Journal of Manufacturing Systems 46 (2018) 103–114

Contents lists available at ScienceDirect

Journal of Manufacturing Systems

journal homepage: www.elsevier.com/locate/jmansys

Staff scheduling in job rotation environments considering ergonomic

aspects and preservation of qualifications
Jan Hochdörffer ∗ , Marc Hedler, Gisela Lanza
wbk Institute of Production Science, Karlsruhe Institute of Technology (KIT), Kaiserstr. 12, 76131 Karlsruhe, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Demographic change is a well-known influencing factor challenging social security systems in industrial-
Received 9 January 2017 ized countries. In a manufacturing context, companies need to cope with an increasingly heterogeneous
Received in revised form 2 November 2017 workforce in terms of qualification and impairments, as well as an increasing average age. The devel-
Accepted 21 November 2017
opment of more standardized processes and the trend to move towards shorter lead times, paired with
Available online 6 December 2017
demographic changes in the workforce, reveal a strong importance of staff planning.
A short-term staff planning system which generates job rotation schedules taking into consideration
Keywords:
workers’ qualifications, the workplace’s ergonomic exposure, and the most recent allocations of each
Job rotation
Staff scheduling
worker is sought to ensure the right worker is allocated to the right work place at the right time. The
Assembly arising complexity of such scheduling problems is met in this paper by using a linear programming based
Ergonomics heuristic, which solves the scheduling problem gradually for each rotation round and generates a holistic
Linear programming job rotation schedule for an entire workday.
Heuristic The presented approach to short-term staff scheduling is implemented in a VBA-based software pro-
totype and was tested in the final assembly line of a German automotive manufacturer.
© 2017 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction cause of illness, followed by other illnesses such as cardiovascular

Along with Italy and Japan, Germany is one of the countries with
the oldest population [1] and according to the most likely future
scenario, the German population will not only get older, but also
decrease significantly over the next decades [2]. Therefore, Ger-
man companies need to prepare for this demographic change by
considering the challenges that an increasingly elderly workforce
presents to their workforce planning.
The commonly believed hypothesis that the general perfor-
mance of a workforce decreases with increasing age, is disproved
in recent studies which find no significant correlation between
age and performance [3]. However, research in this field assumes
that workforce characteristics are becoming more heterogeneous
within an ageing workforce, e. g. in terms of impairments [4]. Addi-
tionally, the occurrence of illness-related absence from work is
found to increase with an increasing age of the workforce. Mus-
culoskeletal disorders (MSD) are identified as the most significant
Abbreviations: MSD, musculoskeletal disorders; BEQR, Balanced Ergonomic
Qualification preserving job Rotation; EAWS, European Assembly Worksheet; IEA,
International Ergonomics Association; WBE, whole body exposure.
∗ Corresponding author.
E-mail address: jan.hochdoerffer@kit.edu (J. Hochdörffer).
diseases and psychological illnesses [5].
One of the essential business sectors within the German econ-
omy is the automotive industry [6]. With respect to age distribution,
it is predicted that the average age among the workforce in this
industry will increase from 40 (2011) to 44 within the next five
years. By 2021, more than half of the workforce will be older than
45, whereas this group represented only 38% in 2011 [7]. Manufac-
turing processes within the automotive industry are characterized
by repetitive and physically demanding work tasks at high fre-
quencies, especially in final assembly lines. When executing final
assembly tasks, the hand-finger system as well as the whole human
body are primarily exposed. Such a working environment may lead
to more frequent occurrences of MSD in the long run [8].
Increasing product variety as well as shorter product lifecycles of
produced car series call for new requirements for the manufactur-
ing process. The workforce does not only have to adapt to changes
faster than before, but workers also have to be able to execute a
wider range of working activities which requires diversified qual-
ifications [9]. These aspects amplify the challenges demographic
change poses on workforce planning and qualification procedures
[7].
To lower or prevent occupational injuries and absenteeism from
work, three general approaches have been identified by [10]. Firstly

