (2  1 ) ; 1.08 2.28 compared to the second encoder feedback,
encoders Evaluation of the end-effector (19.5%) (24.7%) algorithm is able to compensate 4/5 of
deflections using geometric model compliance errors
Controller input adjustment by Degree of compensation is twice better
Single encoder
  (k J  kL )  ; 1.08 2.28 compared to the second encoder feedback,
and force
Evaluation of the end-effector (19.5%) (24.7%) algorithm is able to compensate 4/5 of
sensor
deflections using stiffness model compliance errors

751
IFAC INCOM 2018
Bergamo, Italy, June 11-13, 2018 A. Klimchik et al. / IFAC PapersOnLine 51-11 (2018) 740–745 745

The control strategies considered in this section are ACKNOWLEDGMENTS


summarized in Table 2, which contains both conventional
built-in algorithms and proposed in this work. It clearly The work presented in this paper was supported by the grant
shows that the existing compensation algorithm (embedded of Russian Science Foundation №17-19-01740.
in the robot controller) allows reducing the robot positioning REFERENCES
errors from 4.03 mm to 1.81 mm, while the proposed
technique allows achieving further error reduction down to Alici G, Shirinzadeh B. Enhanced stiffness modeling, identification and
1.08 mm. Its simplified version (easy for industrial characterization for robot manipulators. Robotics, IEEE Transactions
on. 2005;21:554-64.
implementation) ensures the same error reduction. Some Deblaise D, Hernot X, Maurine P. A systematic analytical method for PKM
control strategies ensure the same degree of the compliance stiffness matrix calculation. IEEE International Conference on
error compensation. This fact confirmed by mathematical Robotics and Automation (ICRA 2006): IEEE; 2006. p. 4213-9.
equivale between the model used for error compensation, Deutschmann B, Liu T, Dietrich A, Ott C, Lee D. A Method to Identify the
while in the real industrial environment their efficiency may Nonlinear Stiffness Characteristics of an Elastic Continuum
Mechanism. IEEE Robotics and Automation Letters. 2018;3:1450-7.
differ since they have different inputs and measurement noise Dumas C, Caro S, Cherif M, Garnier S, Furet B. Joint stiffness identification
will affect differently of the manipulator positioning of industrial serial robots. Robotica. 2012;30:649-59.
accuracy. Hence, for the robot with double encoder, essential Dumas C, Caro S, Garnier S, Furet B. Joint stiffness identification of six-
improvement of the robot accuracy can be achieved by revolute industrial serial robots. Robotics and Computer-Integrated
simple modification of the feedback signal in the actuators. Manufacturing. 2011;27:881-8.
Gosselin C. Stiffness mapping for parallel manipulators. Robotics and
The presented study is based on the particular robot model, Automation, IEEE Transactions on. 1990;6:377-82.
but the proposed advance control (error compensation) Guillo M, Dubourg L. Impact & improvement of tool deviation in friction
algorithms will have similar effects on the other robots also. stir welding: Weld quality & real-time compensation on an industrial
robot. Robotics and Computer-Integrated Manufacturing. 2016;39:22-
It should be mentioned that the degree of compliance error 31.
compensation based on the second encoder feedback depends Klimchik A, Ambiehl A, Garnier S, Furet B, Pashkevich A. Efficiency
on the robot geometry and gearboxes compliances, i.e. by evaluation of robots in machining applications using industrial
ratio   kL / k J indicating link compliance impact in the performance measure. Robotics and Computer-Integrated
total robot compliance. This hypothesis has been confirmed Manufacturing. 2017;48:12-29.
Klimchik A, Chablat D, Pashkevich A. Stiffness modeling for perfect and
in the experimental studies with robots from different non-perfect parallel manipulators under internal and external loadings.
manufacturers. In all cases, the stiffness model-based Mechanism and Machine Theory. 2014a;79:1-28.
compliance error compensation technique allowed to improve Klimchik A, Furet B, Caro S, Pashkevich A. Identification of the
robot positioning accuracy compared to the built-in algorithm manipulator stiffness model parameters in industrial environment.
and was able to compensate more than 80% of compliance Mechanism and Machine Theory. 2015;90:1-22.
Klimchik A, Pashkevich A. Serial vs. quasi-serial manipulators: Comparison
errors. analysis of elasto-static behaviors. Mechanism and Machine Theory.
2017;107:46-70.
Klimchik A, Pashkevich A, Chablat D. CAD-based approach for
identification of elasto-static parameters of robotic manipulators. Finite
6. CONCLUSIONS Elements in Analysis and Design. 2013a;75:19-30.
Klimchik A, Wu Y, Caro S, Furet B, Pashkevich A. Geometric and
For many new application domains, the issue of the robot elastostatic calibration of robotic manipulator using partial pose
positioning accuracy improvement under external loading measurements. Advanced Robotics. 2014b;28:1419-29.
becomes a key subject. Current achievements in this area rely Klimchik A, Wu Y, Dumas C, Caro S, Furet B, Pashkevich A. Identification
mainly on the double-encoder technique, that replaces the of geometrical and elastostatic parameters of heavy industrial robots.
conventional motor-side feedback by the link-side feedback. IEEE International Conference on Robotics and Automation (ICRA)
2013b. p. 3707-14.
However, this simple and straightforward solution allows Magrini E, Flacco F, Luca AD. Estimation of contact forces using a virtual
compensating only part of the compliance errors caused by force sensor. 2014 IEEE/RSJ International Conference on Intelligent
the external force/torque (about a half for the manipulator Robots and Systems2014. p. 2126-33.
under study). Mendes N, Neto P, Simão MA, Loureiro A, Pires JN. A novel friction stir
welding robotic platform: welding polymeric materials. The
To utilize the full potential of the double-encoder schema, International Journal of Advanced Manufacturing Technology. 2014:1-
this paper proposes a new control strategy, where the actuator 10.
feedback is created via a linear combination of the primary Pashkevich A, Klimchik A, Chablat D. Enhanced stiffness modeling of
manipulators with passive joints. Mechanism and machine theory.
and secondary encoders signals. Relevant coefficients are 2011;46:662-79.
computed using the manipulator enhanced stiffness model, Pigoski T, Griffis M, Duffy J. Stiffness mappings employing different
which is obtained using the developed identification frames of reference. Mechanism and machine theory. 1998;33:825-38.
technique. It allowed essentially improve the robot Rossi R, Fossali L, Novazzi A, Bascetta L, Rocco P. Implicit force control
positioning accuracy and to achieve the compliance error for an industrial robot based on stiffness estimation and compensation
during motion. 2016 IEEE International Conference on Robotics and
reduction ratio about 4/5. The efficiency of proposed control Automation (ICRA)2016. p. 1138-45.
strategy was confirmed in the experimental study. Salisbury JK. Active stiffness control of a manipulator in Cartesian
In future, for further accuracy improvement, the proposed coordinates. Decision and Control including the Symposium on
Adaptive Processes, 1980 19th IEEE Conference on: IEEE; 1980. p.
model will be enhanced to take into account directly 95-100.
influence of the gravity forces applied to the manipulator Tsai J, Wong E, Tao J, McGee HD, Akeel H. Secondary position feedback
components. control of a robot. Google Patents; 2013.
Vosniakos G-C, Matsas E. Improving feasibility of robotic milling through
robot placement optimisation. Robotics and Computer-Integrated
Manufacturing. 2010;26:517-25.

752

You might also like