9th IFAC Conference on Manufacturing Modelling, Management and

9th
9th IFAC Conference on Manufacturing Modelling, Management and
9th IFAC
IFAC Conference
Control Conference on
on Manufacturing
Manufacturing Modelling,
Modelling,
Available Management
Management
online and
and
at www.sciencedirect.com
Control
Control
Berlin,
9th IFACGermany, August
Conference on 28-30, 2019
Manufacturing Modelling, Management and
Control
Berlin, Germany,
Berlin, Germany,
Control August
Germany, August 28-30,
August 28-30, 2019
28-30, 2019
2019
Berlin,
Berlin, Germany, August 28-30, 2019 ScienceDirect
IFAC PapersOnLine 52-13 (2019) 1860–1865
Assembly Line Balancing with Collaborative Robots under
Assembly
Assembly Line
Line Balancing
Balancing with
with Collaborative
Collaborative Robots
Robots under
under
consideration
Assembly Line of Ergonomics:
Balancing with a cost-oriented
Collaborative approach
Robots under
consideration of Ergonomics: a cost-oriented
consideration of Ergonomics: a cost-oriented approach approach
considerationChristian
of Ergonomics:
Weckenborg*,aThomas
cost-oriented
S. Spengler approach
Christian
Christian Weckenborg*,
Christian Weckenborg*, Thomas
Weckenborg*, Thomas S.
Thomas S. Spengler
S. Spengler
Spengler
Christian
Institute of Automotive Management and Industrial Production, Technische Weckenborg*, Thomas S. Spengler
Universität Braunschweig (GERMANY),
Institute
Institute of Automotive Management and Industrial Production, Technische Universität Braunschweig (GERMANY),
Institute of of Automotive
Mühlenpfordtstr. 23, 38106
Automotive Management
Braunschweig,
Management and
and Industrial
Germany Production,
Industrial Technische
Technische Universität
(Tel: 0049-531-3912207;
Production, Braunschweig
e-mail: {c.weckenborg,
Universität (GERMANY),
Braunschweigt.spengler}@tu-
(GERMANY),
Mühlenpfordtstr.
Institute of
Mühlenpfordtstr. 23, 38106
Automotive
23, 38106 Braunschweig,
Management
Braunschweig, and Germany Production,
Industrial
Germany (Tel: 0049-531-3912207;
(Tel: 0049-531-3912207;
Technische e-mail: {c.weckenborg,
Universität
e-mail: {c.weckenborg,
Braunschweig t.spengler}@tu-
(GERMANY),
t.spengler}@tu-
Mühlenpfordtstr. 23, 38106 Braunschweig, Germany braunschweig.de).
(Tel: 0049-531-3912207; e-mail: {c.weckenborg, t.spengler}@tu-
Mühlenpfordtstr. 23, 38106 Braunschweig, Germany braunschweig.de).
(Tel: 0049-531-3912207; e-mail: {c.weckenborg, t.spengler}@tu-
braunschweig.de).
braunschweig.de).
braunschweig.de).
Abstract: The role of automation in modern manufacture experienced increasing relevance throughout
Abstract: The
Abstract: The role role of of automation in in modern
modern manufacture
manufacture experienced
experienced increasing
increasing relevancerelevance throughout
throughout
past decades.
Abstract: role of automation
The However, manufacturers
automation cannot manufacture
in modern efficiently automate
experiencedmanyincreasing
tasks to date and consequently
relevance throughout
past decades.
past decades.
Abstract: The However,
role
However, of manufacturers
automation
manufacturers in cannot manufacture
modern
cannot efficiently automate
efficiently automate
experiencedmanyincreasing
many tasks to
tasks to date
date and consequently
relevance
and consequently
throughout
utilize
past humanHowever,
decades. workers, manufacturers
which are subject cannot to efficiently
limited physical
automate capacity.
many tasksIn recent to date years, collaborative
and consequently
utilize
past
utilize humanHowever,
decades.
human workers, manufacturers
workers, which are
which are subject
subject
cannot to efficiently
limited physical
physical
automate capacity.
many tasksIn recent
recent
to date years,
and collaborative
consequently
robots
utilize can
human support manual
workers, task execution
which are subject at tolow
to limited
costs. physical
limited This novelcapacity.
technology
capacity. In may
In recent years,
thus collaborative
represent
years, a cost-
collaborative
robots
utilize
robots can
human
can support
support manual
workers,
manual task
which
task execution
are subject
execution at
at low
to
low costs.
limited
costs. This
physical
This novel
novel technology
capacity.
technology In may
recent
may thus represent
years,
thus aa cost-
collaborative
represent
efficient
robots can opportunity
support manual of (partial) automation
task execution of assembly
at low costs. This tasks. Wetechnology
novel consider an may assembly line
thus represent a cost-
balancing cost-
efficient
robots can opportunity of (partial) automation of assembly tasks. Wetechnology
consider an assembly line balancing
efficient
problem
efficient tosupport
opportunity
determine
opportunity manual
of task execution
(partial)
ofcost-efficient
(partial) automation
system
automation at low
of costs. This
assembly
configurations,
of assembly novel
tasks.
in We
tasks. which
We consider an may
collaborative
consider an thus
assembly
assembly represent
robots line ahuman
balancing
lineand
balancing cost-
problem
efficient to determineofcost-efficient
opportunity (partial) system
automation configurations,
of assembly in
tasks. which collaborative robots and human
problem
workers may
problem to
to determine
be assigned
determine cost-efficient
to the stations
cost-efficient system
systemof the configurations,
line. We develop
configurations, in We
in whichconsider
a mathematical
which an assembly
collaborative
collaborative model robots
and
robots line balancing
and
illustrate
and human
human its
workers
problem
workers may
to
may be
be assigned
determine
assigned to
to the
cost-efficient
the stations
system
stations of
of the
the line.
line. We
configurations,
We develop
in
develop a mathematical
which
a collaborative
mathematical model
model and
robots
and illustrate
and
illustratehuman its
its
functionality using a real-world oriented case study. Subsequently,
workers may be assigned to the stations of the line. We develop a mathematical model and illustrate its we identify the determining factors of
functionality
workers using aa real-world oriented case thestudy.
line. Subsequently, a we identify the determining factors of
the use ofmay
functionality
functionality be assigned
using
collaborative
using to theThe
a real-world
robots.
real-world stations
oriented
results
oriented of demonstrate
case
case study.
study. Wethat
develop
Subsequently, mathematical
we
collaborative
Subsequently, we identify
robots
identify the model
theare and illustrate
determining
an economic
determining factors
option
factors its
of
of
the use of of collaborative
functionality
the use using robots. The
a real-world The results
oriented casedemonstrate that collaborative
study. Subsequently, collaborative robots
we identify are
theare an economic
economic
determining option
factors of
the use of collaborative
for automation in manual
collaborative robots.
manufacture.
robots. The results
results demonstrate
demonstrate that
that collaborative robots
robots are an
an economic option
option
for automation
the
for automation
use of in manual
in manualrobots.
collaborative manufacture.
The
manufacture. results demonstrate that collaborative robots are an economic option
for automation in manual manufacture.
© 2019,
Keywords:
for IFAC
automation
Keywords:
assembly
in manual
assembly
line balancing,
(International Federation
manufacture.
line balancing,
balancing,
cost
cost
orientation,
of orientation,
Automatic Control)ergonomics,
Hosting
ergonomics,
collaborative
by Elsevier Ltd.
collaborative
robots,
robots,
industry
Allindustry 4.0.
rights reserved.
4.0.
Keywords: assembly line balancing, cost
Keywords: assembly line cost orientation,
orientation, ergonomics,
ergonomics, collaborative
collaborative robots, robots, industry
industry 4.0. 4.0.
Keywords: assembly line balancing, cost orientation,  ergonomics, collaborative robots, industry 4.0.


 its operation (BMWi, 2017). Though most articles in ALB
1. INTRODUCTION AND BACKGROUND its
its operation (BMWi,
1. INTRODUCTION AND BACKGROUND  its operation
literature
operation (BMWi,
contribute
(BMWi, to 2017).