https://doi.org/10.1016/j.jmsy.2017.11.005
0278-6125/© 2017 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
104 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
(1), working activities can be redesigned in terms of tools or work-
place modifications and the manufacturing process at the early
stages of product development. In addition, (2) the use of per-
sonal protective equipment, such as specific clothing, can help with
injury prevention. Moreover, (3) utilizing administrative means,
e. g. job rotation, may lower or prevent occupational injuries and
absenteeism from work. According to Lussier and Hendon, job rota-
tion can be defined as “[. . .] performing different jobs in a certain
sequence, each for a set period of time” [11].
When analyzing the three outlined approaches, it was found
that the application of engineering solutions (1) are often too
expensive to be fulfilled, and when considering MSD, not many
effective solutions are available for protective equipment (2). Utiliz-
ing administrative means (3) offers the most cost-efficient solution
for the challenges in this field [12].
Through the use of administrative means, working activities can
be organized in a way that reduces the monotony of the assembly
process. Particularly in the automotive industry, job rotation is a
broadly used concept to reduce fatigue and unilateral exposure of
body regions [13]. However, there is typically no specific focus on
the changing needs of elderly workers. In the last decade, com-
panies such as BMW have begun addressing the challenges in the
automotive assembly process stemming from demographic change
by testing measures to cope with the future situation of elderly
workers [14].
Nevertheless, a holistic approach which considers physical
exposure and continuous preservation of the workforce’s qualifi-
cations, as well as incorporates the situation of individual elderly
workers into models to generate job rotation schedules is currently
still missing [15]. In this paper an integer linear programming (ILP)
based heuristic is developed, which eases the outlined aspects and
fills the gap in current research entitled the Balanced Ergonomic
Qualification preserving job Rotation (BEQR) method. The remain-
der of this paper is structured as follows: In Section 2, the relevant
theoretical foundations are outlined followed by an overview of
current literature on short-term staff planning. The implementa-
tion of the BEQR method, including the developed heuristic to solve
the ILP, is introduced in Section 3. Finally, an application example in
the final assembly line of a German automotive manufacturer illus-
trating the planning advantage of the BEQR method is presented in
Section 4, followed by the discussion of the results in Section 5.
2. Literature review
According to Ellström (1997) the term qualification in a pro-
fessional environment refers to the competence that is actually
required by the work task and/or is implicitly or explicitly pre-
scribed by the employer [16]. A job rotation environment with
varying work tasks leads to multi-skilled employees who can han-
dle large product variety [17]. In contrast, related qualifications
become obsolete when the job rotation interval for a specific
workstation is too long. Therefore the workforce has to practice
respective working activities on a regular basis, which claims a
trade-off between the benefits of long and short-term job rotation
intervals [18].
Staff scheduling aims to align a company’s available workforce
with work tasks in the most beneficial way and can also be referred
to as personnel scheduling or rostering. In the context of manu-
facturing, the mapping of the available workforce with different
work tasks may be referred to as allocation and is typically based
on workforce qualifications and the task requirements with respect
to time and capacity constraints [19,20].
The definition of the term “ergonomics”, provided by the Inter-
national Ergonomics Association (IEA), strives for improvement
in two main fields: the optimization of human well-being and
the optimization of the overall system performance [21]. Further-
more, ergonomics can broadly be clustered into three main focus
groups. The field of physical ergonomics (1) focuses on anatomi-
cal or biomechanical human characteristics. Relevant topics within
this field are working posture and the prevention of work-related
MSD. Organizational ergonomics (2) deals with the optimization of
sociotechnical systems, including decisions regarding work design.
Lastly, cognitive ergonomics (3) considers the interaction between
humans and their working environment, such as human-computer
interaction.
Looking at the design of work systems, ergonomics can also be
distinguished based on the point in time of implementation. Schlick
et al. introduce the terms “prospective ergonomics” and “corrective
ergonomics” [22]. Accordingly, prospective ergonomics means
that ergonomic findings are included ex ante while designing work
systems. Corrective ergonomics, on the other hand, is used when
ergonomic aspects are regarded ex post, i. e. after working sys-
tems are created. Short-term staff scheduling systems, which take
ergonomic aspects into account, can be categorized as organiza-
tional corrective ergonomics.
Various ergonomic risk assessment metrics have been devel-
oped in recent years featuring differences in their focused working
activities and levels of detail. These metrics typically aim at evaluat-
ing physical workloads by means of screening tools. The outcomes
of the evaluations can then be used to improve the assessed work-
ing activity ergonomics [22]. Based on the screening procedure
of the Automotive Assembly Worksheet (AAWS), the European
Assembly Worksheet (EAWS) was designed. The EAWS assesses
movements a worker executes while performing work tasks at shop
floor level, and thereby revises the AAWS for an increased field of
application [23]. According to Schaub et al., the EAWS is currently
implemented at several German and European car manufacturers
and automotive suppliers [22].
In particular, the EAWS evaluates the following:
• working postures and movements with low additional physical
efforts
• action forces of the whole body or hand-finger system
• manual materials handling
• repetitive loads on the upper limbs
The evaluation of each topic results in point values indicating
the extent of the physical exposure the worker has to cope with.
In order to identify potential risks regarding MSD when executing
tasks, these generated point values are summed and can then be
associated to a color scheme which is illustrated in Fig. 1 [22].
The green section represents physical loads ranging from 0 to
25, indicating a minor risk for the person performing the task.
The yellow section, which ranges from 26 to 50, marks a possi-
ble risk whereby an entire structural redesign of the respective
task is recommended in order to minimize the workforce’s physical
exposure. However, should a redesign not be feasible, the imple-
mentation of other corrective ergonomic actions such as minor
physical modifications (e. g. adjustments to tool equipment) or
organizational changes (e. g. adaptions to short-term staff schedul-
ing) is recommended. The red section contains all tasks featuring
a high likelihood of getting MSD, which should therefore be modi-
fied. However, the risks to which the workforce is exposed are not
linked to “critical values” at which certainly and instantly an injury
occurs, since the measures are based on the average worker. Indi-
vidual physical conditions and working techniques also have an
influence on such risk [22].
In the following, current approaches in the field of short-term
staff planning dealing in particular with job rotation and staff
scheduling problems are reviewed. The approaches are preselected
for their relevance regarding the considered topic. Therefore, the
 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
Fig. 1. Ergonomic risk assessment color scheme adapted from [22].
focus is on approaches which take at least one of the following
requirements into account: ergonomics and preservation of work-
force qualifications.
In order to better classify the reviewed sources, the require-
ments were broken down into three main aspects of interest,
namely (1) incorporation of ergonomics and how well the reviewed
sources could adapt to characteristics of elderly workers, (2) bal-
anced distribution among the worker allocation to workstations,
as well as the possibility to include individual restrictions such as
qualifications and impairments into rotation schedule generation,
and (3) qualification preservation, assessing whether the gener-
ated schedules consider the timespan between two allocations to
the same workstation. While very short timespans can be hard on
physical exposure and motivation, too long timespans can affect
the worker’s knowledge of the respective work tasks negatively. In
addition to the considered scope, the solving methods used are also
briefly analyzed.
Carnahan et al. introduce the ergonomic job rotation scheduling
problem that aims to smooth the ergonomic risks amongst work-
ers by minimizing their ergonomic load when performing lifting
tasks [24]. Tharmmaphornphilas and Norman extend the model
from Carnahan et al. by taking the quality of preceding worker allo-
cations into account when allocating workers in the next period to
enable a stronger inter-period relationship [12]. Other approaches
focus on the reduction of different exposure types. For instance,
Kullpattaranirun and Nanthavanij present models for the design
of rotation schedules to reduce noise exposure [25,26], while both
noise exposure and back injuries are considered in the model from
Aryanezhad et al. [27].
The majority of the approaches reviewed base their rotation
schedules on either specific or general ergonomic risk assessment
metrics, but differ in respect to the level of detail regarding evalu-
ation sections they cover. For instance, Tharmmaphornphilas and
Norman as well as Carnahan et al. use the Job Severity Index (JSI)
to assess the potential for back injury [12,24] while Yoon et al. use
rough screening procedures [28]. Other authors, such as Asensio-
Cuesta et al. develop and apply their own, case-specific metrics [29].
Furthermore, applied metrics may feature different possibilities to
incorporate physical impairments, which are more likely to arise
due to the ageing workforce. Such metrics are used in [15,29,30].
The succession of stations which bear a risk of exposing the work-
force to critically high physical loads is only explicitly modelled
by Yoon et al. [28]. Considering individual physical capabilities of
the workforce, as in [29], is seen as a major limitation for appli-
cation in industry in view of anti-discrimination laws in several
industrialized countries.
Looking at the second aspect, to create a balanced rotation
schedule and to cope with individual restrictions, the reviewed lit-
erature is vastly diverse. The objective to avoid the succession of the
same workstation in a row is only picked up in the approaches of,
Yoon et al., Asensio-Cuesta et al. as well as Bhadury and Radovilsky
[28,29,31]. However, none of these approaches implemented the
ability to modify a sequence of allocated workstations. Addition-
105
ally, the possibility to adjust the duration of rotation rounds is only
considered by two sources [29,32].
Regarding the third aspect reviewed, only two of the selected
publications meet the requirements for a qualification-preserving
rotation schedule [15,18]. However, the approach elaborated by
Azizi et al. respects individual intellectual characteristics, which
might conflict with current anti-discrimination laws in indus-
trial countries [18]. The model from Stanić lacks application
since the pursued action regarding the preservation of qualifi-
cations is only outlined theoretically [15]. Additionally, looking
at the assumptions made regarding workforce qualifications, the
regarded publications are split into two groups. Most of the sources
expect the workforce to be equally (homogeneously) qualified e. g.
[28,29,31,32] and only differ in the extent of how well an individ-
ual can execute a task (e. g. [33]). Only two approaches distinguish
between being and not-being equally qualified and therefore
incorporate a heterogeneous workforce [15,27]. The possibility
to schedule qualification procedures, e. g. on-the-job training, by
allocating more than one worker to selected workstations is not
considered by any of the reviewed approaches.
A wide range of methods is applied to solve the job rota-
tion or staff scheduling problems presented in these sources.
The majority of the approaches either use integer programming
[12,24,27,28,30,18], metaheuristics such as genetic algorithms
[24–26,29,30,18,33], or simulated annealing [30]. Heuristics are
only applied in a few approaches [12,30,32].
Overall, all reviewed approaches feature limitations. None of the
approaches fully fulfills the requirements to create a job rotation
schedule that reduces physical exposure upon a heterogeneously
qualified workforce, considers feasible variations among the work-
stations, pays attention to the preservation of qualifications, or is
suitable for in-field use with individual adjustments or restrictions.
The approach from Stanić is the only approach that fulfills most
of the requirements, however it still lacks a solving method, and
therefore remains purely conceptual [15].
3. Short-term staff planning modelling
In order to set up a short-term staff planning system, vari-
ous data on the workforce and on the assembly tasks assigned to
workstations need to be gathered. In the BEQR method, several
assumptions about the application scope of the considered man-
ufacturing system, workforce characteristics, and assembly task
characteristics are made to cope with real-life environments in the
automotive industry.
3.1. Modelling assumptions
The first group of assumptions relate to organizational con-
ditions typically found in automotive assembly lines. Rotation
schedules are often created individually for each shift whereby
differences in terms of the length and number of rotation rounds
within shifts may exist. The number of workstations per workgroup
106 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114

to be executed per shift are fixed during the duration of the shift,
however, some workstations may be nonobligatory.
The second group of assumptions is made to better model the
characteristics of the regarded workforce. Following current regu-
lations and anti-discrimination laws (e. g. AGG in Germany [34]),
no differentiation is made amongst workers in terms of gender
or capabilities, such as the efficiency and quality in performing a
task. However, the workforce is assumed to be heterogeneous in
respect to individual qualification profiles which means that not
all workers are qualified to execute all tasks [29]. The numbers of
workers present per shift may vary based on expected or unex-
pected situations. Imbalances in the number can be dealt with to
a certain extent: understaffing may be compensated by eliminat-
ing nonobligatory tasks, and overcrowding can be compensated by
executing qualification procedures with two workers working on
the same tasks at the same time. The total recovery of a worker’s
physical exposure is assumed to occur overnight. Therefore, the
restriction to limit the succession of high exposure within the same
body region is neglected for the allocation to the first workstation
during the shift on the subsequent day.
The third assumption group concerns specific work task charac-
teristics within the automotive assembly process. All workstations
are assumed to be classified in terms of their physical expo-
sure using a common ergonomic risk assessment metric. Derived
ergonomic classification values increase linearly with the time it
takes to execute the task. Interdependencies, such as the accu-
mulation of fatigue or recovery aspects due to the sequence of
workstations allocated in the generated schedule, are expected to
balance out with respect to their negative or positive effects and
are neglected accordingly. Therefore, assessment metric values for
a specific task depend only on the task’s ergonomic risk assessment
and duration.
3.2. Input information
Required subsets of input data for the BEQR method may stem
from different sources such as the company’s enterprise resource
planning system or other databases. Furthermore, data is generated
in the course of the method’s usage, such as information regarding
the most recent rotation schedules. Lastly, the BEQR method fea-
tures several options for the user to influence the generation of the
rotation schedules, by taking individual preferences and restric-
tions into account. The following tables depict the relevant data