2017).
the design
2017). Though
Though
Though most
most articles
of manual
most articles
assembly
articles in
in ALB
in lines,
ALB
ALB
Over the1. INTRODUCTION
1. past
INTRODUCTION
decades, the role AND
AND ofBACKGROUND
BACKGROUND
automation in industrial its literature contribute
operation (BMWi,
literature contribute
contribute to the
2017).
to the
the designThough
design of of manual
of are most
manual assembly
articles
assembly in lines,
ALB
lines,
Over the
Over the1. past
INTRODUCTION
past decades, the the role AND
role ofBACKGROUND
automation in industrial approaches considering ergonomics are scarce. The first to
in industrial approaches
literature considering to ergonomics
design manualscarce. The
assembly first
lines,
production
Over the past hasdecades,
decades, role of
increasedthesignificantly. of automation
In the current
automation state, literature
in industrial approaches
explicitly
approaches contribute
considering
incorporate
considering toergonomic
the designrisk
ergonomics
ergonomics of are
manual
are scarce.
estimation
scarce. assembly
The first
methods
The first to
lines,
to
into
to
production
Over
however,the past
production
production has increased
hasdecades,
increasedthe
manufacturers
has increased significantly.
role of
significantly.
cannot efficiently
significantly. In
In the
automation
In the current
automate
the current state,
in industrial
current state, approaches
many
state, explicitly incorporate
considering ergonomic
ergonomics risk estimation
are scarce. methods
The firstinto
to
explicitly
ALB
explicitly incorporate
models were
incorporate Otto ergonomic
and
ergonomic Scholl risk
risk estimation
(2011), who
estimation methods
proposed
methods into
the
into
however,
production
however, manufacturers
has increased
manufacturers cannot
cannot efficiently
significantly. valueInadded
efficiently automate
the current
automate many
state, explicitly
many ALB
tasks.
however, Therefore, a large part
manufacturers cannotof the efficiently is generated
automate many ALB models
ALB models
application
models ofwere
incorporate
were
were Otto
Otto
several
Otto riskand
ergonomic Scholl
andestimation
and Scholl
Scholl risk(2011),
(2011), who
estimation
methods.
(2011), who proposed
who methodsA surveyinto
proposed
proposed the
the
of
the
tasks.
however,
tasks.
by manual
tasks. Therefore,
Therefore, a
manufacturers
a large
large
work autilizing
Therefore, large partpart
cannot
part of
of
human the value
efficiently
the value
of theworkers
value addedadded is
automate
added is generated
characterized many
generated
is generated by application
ALB models of several
were Otto riskand estimation
Scholl methods.
(2011), who A survey
proposed of
the
application
subsequently
application of
of several
published
several risk
risk estimation
literature in
estimation this methods.
field
methods. is A
given
A survey
by
survey Ottoof
of
by
by manual
tasks.
manual work
Therefore,
work autilizing
large
utilizing parthuman
of
human the workers
value
workers characterized
added is generated
characterized by
by subsequently published literature in this field is given by Otto
flexibility
by manualand workadaptability
utilizing (Antonelli
human workers 2016; Krüger by
et al., characterized et application
subsequently of
and Battaïa (2017).
subsequently several
published
published risk estimation
literature
Recent
literature in
literaturethis methods.
field
in this extends is
field is given A
given survey
by
the problem Otto
by Otto of
flexibility
by 2009).and
manual
flexibility In adaptability
work
and utilizing
adaptability (Antonelli
human novelworkers
(Antonelli et
et al., 2016;
2016;toKrüger
al., characterized
Krüger et
by
et subsequently
and
al.,
flexibility and recent
adaptability years, (Antonelli trends
et al., arise
2016; support
Krüger et and Battaïa
towards
and Battaïa
more
Battaïa (2017).
published
(2017).
detailed
(2017). Recent
literature
Recent
description
Recent literature
in of
literature
literature extends
thisergonomic
field is given
extends
extends the
the
risk
the problem
by Otto
problem
(Xu
problem et
al., 2009).
flexibility
2009).
al., 2009). In
and recent
In adaptability years,
recentInyears,
years, novel
this(Antonelli trends
novelhuman-robot
trends arise
2016;
et al., arise
arise to support
toKrüger
support et towards more detailed description of
manual
al., manufacture.
In recent field,
novel trends collaboration
to support and
towards
al.,
towardsBattaïa
2012;more
more (2017).
detailedRecent
Bortolini
detailed etdescription
al.,
description 2017), of ergonomic
literature
of extends the
ergonomic
integration
ergonomic risk
risk
of
risk (Xu
(Xu et
problem
related
(Xu et
et
manual
al.,
manual manufacture.
2009). In recent
manufacture. In this
this field,
Inyears, novel
field, human-robot
trends
human-robot arisecollaboration
to support
collaboration al.,
is an
manual emerging
manufacture. technology.
In this field, Since collaborative
human-robot collaboration al., 2012;
robots towards
al., 2012;
planning
2012;moreBortolini
detailed
Bortolini
tasks (Battini
Bortolini et
et etal.,al.,2017),
al.,
etdescription
al., 2017),
2017,
2017), integration
of ergonomic
integration
2016a),
integration or of
risk
of related
(Xu of
related
design
of related et
is
is anan emerging
manual manufacture.
emerging technology.
In thisbyfield,
technology. Since collaborative
human-robot
Since robots
collaboration
collaborative robots al., planning
2012;tasks (Battini
(cobots)
is an are characterized
emerging technology. inherent
Since security mechanisms,
collaborative robots planning
composite
planning tasks (Battiniet et
Bortolini
capacity-oriented
tasks (Battini etal.,al.,
et al.,
(e.g.
al., 2017,
2017),
2017,
2017, 2016a),
integration
2016a), cycle
minimizing
2016a), or
or of
or design
related
design
time
design of
of
or
of
(cobots)
is an
(cobots) are
emerging
are
the installation
(cobots) characterized
technology.
characterized
of external
are characterized by
by inherent
Since
inherent security
devices security
by inherent mechanisms,
collaborative
security mechanisms,
(for instance,
mechanisms, robots
safety composite
planning capacity-oriented
tasks (Battini et (e.g.
al., minimizing
2017, 2016a), cycle
or time
design or
of
composite
number
composite of capacity-oriented
stations) and
capacity-oriented (e.g.
ergonomic
(e.g. minimizing
objectives
minimizing cycle
(Battini
cycle time
timeet or
al.,
or
the
the installation
(cobots) isare
installation of
of external
notcharacterized external devices
byestablishing
inherent (for
devices security
(for instance,
mechanisms,
instance, safety
safety composite
number of stations) and ergonomic objectives
fences)
the installation necessary.
of externalBy devices stations
(for comprising
instance, safety of capacity-oriented
numberBautista
2016b;
number of stations)
et al.,
stations) and
and ergonomic
2016).
ergonomic objectives (Battini
(e.g. minimizing
objectives cycle time
(Battini
(Battini et
et al.,
et or
al.,
al.,
fences)
the
fences)
human is
is
fences) worker not
installation
not necessary.
of
necessary. external
and cobot,
is not necessary. By
By establishing
devices
establishing
Byadvantages stations
(for
establishingofstations both manualcomprising
instance,
stations safety
comprising
comprising and 2016b;number Bautista
of et
stations) al.,
and2016).
ergonomic objectives (Battini et al.,
2016b; Bautista et al.,
2016b; Bautista et al., 2016). 2016).
human
fences) worker
is not and
necessary. cobot, By advantages
establishing of both manual and
human
automated
human worker
worker and
and cobot,
manufacture cobot, canadvantages
be realized
advantages ofstations
of both
both comprising
in amanual combined
manual and
and The 2016b;capacity-oriented
The Bautista et al., 2016).