using examples  a workgroup with a total number of seven work-
of
ers I = 1, . . ., 7 assigned to it and a regular assembly process
 
including five obligatory J o = 1, 3, 4, 5, 6 and one nonobligatory
 
workstation J no = 2 with J = J o ∩ J no .
Table 1 illustrates the workers qualification profiles in terms of
their assignment to workstations.
The used notation in Table 1 specifies whether a worker is qual-
ified (1) or not qualified (0) to perform tasks at a workstation.
 
qij ∈ 0, 1 for i = 1, . . ., nI ; j = 1, . . ., nJ (1)
As illustrated in Table 2, the number of workers assigned to a
workgroup is normally higher than the number of workstations to
be executed since workers may be absent from work (noted by 0)
due to illness, scheduled holidays, or other reasons.
The ergonomic risk assessment metric used is comparable to
the European Assembly Worksheet (EAWS), which is explained in
detail in [22]. In response to working environment characteristics
within automotive assembly lines and the increasing awareness of
MSD and its causes, particular focus is set on the whole body subdi-
vided into four categories as well as the hand-finger system [7]. An
exemplary overview of ergonomic risk assessment by workstations
is illustrated in Table 3 using the color scheme outlined in Section
2.
As illustrated in Table 4, the general length and number of rota-
tion rounds may differ among workgroups and shift types (i. e.
early shift, late shift, and night shift). Due to these differences and
their associated consequences, all relevant information needs to be
gathered before applying the BEQR method.
In order to fulfill the proposed objective to develop balanced
ergonomic qualification preserving job rotation schedules, the cre-
ation of an internal database is necessary. As illustrated in Table 5,
every allocation for every worker within the workgroup is stored
after each rotation round. This information is complemented with
the individual frequency of workstations and the respective length
of each round. In case of a worker’s absence, the corresponding
fields in the database are left blank (e. g. Worker 07, who was absent
on the 23.02.2016). The basis for a new allocation of a worker is
set by considering the previous n rotation rounds. When choosing
a longer history of past allocations, effects such as illness-related
absences or holidays can be coped with. Datasets exceeding the

Table 1
Individual qualification profile matrix.

Worker 01 Worker 02 Worker 03 Worker 04 Worker 05 Worker 06 Worker 07

Station 01 1 1 1 1 1 1 1
Station 02 1 0 1 0 1 1 0
Station 03 1 1 0 1 1 1 0
Station 04 1 1 0 1 0 1 1
Station 05 1 0 1 1 1 1 1
Station 06 1 0 1 1 0 0 1

Table 2
Individual attendance matrix.

Name February March

Mo Tu We Th Fr Sa So Mo Tu We Th ... ... ...

22 23 24 25 26 27 28 29 1 2 3 ... ... ...

Worker 01 0 1 1 1 0 1 0 1 1 1 0 ... ... ...

Worker 02 1 0 1 0 1 1 0 0 1 0 1 ... ... ...
Worker 03 1 1 1 1 1 1 0 1 1 1 1 ... ... ...
Worker 04 0 1 0 1 0 0 0 1 0 0 0 ... ... ...
Worker 05 1 1 0 0 1 1 0 1 0 1 1 ... ... ...
Worker 06 1 1 1 1 1 0 0 1 1 1 1 ... ... ...
Worker 07 1 0 1 0 1 1 0 0 1 1 1 ... ... ...
 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114 107

Table 3
Ergonomic classification and specific characteristics of workstations.

Table 4 Table 6
Lengths of rotation rounds in regards to shift types. Temporary impairments for individual workers.

Early shift Start End Length (min) Worker Station CW Dates with temp. restrictions

Rotation 1 06:00 07:50 110 Mo Tu We Th Fr Sa

Rotation 2 08:15 10:00 105
Rotation 3 10:10 11:50 100 Worker 02 Station 04 CW 08 22 23
Rotation 4 12:25 13:20 55 Worker 07 Station 06 CW 08 22 23 24
Rotation 5 13:32 14:35 63

Late shift Start End Length (min)

In the BEQR method, impairments are distinguished in two
Rotation 1 14:35 16:30 115 categories: permanent and temporary. Permanent impairments
Rotation 2 16:45 18:00 75
Rotation 3 18:13 19:00 47
impede workstation execution in the long run and lead to an adjust-
Rotation 4 19:30 21:00 90 ment of a worker’s qualification profile. However, situations may
Rotation 5 21:10 22:00 50 arise, which prevent a worker from executing a workstation task
Rotation 6 22:10 22:55 45 only temporarily, such as sports injuries. The BEQR method fea-
tures the functionality to constrain individual qualification profiles
manually for the expected time of the impairment. In the example
illustrated in Table 6, worker 2 s qualification for workstation 4 will
specified timeframe n are deleted from the database as new ones be disregarded in calendar week (CW) 8 for Monday, the 22nd and
are added. Tuesday, the 23rd.
A distinctive feature of the BEQR method is the possibility to
manually restrict the sequences of allocated workstations within
the process of generating job rotation schedules. This function is 3.3. Formulation of the allocation model
useful when two workstations expose the same body region to a
high degree ergonomic risk. The user is also given the option to The proposed allocation model is based on the integer linear pro-
prioritize specified constraints within the range from one (high- gramming (ILP) allocation model described by Domschke et al. [35].
est priority) to five (lowest priority). In anticipation to Section 3.3, In order to determine job rotation schedules for multiple rotation
every possible worker allocation is given a penalty cost. The aim rounds, the respective model is extended by the dimension of time
of the presented linear program is to search for the best possi- in terms of rotation rounds k which are taking place at dedicated
ble solution, minimizing the overall penalty costs originating from shifts on dedicated days. Furthermore, the model is extended by
the workforce allocation. In this respect, sequences, which should considering workers’ qualifications qij,k and all functions which are
be avoided due to manually included restrictions, are not fully necessary to cope with the input information presented in Section
excluded, although they are not likely to occur in the incorporated 3.2.
procedures. Theoretically, situations may arise in which one con-
nK nJ
straint is violated. By assigning ranks to the constraints, the user   nI

Min cij,k xij,k (2)

can prioritize which violation is more acceptable to violate than
others. k=1 j=1 i=1

Table 5
Historical backtrack of past allocations.

Rotation round Date Length (min) Worker 01 ... Worker 04 ... Worker 07

Allocation Frequ. ... Allocation Frequ. ... Allocation Frequ.

1 23.02.16 63 Station 01 49 ... Station 05 16 ...

2 23.02.16 55 Station 02 63 ... Station 04 101 ...
3 23.02.16 100 Station 06 100 ... Station 03 68 ...
... ... ... ... ... ... ... ... ... ... ...
n␶ − 2 18.01.16 63 Station 05 117 ... Station 02 62 ... Station 06 120
n␶ − 1 18.01.16 55 Station 01 49 ... Station 03 68 ... Station 05 98
n␶ 18.01.16 100 Station 06 100 ... Station 02 62 ... Station 01 52
108 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
s.t.
nJ,k
 Regarding the complexity of short-term staff scheduling, Ernst
xij,k = 1 for i = 1, . . ., nI,k ; nI,k ⊆ nI ; nJ,k ⊆ nJ ; k = 1, . . ., nK (3)
j=1
nI,k
 [19]. Staff scheduling problems considering the allocation of mul-
xij,k = 1 for j = 1, . . ., nJ,k ; nJ,k ⊆ nJ ; nI,k ⊆ nI ; k = 1, . . ., nK (4)
i=1
xij,k ≤ qij,k for i = 1, . . ., nI,k ; nI,k ⊆ nI ; j = 1, . . ., nJ,k ; br nJ,k ⊆ nJ ; k = 1, . . ., nK
 
xij,k ∈ 0, 1 for i = 1, . . ., nI,k ; nI,k ⊆ nI ; j = 1, . . ., nJ,k ; br nJ,k ⊆ nJ ; k = 1, . . ., nK
 