capacity-oriented
objectives pursued in previous
objectives pursued in previous
automated
human
automated worker
production system
automated manufacture
and
manufacture
manufacture cobot,
at low can can
can be
advantages
be realized
realized
costs.beAsrealized of in
both
in a
a combined
manual
combined
reported inby aStatistisches and The capacity-oriented
literature,
The however,
capacity-oriented
combined literature, however, are mere are objectives
mere substitutes
objectives pursued
pursued for the in monetary
in previous
previous
production
automated
production system
manufacture
system at
at low
low costs.
can
costs. be As
As reported
realized
reported inby
by Statistisches
a combined
Statistisches The capacity-oriented
literature, however, are mere substitutes
objectives
substitutespursued for
for the
in
the monetary
previous
monetary
production system at low costs. As reported by Statistisches value associated with them and become less applicable, the
Bundesamt (2018), cobots face significantly increasing value associated
literature, however,with them
are mere and become
substitutes less
for applicable,
the monetary
the
Bundesamt
production
Bundesamt (2018),
system
(2018),at low cobots
cobots costs. face
face significantly
As reported
significantly increasing
by Statistisches
increasing value literature,
more however,
value associated
associated
generalized withthe
with are
them
them mere andsubstitutes
problem
and become
becomes
become for applicable,
less
less the monetary
applicable,
(Boysen the
and
the
market
Bundesamt demand. (2018), cobots face significantly increasing more generalized the problem becomes (Boysen and
market
Bundesamt demand.
market demand.
demand. (2018), cobots face significantly increasing value
more
Fliedner,
more associated
generalized with
2008). We the
generalized thethem
therefore and
problem
problem become
propose becomes less
an approach
becomes applicable,
(Boysen
(Boysen for cost-the
and
and
market Fliedner, 2008). We We the therefore propose an approach
approach for cost-
cost-
Due todemand.
market its efficiency, industrial manufacture is yet more Fliedner,
oriented
Fliedner, generalized
2008).
assembly
2008). Weline thereforeproblem
balancing
therefore propose
propose becomes
with an
human
an approach(Boysen for
workers
for and
and
cost-
Due
Due to to
to its
predominantly
Due its efficiency,
its efficiency,
efficiency, industrial
industrial
organized inindustrial manufacture
manufacture
Assemblymanufacture
Lines utilizing is
is the
is yet
yet Fliedner,
yet oriented assembly
2008). We line balancing
therefore proposewith human
an approach workers
for and
cost-
oriented
cobots
oriented assembly
under line
consideration
assembly line balancing
of
balancing with
ergonomics.
with human
human workers
workers and
and
predominantly
Due
flow to
predominantly its
principle. organized
efficiency,
organized
Assembly in
in
Line Assembly
industrial
Assembly
Balancing
predominantly organized in Assembly Lines utilizing the oriented Lines
manufacture
Lines
(ALB) utilizing
utilizing
is is
concerned the
yet
the cobots under consideration
assembly line of
balancingergonomics.
with human workers and
cobots under consideration
cobots under consideration of ergonomics.of ergonomics.
flow
flow principle.
predominantly
with
flow principle.
the allocation
principle. Assembly
organized
Assembly
Assembly of tasks Line
in to
Line
Line Balancing
stations, (ALB)
Assembly
Balancing
Balancing Lines
(ALB) is
is concerned
utilizing
for instance
(ALB) is in the
concerned
concerned the To cobotstheunder
best consideration
of authors’ knowledge,of ergonomics. only two articles in
with principle.
flow
with the allocation
the allocation
Assembly of tasks
of tasks
Line toBalancing
to stations, (ALB)
stations, for instance
for instance
is in the
concerned
in the To To the
To the best
best of of authors’
authors’ knowledge,
knowledge, only two
only two articles
articles in in
initialthedesign
with allocation of ofthetasks production
to stations,system. for instance in the academic
Recognized the best
academic
literature
literature
contribute
of authors’
contribute
to cost-oriented
knowledge,
to
only
cost-oriented
two ALB articles
ALB
under
under
in
initial
initialthedesign
with
summaries
initial allocation
design
designon the of ofthe
of state
of the ofproduction
thetasks to stations,
production
research
production in system.
for field
system.
this
system. Recognized
instancehaveinbeen
Recognized
Recognized the To the best
academic
consideration
academic of authors’
literature contribute
of ergonomics.
literature contribute knowledge,
Kara
to et al.only
to cost-oriented
cost-oriented
(2014) twoconsider
articles
ALB
ALB underin
the
under
summaries
initial
summaries designon
on the
theof state
the
state of
of research
production
research in
in this
system.
this field
field have
Recognized
have been
been consideration
academic
consideration
assignment of of
of ergonomics.
literature
workers contribute
ergonomics.
to Kara
stations, to
Kara et
et
where al.
al. (2014)
cost-oriented
(2014)
workers consider
ALB
consider
maintain the
under
the
an
provided byon
summaries Boysen
the state et al. of (2007),
researchBoysen al. (2008),
in thisetfield have been and consideration of ergonomics. Kara et al. (2014) consider the
provided
summaries
providedand
Battaïa
provided by
byon
by Boysen et
et al.
the (2013).
Boysen
Dolgui
Boysen state
et of (2007),
al.
al. researchBoysen
(2007),
(2007), in thiset
Boysen
Boysen al. (2008),
al.
etfield
et al. have been
(2008),
(2008), and assignment
and
and assignment of
consideration
individual
assignment of workers
of
workers
maximum
of workers to
toofstations,
ergonomics.
to Karawhere
stations,
acceptable
stations, where
where workers
workers
workers maintain
(2014) expenditure.
et al.energy consider
maintainthe
maintain an
an
an
Battaïa
provided
Battaïa and
and by Dolgui
Boysen
Dolgui (2013).
et
(2013). al. (2007), Boysen et al. (2008), and individual
assignment
individual
Besides
individual maximum
of
resource workers
maximum
maximum and to of
of
station
of acceptable
stations,
acceptable
costs,
acceptable where
they energy
workers
energy
consider
energy expenditure.
maintain
expenditure.
additional
expenditure. an
Battaïa and Dolgui (2013). Besides resource
When humans
Battaïa
When and Dolgui are(2013).
engaged in the assembly process, the individual Besides
cost
Besides types. Tiacciand
maximum
resource
resource andand
and station
of Mimmi
station
station costs,
acceptable they
costs,(2018)
costs, they consider
theyenergy
consider
suggest
consider additional
expenditure.
additional
an ALB
additional
When humans
When humansofare
consideration
humans are
are engaged
engaged in
ergonomics
engaged is the
in
in the
the assembly
assembly
required
assemblyin theprocess,
design the
process,
process, the cost
of approach
the Besides
cost types. Tiacci
resource
types. Tiacci and and
andstationMimmi
Mimmi costs, (2018)
they
(2018) suggest
consider
suggest an
an ALB
additional
ALB
consideration
When humans of ergonomics is required in the design of cost for
types. multiple
Tiacci andvariants.
Mimmi They utilize
(2018) the
suggest Occupational
an ALB
consideration
modern
consideration ofare
workplaces
of engaged
ergonomics
to optimize
ergonomics in
is the
is both assembly
required
human
required in theprocess,
in well-being
the designand
design the
of approach
of cost
approach for
types.
for multiple
Tiacci
multiple andvariants.
Mimmi
variants. They
They utilize
(2018)
utilize the
suggest
the Occupational
an
Occupational ALB
modern workplaces to RepetitiveforActions
approach multiple (OCRA)
variants. They indexutilize
for the ergonomic
Occupational risk
consideration
modern system
overall
modern workplaces
workplaces to optimize
ofperformance
ergonomics
to optimize
optimize is both
both
(IEA,
both human
required
human
2018).
human well-being
inThe designand
the neglect
well-being
well-being of
and RepetitiveforActions
of assessment
and approach
Repetitive Actions
multiple (OCRA)
variants.
(OCRA) indexutilize
They
index for the
for ergonomic
Occupational
ergonomic risk
risk
overall
modern
overall system
workplaces
system performance
to
performanceoptimize (IEA,
both
(IEA, 2018).
human
2018). The neglect
well-being
The neglect of
and
of Repetitive and pursue
Actions minimizing
(OCRA) index overall
for costs considering
ergonomic risk
physical system
overall ergonomicsperformance (as one (IEA,domain 2018). of Theergonomics,
neglect of assessmentassessment
Repetitive
assessmentcost and pursue
Actions
and pursue(OCRA) minimizing
minimizing index overall
forhave
overall costs considering
ergonomic
costs considering risk
physical
overall
physical
henceforth
physical
ergonomics
system
ergonomicsperformance
synonymously)
ergonomics
(as
(as
(as one
one
in(IEA,
one the
domain
2018).
domain
domaindesignof
of
The
of
ergonomics,
phaseneglect
ergonomics,
of an
ergonomics, of multiple
multiple
assessment
multiple
and
cost
and
cost
types.
pursue
types.
pursue
types.
Both
Both
Both
approaches
minimizing
approaches
minimizing
approaches
overall
overall have
have
costs
costs
inconsidering
in
in
common,
common,
considering
common,
henceforth synonymously) in however, cost
multiple that only
types.one Both resource
approachescan behave assigned to each
in common,
physical
assembly
henceforth ergonomics
henceforthsystem synonymously) (as one
may consequently
synonymously) in the
in the
thedomain
causedesign
design of phase
phase
unforeseen
design phase of
of an
ergonomics,
costs
of an
in however,
an multiple that
that only
cost
however,Consequently,types.one
only oneBoth resource
approaches
resource can be
behave
canparallel assigned to
to each
in common,
assigned each
assembly
henceforth system may
synonymously) consequentlyin thecausecause unforeseen
design costs
phase costs of an in station.
however, that only one they neglect
resource can be execution
assigned to eachby
assembly system
assembly system may consequently
may consequently cause unforeseen
unforeseen costs in however,
in station. Consequently,
that only
station. Consequently,
Consequently, theyone they neglect
resource
they neglect can parallel
be
neglect parallel execution
assigned
executioneach
parallel execution to by
by
assembly system may consequently cause unforeseen costs in station. Consequently, they neglect parallel execution by
station.
by
2405-8963 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2019 IFAC 1890
Peer review©under
Copyright 2019 responsibility
IFAC of International Federation of Automatic
1890 Control.
Copyright
Copyright ©
© 2019
2019 IFAC
IFAC 1890
1890
10.1016/j.ifacol.2019.11.473
Copyright © 2019 IFAC 1890
 2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Christian Weckenborg et al. / IFAC PapersOnLine 52-13 (2019) 1860–1865 1861
multiple resources and the resulting necessity of scheduling
tasks within stations. Furthermore, resources are assumed to
have the identic performance, i.e., resource-dependent
processing times are not considered. The approaches
therefore neglect the time-cost tradeoff between different
resources we exhibit in our study.
The remainder of this contribution is structured as follows: in
Section 2, we give a detailed description of the decision-
making situation we consider. Our mathematical model is
presented in Section 3. Results of an illustrative example and
a multivariate analysis are provided in Section 4. Finally, we
conclude our study in Section 5.
2. PROBLEM SETTING
We face the problem of designing an assembly line for a
particular product allocated from Network Planning. A
maximum cycle time 𝑐𝑐𝑐𝑐 is assigned in order to meet sales
targets. The expected revenue per product unit is determined
by sales planning. Since production volume and product
revenue are externally given, we assume the revenues of our
mutually exclusive investment alternatives (i.e. assembly line
configurations) to be identical. Therefore, we utilize the Cost
Comparison Method (CCM) for static investment appraisal of
one average period. In the contribution at hand, we suggest
one cycle time as a representative time span. We pursue
minimizing costs per cycle and consider salaries,
depreciation, and interest costs, which we aggregate to costs
of stations 𝐶𝐶𝑘𝑘𝑆𝑆 and costs of utilized resources 𝐶𝐶𝑟𝑟𝑅𝑅 (in EUR)
(Götze et al., 2015).
A set of 𝑛𝑛 tasks 𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼 have to be allocated among the
stations of the system with respect to their precedence
relations (𝑖𝑖, 𝑗𝑗) ∈ 𝐸𝐸. A set 𝑟𝑟 ∈ 𝑅𝑅 containing 𝑅𝑅 resource types
are available. Resources are assumed to be either a human
worker (𝑟𝑟 ∈ 𝑅𝑅𝑊𝑊 ⊆ 𝑅𝑅) or a cobot (𝑟𝑟 ∈ 𝑅𝑅𝐶𝐶 ⊆ 𝑅𝑅). Processing
times 𝑡𝑡𝑖𝑖𝑖𝑖 (in min) are deterministic and constant for each task
𝑖𝑖 and resource 𝑟𝑟. Whether resource 𝑟𝑟 is able to perform task 𝑖𝑖
is indicated in the capability matrix 𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 (𝑛𝑛 ⋅ 𝑅𝑅 binaries).
Binary variables 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 and 𝑦𝑦𝑘𝑘𝑘𝑘 indicate whether task 𝑖𝑖 is
assigned to station 𝑘𝑘 and resource 𝑟𝑟, and resource 𝑟𝑟 is
assigned to station 𝑘𝑘, respectively. The maximum dimension
of the line is constrained to 𝐾𝐾 stations due to spatial
restrictions.
Since human workers can be assigned to the stations of the
system, ergonomic risks have to be considered. We therefore
adapt the basic concept proposed by Price (1990). He
determines the mean work rate (𝑀𝑀𝑀𝑀𝑀𝑀) of workers based on
energy expenditure (a physical, non-financial measurand) of
the conducted tasks. Under consideration of an acceptable
work level (𝐴𝐴𝐴𝐴𝐴𝐴) to be maintained throughout a work day
and a relaxation rate (𝑅𝑅𝑅𝑅), he determines relaxation
allowances (𝑅𝑅𝑅𝑅) required for the worker to recover from load
exceeding 𝐴𝐴𝐴𝐴𝐴𝐴, thus adding a time increment to the
respective task. Since it simplifies ergonomics assessment
and reduces computational complexity of the resulting
problems, his approach is particularly attractive for
consideration in ALB and has recently received increasing
attention (Battini et al., 2015; Battini et al., 2016a, 2017;
Battini et al., 2016b; Finco et al., 2018; Sgarbossa et al.,
1891
2016). Additionally, energy expenditure can be predicted
particularly well (Garg et al., 1978). Following Price (1990)
and Battini et al. (2017), workers may maintain 𝐴𝐴𝐴𝐴𝐴𝐴 =
4.3 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘/𝑚𝑚𝑚𝑚𝑚𝑚 throughout their work day, while 𝑅𝑅𝑅𝑅 =
1.86 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘/𝑚𝑚𝑚𝑚𝑚𝑚 are required while relaxing in standing
𝑒𝑒
position. Since 𝑀𝑀𝑀𝑀𝑀𝑀 = 𝑖𝑖𝑖𝑖 , where 𝑒𝑒𝑖𝑖𝑖𝑖 (in kcal) is the energy
𝑡𝑡𝑖𝑖𝑖𝑖
expenditure of human worker 𝑟𝑟 conducting task 𝑖𝑖, 𝑅𝑅𝑅𝑅
calculates to
𝑀𝑀𝑀𝑀𝑀𝑀−𝐴𝐴𝐴𝐴𝐴𝐴 1 𝑒𝑒𝑖𝑖𝑖𝑖
𝑅𝑅𝑅𝑅 = = 0.41 ⋅ − 1.76. (1)
𝐴𝐴𝐴𝐴𝐴𝐴−𝑅𝑅𝑅𝑅 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 ⁄𝑚𝑚𝑚𝑚𝑚𝑚 𝑡𝑡𝑖𝑖𝑖𝑖
To allow for relaxation of workers, the corrected processing
times 𝜂𝜂𝑖𝑖𝑖𝑖 are calculated according
𝜂𝜂𝑖𝑖𝑖𝑖 = 𝑡𝑡𝑖𝑖𝑖𝑖 (1 + 𝑅𝑅𝑅𝑅)
1 𝑒𝑒
0.41 ⋅ 𝑒𝑒𝑖𝑖𝑖𝑖 − 0.76 ⋅ 𝑡𝑡𝑖𝑖𝑖𝑖 , 𝑖𝑖𝑖𝑖 > 4.3 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘/𝑚𝑚𝑚𝑚𝑚𝑚
={ 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 ⁄𝑚𝑚𝑚𝑚𝑚𝑚 𝑡𝑡𝑖𝑖𝑖𝑖
(2)
𝑡𝑡𝑖𝑖𝑖𝑖 , 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒.