qij,k ∈ 0, 1 for i = 1, . . ., nI,k ; nI,k ⊆ nI ; j = 1, . . ., nJ,k ; br nJ,k ⊆ nJ ; k = 1, . . ., nK
nI,k = nJ,k nI,k ⊆ nI ; nJ,k ⊆ nJ ; k = 1, . . ., nK
The objective of the presented model is to minimize the sum
of individual penalty costs (cij,k ) originating from the allocation of
present workers i= (1, . . ., nI ) to work tasks which are performed at
workstations j = 1, . . ., nJ within all rotation rounds during a shift
k = (1, . . ., nK ) in the regarded workgroup (Eq. (2)). The calculation
of penalty cost is outlined in Section 3.4. The actual allocation of
a single worker to a single workstation within a specific rotation
round is defined by the binary decision variable xij,k .
Penalty cost minimization is subject to constraints given in Eq.
(3)–(8). The discrete worker-workstation allocation in terms of one
workstation per worker per rotation round (3) and respectively
the workstation-worker relationship in terms of one worker per
workstation per rotation round (4) is defined in the first two con-
straints. However, as described in Section 3.1, this constraint could
be opposed to situations of understaffing and overcrowding, which
have to be intercepted prior to solving the ILP. As illustrated by
the pseudocode in Fig. 2, understaffing in a rotation round ku is
compensated by eliminating nonobligatory workstations jno* with
lowest priority from the set of workstations J until the number of
workstations matches the number of workers nJ,ku = nI,ku :
Overcrowding can be compensated for by allocating an unquali-
fied worker inq with qinq j ,ko = 0 to a workstation j during a rotation
round ko that is occupied by a qualified worker at the same time. In
such situations, qualification procedures for the unqualified worker
inq at the considered workstation j can be performed, which is
illustrated by the pseudocode in Fig. 3.
The allocation (xij,k ) of qualified workers (qij,k ) to a workstation
in a rotation round is guaranteed by Eq. (5), and respects temporary
impairments when specified in the input data. The condition of
non-negativity of the decision variable is stated in Eq. (6). In Eq. (7) it
is defined whether or not workers are qualified to process a task at a
workstation. An exact match of the number of present workers and
the workstations to be executed at every rotation round is ensured
by Eq. (8).
Fig. 2. Compensation of understaffing.
3.4. Heuristic approach to generate rotation schedules
et al. point out the complexity in finding an optimal solution to the
problem, when aspects such as minimal costs or equal distribu-
tion of tasks among employees have to be taken into consideration
tiple employees to multiple workplaces over a multi-period time
horizon are three-dimensional and can be categorized as NP-hard
(5) [36]. Since NP-hard problems may not be solved optimally in poly-
nomial time, heuristics are often chosen as a compromise between
computation time needed to solve the problem and the quality of
the solution [35].
In anticipation of implementation of the BEQR method within
(6)
a software tool, the following heuristic was developed to solve the
proposed ILP. This was necessary to allow for reasonable compu-
tation times on medium performance personal computers. When
solving the original ILP described in Section 3.3, the computation
time exceeded 45 min which was the maximum time available to
(7) create a shift schedule. As illustrated in Fig. 4, the short-term staff
scheduling problem is broken down into rotation round-specific
ILPs featuring only two dimensions, which can be solved optimally
in significantly lower computation time. The problem is therefore
(8)
solved interactively for each rotation round, and the results are then
strung back together in order to generate a feasible solution for one
shift.
Three matrices, identical in size from the nature nJ,k × nI,k ,
are generated for every rotation round k, with to-be executed
workstations (nJ,k ) in rows and present workers (nI,k ) in columns,
respectively. Qualifications of workers for to-be executed work-
stations qij,k are written in the first matrix as the result of step 3.
Penalty costs (cij,k ) for all possible allocations (step 4) are written
in the second matrix. Eq. (9) illustrates multiple parameters taken
into consideration when calculating penalty costs for each worker,
workstation and rotation round combination.
 
  Lij,

5
cij,k = HisRaij, ∗ HisFrij,n + HisRaij, ∗ ∗ ErgPj,cr + ManRij,k (9)
L
cr=1
Parameter HisRaij, refers to recent allocations  of the regarded
worker to a respective workstation.  is linked to the internal
database of the BEQR method and features the running number
within the range of the last n allocation, whereas index  = 1 rep-
resents the most recent allocation. Most recent allocations feature
higher HisRaij, values. Furthermore, the number of allocations to
a workstation (HisFrij,n ), as well as the length of the most recent
allocation (Lij, ), are considered. Hereby the ratio of the length of the
most recent allocation (Lij, ) to the average length of a daily shift
(␾L) is calculated. Larger numbers of allocations as well as longer
allocations lead to higher penalty cost and are therefore avoided
due to the structure of the penalty cost minimizing ILP. Thereby,
the preservation of workforce’s qualifications is supported.
The ergonomic exposure at a workstation j is considered by the
parameter ErgPj,cr which is calculated based on the task’s ergonomic
risk assessment values in four categories for the whole body and
one category for the hand-finger system that is referred to by
parameter cr.
The last parameter ManKij,k represents the manually-set restric-
tions for the sequences of allocated workstations. The value for
ManKij,k may significantly differ in magnitude, as the sum of all
affected restrictions is determined. Generally, the magnitude of the
manually set restrictions is limited by a sufficiently big number
(BIG M) which is calculated based on the application case-specific,
maximum possible penalty cost, coming from the first two brack-
eted terms in Eq. (9). Depending on the selected priorities fractions
of respective BIG M can be used.
 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114 109

Fig. 3. Compensation of overcrowding.

Fig. 4. Heuristic implemented within the BEQR method.