Since energy expenditure is utilized as a measure of human
ergonomic load, (1) and (2) are limited to ∀𝑖𝑖 ∈ 𝐼𝐼, 𝑟𝑟 ∈ 𝑅𝑅𝑊𝑊 and
thus do not apply for cobots. Please note that 𝜂𝜂𝑖𝑖𝑖𝑖 > 𝑡𝑡𝑖𝑖𝑖𝑖 for
any task with 𝑀𝑀𝑀𝑀𝑀𝑀 > 𝐴𝐴𝐴𝐴𝐴𝐴. Henceforth, we utilize the
corrected processing times 𝜂𝜂𝑖𝑖𝑖𝑖 to limit the ergonomic load of
workers, which (independent of actual processing times 𝑡𝑡𝑖𝑖𝑖𝑖 )
must not exceed 𝑐𝑐𝑐𝑐.
Additional complexity is induced by the presence of cobots.
Since multiple resources can be assigned to each station,
parallel execution of tasks within stations is allowed.
Therefore, the necessity of task scheduling within stations
arises. We introduce variables 𝑠𝑠𝑖𝑖 to denote the start time of
task 𝑖𝑖 in the station it is assigned to. Regarding precedence
relations of tasks (𝑖𝑖, 𝑗𝑗) ∈ 𝐸𝐸, we not only have to ensure their
compliance between stations (as is common in ALB) but also
within stations, i.e. 𝑗𝑗 can start earliest, after 𝑖𝑖 is finished in
their common station. This is not only true for tasks subject
to precedence relations, but also for any tasks 𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼 which
are assigned to both same station and resource. For these
tasks, one task consequently has to precede the other on their
common resource.
3. MODEL FORMULATION
Based on the description of the problem setting and the
notation as summarized in Table 1, we set up the following
model.
Minimize
𝑍𝑍(𝑥𝑥, 𝑦𝑦) = 𝑐𝑐𝑐𝑐 ⋅ ∑𝑘𝑘∈𝐾𝐾 ∑𝑟𝑟∈𝑅𝑅 𝐶𝐶𝑘𝑘𝑆𝑆 ⋅ 𝑥𝑥𝑛𝑛𝑛𝑛𝑛𝑛 ⋅ 𝑘𝑘
+𝑐𝑐𝑐𝑐 ⋅ ∑𝑘𝑘∈𝐾𝐾 ∑𝑟𝑟∈𝑅𝑅 𝐶𝐶𝑟𝑟𝑅𝑅 ⋅ 𝑦𝑦𝑘𝑘𝑘𝑘
(3)
∑𝑘𝑘∈𝐾𝐾 ∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 = 1
∀𝑖𝑖 ∈ 𝐼𝐼 (4)
𝑠𝑠𝑖𝑖 + ∑𝑘𝑘∈𝐾𝐾 ∑𝑟𝑟∈𝑅𝑅 𝑡𝑡𝑖𝑖𝑖𝑖 ⋅ 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ≤ 𝑐𝑐𝑐𝑐 ∀𝑖𝑖 ∈ 𝐼𝐼
(5)
∑𝑖𝑖∈𝐼𝐼 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ⋅ 𝜂𝜂𝑖𝑖𝑖𝑖 ≤ 𝑐𝑐𝑐𝑐 ∀𝑘𝑘 ∈ 𝐾𝐾, 𝑟𝑟 ∈ 𝑅𝑅𝑊𝑊 (6)
∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑗𝑗𝑗𝑗𝑗𝑗 ≤ ∑𝑘𝑘ℎ=1 ∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑖𝑖ℎ𝑟𝑟 ∀(𝑖𝑖, 𝑗𝑗) ∈ 𝐸𝐸, 𝑘𝑘 ∈ 𝐾𝐾 (7)
𝑠𝑠𝑖𝑖 + ∑𝑟𝑟∈𝑅𝑅 𝑡𝑡𝑖𝑖𝑖𝑖 ⋅ 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ≤ 𝑠𝑠𝑗𝑗 + 𝑀𝑀(1 − 𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 ) ∀𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼, 𝑘𝑘 ∈ 𝐾𝐾
(8)
𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 1 − 𝑀𝑀(1 − ∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ) − 𝑀𝑀(1 − ∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑗𝑗𝑗𝑗𝑗𝑗 )
∀(𝑖𝑖, 𝑗𝑗) ∈ 𝐸𝐸, 𝑘𝑘 ∈ 𝐾𝐾
(9)
𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 + 𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 1 − 𝑀𝑀(1 − 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ) − 𝑀𝑀(1 − 𝑥𝑥𝑗𝑗𝑗𝑗𝑗𝑗 )
∀𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼, 𝑖𝑖 ≠ 𝑗𝑗, 𝑘𝑘 ∈ 𝐾𝐾, 𝑟𝑟 ∈ 𝑅𝑅
(10)
2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Christian Weckenborg et al. / IFAC PapersOnLine 52-13 (2019) 1860–1865
1862

1
𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 ≤ ⋅ (∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 + ∑𝑟𝑟∈𝑅𝑅 𝑥𝑥𝑗𝑗𝑗𝑗𝑗𝑗 ) ∀𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼, 𝑘𝑘 ∈ 𝐾𝐾 stations, however, neither has to precede the other. By
bounding auxiliary variables 𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 in constraints (9)–(11), we
2
(11)

𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ≤ 𝑦𝑦𝑘𝑘𝑘𝑘 ⋅ 𝐶𝐶𝑀𝑀𝑖𝑖𝑖𝑖 ∀𝑖𝑖 ∈ 𝐼𝐼, 𝑘𝑘 ∈ 𝐾𝐾, 𝑟𝑟 ∈ 𝑅𝑅 (12) are able to limit the activation of computationally complex

𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 ∈ {0,1}
scheduling constraints (8) to the relevant cases. Constraints
∀𝑖𝑖 ∈ 𝐼𝐼, 𝑘𝑘 ∈ 𝐾𝐾, 𝑟𝑟 ∈ 𝑅𝑅 (13) (12) ensure that task 𝑖𝑖 is only assigned to resource 𝑟𝑟 in station
𝑘𝑘, if resource 𝑟𝑟 is allocated to station 𝑘𝑘 and it is capable of
𝑦𝑦𝑘𝑘𝑘𝑘 ∈ {0,1} ∀𝑘𝑘 ∈ 𝐾𝐾, 𝑟𝑟 ∈ 𝑅𝑅
performing task 𝑖𝑖. Finally, constraints (13)–(16) enforce the
(14)
𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 ∈ {0,1} ∀𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼, 𝑘𝑘 ∈ 𝐾𝐾 (15) domains of the decision variables. The resulting model can be
classified as mixed-integer linear program.
𝑠𝑠𝑖𝑖 ≥ 0 ∀ 𝑖𝑖 ∈ 𝐼𝐼 (16)
4. NUMERICAL EXAMPLES
Table 1. Introduced sets, parameters, and variables. To illustrate the effectiveness of our approach and derive
Sets and parameters quantitative evidence on the described problem, we present
𝑖𝑖, 𝑗𝑗 ∈ 𝐼𝐼 ≡ {1, … , 𝑛𝑛} Set of assembly tasks.
̅}
computational experiments in the following section.
𝑘𝑘, ℎ ∈ 𝐾𝐾 ≡ {1, … , 𝐾𝐾
Definition
Set of stations.
𝑟𝑟 ∈ 𝑅𝑅 ≡ {1, … , 𝑅𝑅̅}
Therefore, we implemented the model in Java 8 and solved it
Set of resource types, where using the Java CPLEX API (version 12.7.1). The
𝑅𝑅 = 𝑅𝑅𝑊𝑊 ∪ 𝑅𝑅𝐶𝐶 , with 𝑅𝑅𝑊𝑊 set of computations were run on a standard computer with Intel
worker resource types, and Core i7-8550U CPU @ 1.80 GHz and 16 GB RAM. All
𝑅𝑅𝐶𝐶 set of cobot resource types. instances discussed in the following could be solved to
𝐸𝐸
Set of direct proven optimality in average computational time of 5.4
precedence relations (𝑖𝑖, 𝑗𝑗). seconds with a maximum of 15.0 seconds.
𝑐𝑐𝑐𝑐 Cycle time (in min).
𝐶𝐶𝑘𝑘𝑆𝑆
Costs per station and minute (in 4.1 Instance generation
EUR), if 𝑘𝑘 stations are utilized.
of resource 𝑟𝑟 per entity For our experiments, the real-life case study provided by
𝐶𝐶𝑟𝑟𝑅𝑅
Costs
and minute (in EUR). Battini et al. (2016b) serves as a basis scenario we extend to
Capability matrix (1, if 𝑟𝑟 is suit our problem. In their study, the authors describe the
𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖
capable of performing 𝑖𝑖). assembly process of a high-pressure cleaner for home and
Processing time of task 𝑖𝑖 on garden applications which yet is conducted utilizing manual
𝑡𝑡𝑖𝑖𝑖𝑖
resource 𝑟𝑟 (in min). stations. We adopt the product-related data and the maximum
Energy expenditure of task 𝑖𝑖 on cycle time of three minutes as provided in their article
𝑒𝑒𝑖𝑖𝑖𝑖
resource 𝑟𝑟 (in kcal). (Fig. 1). Also, we limit the maximum dimension of the line to
four stations.
𝜂𝜂𝑖𝑖𝑖𝑖
Corrected processing time of
task 𝑖𝑖 on resource 𝑟𝑟 (in min). Using constraints (2) and the information provided in the case
𝑀𝑀 Big M parameter, e.g. 𝑀𝑀 = 𝑐𝑐𝑐𝑐. study, we calculate the corrected processing times. For the
Decision and auxiliary variables given example problem, only Task 9 requires additional
1, if task 𝑖𝑖 is assigned to station
𝑥𝑥𝑖𝑖𝑖𝑖𝑖𝑖 relaxation time when conducted by human workers (𝐻𝐻):
𝑘𝑘 and resource 𝑟𝑟. 1
1, if resource 𝑟𝑟 is assigned to 𝜂𝜂9𝐻𝐻 = max (𝑡𝑡9𝐻𝐻 , 0.41 ⋅ 𝑒𝑒9𝐻𝐻 − 0.76 ⋅ 𝑡𝑡9𝐻𝐻 )
𝑦𝑦𝑘𝑘𝑘𝑘 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 ⁄𝑚𝑚𝑚𝑚𝑚𝑚
station 𝑘𝑘. 1
= max (0.183 𝑚𝑚𝑚𝑚𝑚𝑚 , 0.41 ⋅ 0.96 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 − 0.76 ⋅
1, if task 𝑖𝑖 precedes task 𝑗𝑗 in 𝑘𝑘𝑘𝑘𝑘𝑘𝑘𝑘 ⁄𝑚𝑚𝑚𝑚𝑚𝑚
𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘
(17)
station 𝑘𝑘. 0.183 𝑚𝑚𝑚𝑚𝑚𝑚)
Start time of task 𝑖𝑖 in its = max(0.183 𝑚𝑚𝑚𝑚𝑚𝑚, 0.255 𝑚𝑚𝑚𝑚𝑚𝑚) = 0.255 𝑚𝑚𝑚𝑚𝑚𝑚
𝑠𝑠𝑖𝑖
respective station.
In order to simplify the reproducibility of our investigations,
Objective (3) is to minimize costs per cycle, taking into we restrict the number of resource types to two (human
consideration costs of stations and resources per minute, and worker and cobot) and consequently do not differentiate their
the cycle time. Constraints (4) ensure that each task is individual entities. Additionally, we assume both resource
assigned to one station and resource. Compliance with cycle types to be capable of performing each task. Also, we assume
time is ensured in constraints (5). Constraints (6) limit the costs of stations to scale proportionately to their number.
ergonomic load of workers in each station. Precedence However, our generic model can accommodate more
relations between stations are enforced by constraints (7). resource types, their potentially limited capabilities, and
Constraints (8) are the general scheduling constraints
indicating that task 𝑖𝑖 has to precede task 𝑗𝑗 in their common
economies of scale in material flow and station installation
without additional customization.
station. This holds true, if 𝑧𝑧𝑘𝑘𝑘𝑘𝑘𝑘 = 1. These cases are further
specified by constraints (9)–(11): 𝑖𝑖 has to precede 𝑗𝑗, if 𝑖𝑖 is a Since the study of Battini et al. (2016b) does not provide all
direct predecessor of 𝑗𝑗 and both are assigned to the same the required parameters, we have to justify additional
station. If two tasks 𝑖𝑖 and 𝑗𝑗 are assigned to the same station assumptions regarding costs of workers and stations, and
and resource, one has to precede the other independent of costs and processing speed of cobots. To determine
precedence relations. If these tasks are allocated to different depreciation of stations and cobots, we have to take

1892
 2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Christian Weckenborg et al. / IFAC PapersOnLine 52-13 (2019) 1860–1865 1863

assumptions on operation in the model company. In this If cobots are available, three workers and two cobots are
study, we assume a total of 552,000 minutes to be considered allocated among four stations. Tasks conducted by cobot
in machine hour rate depreciation, which corresponds to a require three times the human processing time (for instance,
five-year depreciation period with 230 work days per year observe length of Task 1 in both configurations). In Station 2,
and one daily eight-hour shift. both worker and cobot are assigned and consequently execute
0.767 0.217
tasks in parallel. Please note that precedence relations within
2 3 stations are respected. This can be observed particularly well
2.69 0.9
for Tasks 3 and 10. Since Task 3 has to be finished by the
0.117 1.333 1.417 1
cobot prior Task 10 starts, the human experiences idle time
4 10 11 13
0.18 4.56 5.07 1.08 between Tasks 3 and 10 resulting in a delayed start of Task
0.4 0.417 0.25 0 0 0.75 0.417 0.267 10. In case cobots are available, costs per cycle result to
1 5 6 V1 V2 15 16 17
𝐸𝐸𝐸𝐸𝐸𝐸
𝑍𝑍𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 3 𝑚𝑚𝑚𝑚𝑚𝑚 ⋅ 0.06 ⋅ 4 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
0.95 1.19 0.88 0 0 0.92 0.5 0.63

𝑚𝑚𝑚𝑚𝑚𝑚⋅𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝐸𝐸𝐸𝐸𝐸𝐸
0.083 0.417 1.083
7 12 14 + 3 𝑚𝑚𝑚𝑚𝑚𝑚 ⋅ 0.6 ⋅ 3 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 (19)
1 1.17 𝑚𝑚𝑖𝑖𝑖𝑖⋅𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝐸𝐸𝐸𝐸𝐸𝐸
0.15
0.633 0.183 Legend:
+ 3 𝑚𝑚𝑚𝑚𝑚𝑚 ⋅ 0.08 ⋅ 2 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 6.6 𝐸𝐸𝐸𝐸𝐸𝐸,
8 9 tiH (Human processing time in min) 𝑚𝑚𝑚𝑚𝑚𝑚⋅𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
1.23
i
0.96 eiH (Human energy expenditure in kcal) indicating a 16.7% cost reduction compared to the manual
Fig. 1. Precedence graph of the example problem. case. In this example, deployment of cobots consequently
represents an economic option for (partial) automation.