After having generated both (qualification & penalty cost) input
matrices, the ILP for rotation round k is set up (step 5) and solved
using the Simplex algorithm. The solution is used to generate dis-
tinct allocations (xij,k ) for present workers to workstations (step
6). Based on the solution, the internal database is updated (step 7)
and the algorithm is pursued with the next iteration, as long as the
termination condition is not yet fulfilled.
4. Results
The BEQR method was implemented in a software prototype
®
using Microsoft Excel , Visual Basic for Applications, and an Open-
Solver for testing in a real-life environment. In order to simplify
the application of the BEQR method, the process of generating the
job rotation schedules was integrated into a graphical user inter-
110 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114

face (GUI). The developed software prototype was tested for six
months, generating staff schedules for a workgroup at a German
automotive manufacturer’s final assembly line in Germany. The
considered workgroup alternated between early and late shifts
and therefore the number of rotation rounds varied between four,
five, and six rounds with differing lengths of rotation rounds. The
considered workgroup consisted of 40 permanent workers with
individual qualification profiles and 24 workstations from which
seven featured the same work tasks and thereby the same EAWS
risk points. Since identical tasks were to be performed in a side-
inverted position at two workstations, individual settings regarding
the allocation sequence were defined so that workers would not be
allocated to these two workstations back-to-back. During the test-
ing period, several temporary impairments for individual workers
were considered, such as a one week exclusion from certain work-
stations due to a sports injury. This was implemented in the GUI in
a way that qualifications (qij,k ) for the respective workstations were
removed for certain dates from the workers’ qualification profiles
and automatically restored in the following dates. Apart from tem-
porary impairments, qualification profiles of all workers were kept
fixed during the testing period of 250 rotation rounds to allow for
a better comparability of results.
As illustrated in Fig. 5, the number of workstations that work-
ers are qualified for (henceforth referred to as “qualifications per
worker”) ranged from 6 to 24, averaging out at 16.2 qualifications
per worker. Based on these qualifications per worker, the theoreti-
cal average ergonomic exposure per worker is depicted in Fig. 6.
The average ergonomic exposure is calculated by averaging the
ergonomic classification in terms of their EAWS risk points of all
qualifications per worker, ranging from 19.6 to 24.6 per worker.
In cases where a worker is qualified for all workstations, its
average ergonomic exposure is equal to the average ergonomic
classification of all workstations, which corresponds to 22.9 in
the considered workgroup. When a worker is not qualified for
all workstations, its theoretical ergonomic exposure may differ
according to its particular qualification profile. In the considered
workgroup, workstations with a higher ergonomic classification
feature a slightly lower number of qualified workers so that the
average ergonomic exposure over all workers is a little lower at
22.3 compared to the average ergonomic classification of all work-
stations. Overall, the average value of EAWS risk points for whole
body exposure (WBE) are significantly higher than those for hand-
finger exposure for the considered workgroup, and therefore the
focus is on WBE in the following.
As defined in chapter 4.2, the number of previous rotation
rounds n , which are considered in the generation of new rotation
rounds, is referred to as length of historical backtracks. In the fol-
lowing scenarios of n = 50 (scenario 1), n = 150 (scenario 2), and
n = 250 (scenario 3) previous rotation rounds, which are consid-
ered in the generation of new rotation rounds, are used to analyze

Fig. 5. No. of qualifications per worker.

Fig. 6. Workers’ theoretical average ergonomic exposure based on EAWS risk points.
 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
Fig. 7. Workers’ WBEs based on EAWS risk points.
Table 7
Averaged deviations of workers’ theoretical average WBE from actual WBE.
Scenario 1 Scenario 2
n 50 150
averaged deviation 0,71 0,16
Table 8
Rate of not allocated workstation numbers to number of qualifications per worker.
Scenario 1 Scenario 2
n 50 150
rate of not allocated workstations 21,1% 1,3%
the staff scheduling results of the BEQR method. One worker was
excluded from the scenario results since he was absent for a large
number of the first 100 rotation rounds. The workers’ WBEs based
on EAWS risk points for the three scenarios are plotted in Fig. 7 in
comparison to each worker’s theoretical average ergonomic expo-
sure.
It can be derived from the diagram that a larger n results
in smaller differences of a workers’ theoretical average WBE and
actual WBE. This can be quantified by looking at the averaged devi-
ations of workers’ theoretical average WBE and actual WBE over all
workers which is depicted in Table 7. This leads to the conclusion
that considering longer historical backtracks when generating staff
schedules with the BEQR method, results in a more fair distribution
of workers to workstations regarding their WBE, according to their
qualification profiles.
Additionally, it can be concluded that the larger n , the lower
the number of not allocated different workstations when consid-
ering workers’ qualifications. This is illustrated in Fig. 8, where the
number of different workstations that workers are allocated to over
the time period is compared to the qualifications per workers. The
seven workstations featuring the same work tasks are summarized
in one workstation since the allocation of workers to these work-
stations was not differentiated. A quantification of these results
is depicted in Table 8 showing the rate of not allocated worksta-
111
Table 9
Sums of adapted standard deviation differences between allocation quantities to
each qualified workstation per worker.
Scenario 3
Scenario 1 Scenario 2 Scenario 3
250
0,08 n 50 150 250
sum of adapted standard deviations 1,89 1,37 1,31
Scenario 3
tion numbers to number of qualifications per worker for a varying
number of rotation rounds.
250
It can be derived from these results that by increasing n , the
0,0%
preservation of qualification can be significantly improved since
workers are allocated to all workstations which they are qualified
for. However, this conclusion is biased since with an increasing n
there is also a larger number of time slots at disposal to allocate
workers to qualified workstations.
Therefore, it is necessary to also look at the balanced distribu-
tion of worker allocation to workstations in terms of the differences
between allocation quantities to each qualified workstation per
worker. In Fig. 9, the standard deviation of this difference divided
by the total quantity of allocations per worker in the scenarios of
50, 150, and 250 rotation rounds is plotted, showing a significant
reduction from which a more balanced distribution of worker allo-
cation can be derived. It can be derived that workstations were
blocked over short time periods due to certain constellations of
workers’ attendance and qualifications, in which the remaining
workers have to work around the blocked stations. This trend can
also be quantified by looking at the sums of these adapted standard
deviations as illustrated in Table 9.
The generation of a job rotation schedule was typically executed
one day prior to the day of interest, in order to display the next
day’s rotation schedule to the workforce in advance. The basis for
the generation was the input data outlined in Section 3. Hence,
the prototype software was used on a daily basis. The time for
solving the mathematical model was found to be highly depen-
dent on individually set constraints. However, despite the variance
in set constraints, the computation time for solving a single rota-
112 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114