Eurostat (2018) estimates hourly labor costs in Germany at
35 EUR. Therefore, we assume worker costs of 0.6 EUR/min. Manual assembly line
Data to estimate the investment in stations are generally
Station 1 Station 2 Station 3 Station 4
difficult to obtain and highly depend on the actual assembly
processes. We suppose costs of 35,000 EUR per station, ct ct ct ct
which comprise for equipment and installation of the station H 1 8 56 2 10 11 12 14 13 15 16
itself, the related material flow technology, and interest costs. 0 3 0 73 4 9 3 0 3 0 17 3
The resulting depreciation consequently amounts to 0.06 1.700 2.770 2.917 2.433
EUR/min per station. Human ergonomic load on the station
Optimal line balance for manual assembly line
Cobots usually have a basic price between 25,000 EUR and Collaborative assembly line
35,000 EUR (Robotiq, 2016). Taking into account additional
accessories, installation, and interests, we assume overall Station 1 Station 2 Station 3 Station 4
costs of 45,000 EUR per cobot entity, which results in ct 9 ct ct 17 ct
depreciation of 0.08 EUR/min. Consequently, cobots are H not assigned 2 8 10 11 12 14 13 15 16
rather low-priced, but they also lack of efficiency in the C 1 4 5 6 3 not assigned not assigned

human environment. The Bosch APAS cobots, as an 0 3 0 3 0 3 0 3

7
example, reduce working speed to 0.5 m/s in presence of --- 2.987 2.917 2.433
persons (Bosch, 2018). We therefore conservatively assume Human ergonomic load on the station

that the robot requires three times the human processing time Optimal line balance for collaborative assembly line
Relaxation time: Idle times:
and thus introduce a time-cost tradeoff between cobots’ low
costs and high processing times and workers’ high costs and Fig. 2. Configuration of example systems.
low processing times. The presence of cobots could also lead to a reduction of
ergonomic load of human workers. Please note that this
4.2 Illustrative example relation does not apply for the example illustrated. While
workers experience an average ergonomic load of 2.46 min in
Based on the assumptions outlined in Section 4.1, we present the manual scenario, the (remaining) workers in the
a numerical example with consideration of cobots and automated scenario witness an average ergonomic load of
without (both calculated utilizing our model). The resulting 2.78 min.
system configurations are illustrated in Fig. 2 for both cases.
In the manual case, one worker is allocated to each of the 4.3 Determining factors of cobot deployment
four stations. Since only one resource is available in each
Since the illustrative example suggests high potential, we
station, tasks are executed serially with respect to their
further investigate the determining factors of cobot
precedence relations. Task 9 requiring relaxation time of
deployment in manual lines. We therefore conduct a
0.072 min is assigned to Station 2 (relaxation time
multivariate analysis in a full factorial design. The parameter
emphasized in orange), increasing human ergonomic load in
variations considered are given in Table 2. Characteristics of
this station (and thus its finish time) to 2.77 min. The
the basis scenario described in the preceding sections are
resulting costs per cycle are calculated using objective
indicated by asterisks. Parameters not listed here remain as
function (3) and amount to
introduced.
𝐸𝐸𝐸𝐸𝐸𝐸
𝑍𝑍𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = 3 𝑚𝑚𝑚𝑚𝑚𝑚 ⋅ 0.06 ⋅ 4 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝑚𝑚𝑚𝑚𝑚𝑚⋅𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝐸𝐸𝐸𝐸𝐸𝐸
+ 3 𝑚𝑚𝑚𝑚𝑚𝑚 ⋅ 0.6 ⋅ 4 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 = 7.92 𝐸𝐸𝐸𝐸𝐸𝐸.
(18)
𝑚𝑚𝑚𝑚𝑚𝑚⋅𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤

1893
2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Christian Weckenborg et al. / IFAC PapersOnLine 52-13 (2019) 1860–1865
1864
Table 2. Parameters of multivariate analysis.
Parameter Characteristics
Station costs in EUR/min 0.06* 0.12 0.18
Workers costs in EUR/min 0.06 0.18 0.6*
Energy expenditure (factor) 1.00* 1.25 1.50
Speed of cobot (factor) 1.00 2.00 3.00*
In our analysis, we consider variation of four parameters. An
increase in station costs serves to investigate scenarios with
more sophisticated station and material flow equipment. The
reduction of worker costs to 0.18 EUR/min relates to labor
costs experienced in Eastern European countries, for instance,
Czech Republic, Estonia, and Poland (Eurostat, 2018).
Worker costs of 0.06 EUR/min are below labor costs of any
European country. Since cobots are assumed to induce costs
of 0.08 EUR/min, however, this scenario serves to investigate
environments with comparatively expensive cobots compared
to human labor. An increase in energy expenditure serves to
research ergonomically more demanding environments (value
serves as an increase factor applied on the original scenario).
Finally, we investigate scenarios with faster cobots. The
increase factors of the latter two parameters are applied on
the energy expenditure characteristics and human processing
times of the original case provided by Battini et al. (2016b).
Additionally, we subsequently allow the line to comprise a
maximum of six stations to provide for higher freedom of
system design.
Table 3. Incidence on cost-efficient configurations
subdivided by the resulting number of stations.
of stations
ce (of 81)
Occurren
Resource
Number
per cycle
Workers
Cobots
station
station
station
Costs
∅(𝜎𝜎)
∅(𝜎𝜎)
∅(𝜎𝜎)
∅(𝜎𝜎)
s per
per
per
3 15 2.20 (0.53) 1.33 (0.00) 0.53 (0.28) 0.80 (0.28)
4 38 2.99 (1.54) 1.07 (0.11) 0.59 (0.42) 0.48 (0.41)
5 12 3.34 (1.05) 1.03 (0.08) 0.80 (0.26) 0.23 (0.32)
6 16 5.98 (1.82) 1.06 (0.08) 0.26 (0.14) 0.80 (0.07)
Information on the resulting system configurations are
reported in Table 3, subdivided by the resulting number of
stations. The most cost-efficient configurations are realized
utilizing three stations. These solutions are characterized by
high degree of parallelization (1.33 resources per station on
average of all three-station systems), high degree of cobot use
(0.80 cobots per station), and comparatively few workers
(0.53 workers per station). These systems are realized for any
level of station costs and energy expenditure, however, only
if cobots are as fast as workers and workers induce low or
medium costs. Please note that cobots are also utilized in
three-station systems in cases, in which cobot costs are higher
than worker costs. Making advantage of parallel work,
assignment of a cobot to an (existing) manual station is
beneficial compared to introducing an additional manual
station.
Configurations with four stations are realized comprising a
variety of characteristics, for instance, (i) human workers
may exclusively be utilized throughout all levels of station
costs, if worker costs are low and cobot speed and ergonomic
requirements are low or medium, or (ii) cobots may
1894
exclusively be utilized throughout all levels of station costs
and energy expenditures, if worker costs are medium or high
and cobots are as fast as humans. Subject to high ergonomic
requirements, five-station systems with similar characteristics
result at higher costs.
Six-station systems are configured throughout all levels of
station costs and energy expenditures, if cobot speed is
medium or slow and worker costs are high. Though cobots
are slow in these cases, it is beneficial to introduce an
additional station with a cobot rather than using an expensive
worker at an existing station in parallel.
Finally, we report results on system configurations
subdivided by the level of induced ergonomic requirements
in Table 4. If the ergonomic environment gets more
challenging, i.e. higher energy expenditure is required for
manual task execution, fewer workers and more cobots are
used for assembly. Consequently, a trend towards higher
degree of automation can be observed for ergonomically
more demanding environments. Though cost-efficient
configurations comprise increasing utilization of cobots, the
overall costs per cycle increase with the ergonomic
requirements. In order to achieve cost-efficient assembly
system configurations, decisions should not only be taken
about (partial) automation. Supplementary decisions lowering
the energy expenditure experienced within manual task
conduction may thus also contribute to cost-efficiency in
manual production systems.
Table 4. Impact of ergonomic requirements
on resulting system configurations.
ce (of 81)
Occurren
Resource
per cycle
expend.
Workers
Energy
Cobots
station
station
station
Costs
∅(𝜎𝜎)
∅(𝜎𝜎)
∅(𝜎𝜎)
∅(𝜎𝜎)
s per
per
per
1.00 27 3.32 (1.77) 1.11 (0.14) 0.58 (0.39) 0.53 (0.39)
1.25 27 3.38 (1.84) 1.14 (0.15) 0.58 (0.36) 0.56 (0.36)
1.50 27 3.75 (2.12) 1.10 (0.14) 0.48 (0.36) 0.61 (0.41)
As illustrated in our analysis, cost-efficient assembly lines
may yield a vast variety of characteristics ranging from fully
automated to exclusively manual configurations (even for the
very same product). Consequently, a thorough analysis of the
considered production environment is necessary in order to
generate economic assembly system designs. Utilizing our
approach, we are able to support decision making processes
and contribute to cost-efficiency in manual manufacture.
5. OUTLOOK
Our field requires further research. During the development
of this contribution we initially additionally considered costs
of the energy consumption of collaborative robots. Since this
cost component did not have any explanatory effect, we
decided to neglect it in the contribution at hand. However, in
further research we strive to extend our planning approach
towards recognized management accounting methodologies.
In the current state, we model the processing times of
collaborative robots as constant. The processing speed of
collaborative robots, however, varies depending on the
presence or absence of humans. Our assumption of fully
capable cobots needs to be refined such that tasks may be
 2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Christian Weckenborg et al. / IFAC PapersOnLine 52-13 (2019) 1860–1865 1865

incompatible with execution by cobot. Additionally, actual Eurostat (2018), “Estimated hourly labour costs, 2017”,
collaboration of human and robot may reduce the energy available at: https://bit.ly/2zge5ky (accessed 22 February
expenditure of the human worker. We therefore pursue more 2019).
realistic modeling of the characteristics of collaborative Finco, S., Battini, D., Delorme, X., Persona, A. and
robots. Sgarbossa, F. (2018), “Heuristic methods to consider rest
allowance into assembly balancing problem”, IFAC-
REFERENCES
PapersOnLine, Vol. 51 No. 11, pp. 669–674.
Antonelli, D., Astanin, S. and Bruno, G. (2016),
Garg, A., Chaffin, D.B. and Herrin, G.D. (1978), “Prediction
“Applicability of human-robot collaboration to small
of metabolic rates for manual materials handling jobs”,
batch production”, in Afsarmanesh et al., H. (Ed.),
American Industrial Hygiene Association journal,
Collaboration in a hyperconnected world, IFIP Advances
Vol. 39 No. 8, pp. 661–674.
in Information and Communication Technology, Vol.
Götze, U., Northcott, D. and Schuster, P. (2015), Investment
480, Springer International Publishing, Cham, pp. 24–32.
Appraisal, Springer, Berlin, Heidelberg.
Battaïa, O. and Dolgui, A. (2013), “A taxonomy of line
IEA (2018), “Definition and Domains of Ergonomics”,
balancing problems and their solution approaches”,
available at: https://bit.ly/2OP8poa (accessed 22
International Journal of Production Economics, Vol. 142
February 2019).
No. 2, pp. 259–277.
Kara, Y., Atasagun, Y., Gökçen, H., Hezer, S. and Demirel,
Battini, D., Calzavara, M., Otto, A. and Sgarbossa, F.
N. (2014), “An integrated model to incorporate
(2016a), “The tntegrated assembly line balancing and
ergonomics and resource restrictions into assembly line
parts feeding problem with ergonomics considerations”,
balancing”, International Journal of Computer
IFAC-PapersOnLine, Vol. 49 No. 12, pp. 191–196.
Integrated Manufacturing, Vol. 27 No. 11, pp. 997–
Battini, D., Calzavara, M., Otto, A. and Sgarbossa, F. (2017),
1007.
“Preventing ergonomic risks with integrated planning on
Krüger, J., Lien, T.K. and Verl, A. (2009), “Cooperation of
assembly line balancing and parts feeding”, International
human and machines in assembly lines”, CIRP Annals -
Journal of Production Research, Vol. 55 No. 24, pp.
Manufacturing Technology, Vol. 58 No. 2, pp. 628–646.
7452–7472.
Otto, A. and Battaïa, O. (2017), “Reducing physical
Battini, D., Delorme, X., Dolgui, A., Persona, A. and
ergonomic risks at assembly lines by line balancing and
Sgarbossa, F. (2016b), “Ergonomics in assembly line
job rotation. A survey”, Computers & Industrial
balancing based on energy expenditure: a multi-objective
Engineering, Vol. 111, pp. 467–480.
model”, International Journal of Production Research,
Otto, A. and Scholl, A. (2011), “Incorporating ergonomic
Vol. 54 No. 3, pp. 824–845.
risks into assembly line balancing”, European Journal of
Battini, D., Delorme, X., Dolgui, A. and Sgarbossa, F.
Operational Research, Vol. 212 No. 2, pp. 277–286.
(2015), “Assembly line balancing with ergonomics
Price, A.D.F. (1990), “Calculating relaxation allowances for
paradigms: two alternative methods”, IFAC-
construction operatives — Part 1: Metabolic cost”,
PapersOnLine, Vol. 48 No. 3, pp. 586–591.
Applied Ergonomics, Vol. 21 No. 4, pp. 311–317.
Bautista, J., Batalla-García, C. and Alfaro-Pozo, R. (2016),
Robotiq (2016), “What is an average price for a collaborative
“Models for assembly line balancing by temporal, spatial
robot ?”, available at: https://bit.ly/2Q9KHDR (accessed
and ergonomic risk attributes”, European Journal of
22 February 2019).
Operational Research, Vol. 251 No. 3, pp. 814–829.
Sgarbossa, F., Battini, D., Persona, A. and Visentin, V.
BMWi (2017), “Bundesministerium für Wirtschaft und
(2016), “Including ergonomics aspects into mixed-model
Energie. Gesundheitswirtschaft. Fakten & Zahlen.
assembly line balancing problem”, in Goonetilleke, R.
Ausgabe 2016.”, available at: https://bit.ly/2OTQMUm
and Karwowski, W. (Eds.), Advances in Physical
(accessed 22 February 2019).
Ergonomics and Human Factors, Advances in Intelligent
Bortolini, M., Faccio, M., Gamberi, M. and Pilati, F. (2017),
Systems and Computing, Vol. 489, Springer International
“Multi-objective assembly line balancing considering
Publishing, Cham, pp. 991–1001.
component picking and ergonomic risk”, Computers &
Statistisches Bundesamt (2018), “Produktionsmenge von
Industrial Engineering, Vol. 112, pp. 348–367.
Mehrzweck-Industrierobotern in Deutschland in den
Bosch (2018), “Human-robot collaboration”, available at:
Jahren 2008 bis 2017 (in 1.000 Stück)”, available at:
https://bit.ly/2Ju4BH0 (accessed 22 February 2019).
https://bit.ly/2Xibm5m (accessed 22 February 2019).
Boysen, N. and Fliedner, M. (2008), “A versatile algorithm
Tiacci, L. and Mimmi, M. (2018), “Integrating ergonomic
for assembly line balancing”, European Journal of
risks evaluation through OCRA index and
Operational Research, Vol. 184 No. 1, pp. 39–56.
balancing/sequencing decisions for mixed model
Boysen, N., Fliedner, M. and Scholl, A. (2007), “A
stochastic asynchronous assembly lines”, Omega,
classification of assembly line balancing problems”,
Vol. 78, pp. 112–138.
European Journal of Operational Research, Vol. 183
Xu, Z., Ko, J., Cochran, D.J. and Jung, M.-C. (2012),
No. 2, pp. 674–693.
“Design of assembly lines with the concurrent
Boysen, N., Fliedner, M. and Scholl, A. (2008), “Assembly
consideration of productivity and upper extremity
line balancing: which model to use when?”, International
musculoskeletal disorders using linear models”,
Journal of Production Economics, Vol. 111 No. 2, pp.
Computers & Industrial Engineering, Vol. 62 No. 2, pp.
509–528.
431–441.

1895