Fig. 8. No. of allocated different workstations per worker.

Fig. 9. Adapted standard deviation of allocation quantities to qualified workstation per worker.

tion round never exceeded 30 s on a medium performance personal
computer.
5. Discussion
Overall, significant improvements in generating short-term staff
schedules can be identified by applying the BEQR method in a real-
life environment. These improvements were shown in the areas
of balanced distribution, qualification preservation, and fair allo-
cation of workers to workstations when incorporating ergonomic
aspects, by integrating a history of past allocation rounds, into the
staff scheduling problem.
In addition to the quantitative application results, the quality
of the generated job rotation schedules using the BEQR method
 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
was also evaluated using a survey to capture the software pro-
totype users’ opinions. These qualitative survey results show that
BEQR method users were highly satisfied due to the greatly reduced
time needed to generate a shift’s rotation schedule in comparison
to manual schedule generation. Furthermore, the surveyed users
ranked and generated rotation schedules as fair (second best out
of a six point ranking scale) regarding the defined objectives and
pointed out their strong desire to continue working with the soft-
ware prototype even after the testing period was terminated.
However, it also became clear, that the quality of the BEQR
method results, striving for a fulfilment of the outlined objectives,
fully depends on the input data, i. e. the qualification profiles and
the risk assessment of the workstations’ ergonomic exposure of
workers. Therefore, the BEQR method should be accompanied by
a mid- to long-term staff scheduling approach to adapt qualifi-
cation profiles which may add significant value to the quality of
the generated job rotation schedules. The development, as well
as the functional incorporation, of this feature into the developed
software prototype field accompanied with field testing should be
addressed in future research.
The BEQR method addresses those ergonomic aspects which
were identified as most important, using common ergonomic risk
assessment metrics. However, consideration of other causes of
absenteeism such as psychological issues caused by the work-
ing environment are also important. Recent observations show
that psychological diseases are responsible for every second early
retirement, and this trend is expected to increase [37]. Although
this problem is, in comparison with other industries, below aver-
age in the automotive industry, it cannot be neglected due to the
monotony and mental fatigue typical for automotive assembly pro-
cesses. A recent publication suggests to incorporate the factor of
stress and mental fatigue into the ergonomic risk assessment of
working activities [38]. Additionally, motivational aspects within
the job rotation could be integrated [39]. The incorporation of these
aspects into the assessment methods of workstations’ exposure
may also be of interest for further research.
Appendix A. Model notation
Sets and subsets
cr ∈ Cr Ergonomic risk assessment categories
i ∈ I Workers assigned to workgroup I
j ∈ J Workstations of assembly process J
jo ∈ J o Obligatory workstations of assembly process J
jno ∈ J no Nonobligatory workstations of assembly process J
qij ∈ Q Qualification of worker i for workstation j
k ∈ K Rotation rounds per shift
Decision variable
xij,k ∈ X Allocation of worker i to workstation j at rotation round k
Parameters
 Index for past rotation rounds
cij,k Penalty cost for allocation of worker i to workstation j at
rotation round k
ErgPj,cr Ergonomic exposure at workstation j in ergonomic risk
assessment category cr
HisRaij, Indexed most recent allocation of worker i to workstation j
HisFrij,n Frequency of allocations of worker i to workstation j
during considered range of past rotation rounds n
inq Unqualified worker
j’ Workstation with qualification procedure
jno∗ Nonobligatory workstation of assembly process J with
lowest priority
ko Rotation round with overcrowding
ku Rotation round with understaffing
Lij, Length of most recent allocation of worker i to workstation
j review of applications, methods and models. Eur J Oper Res 2004;153:3–27.
L Average length of daily shift
ManRij,k Manually-set restrictions for sequences of worker i to
workstation j allocation in rotation round k
n Considered range of past rotation rounds
nK Number of rotation rounds per shift
113
nJ Number of workstations to be filled with workers
nJ,k Number of workstations to be filled with workers in
rotation round k
nJ,ko Number of workstations to be filled with workers in
rotation round k with overcrowding
nJ,ku Number of workstations to be filled with workers in
rotation round k with understaffing
nI Number of present workers
nI,k Number of present workers in rotation round k
nI,ko Number of present workers in rotation round k with
overcrowding
nI,ku Number of present workers in rotation round k with
understaffing
qij,k Qualification of worker i for workstation j at rotation
round k
qinq j ,ko Qualification of unqualified worker inq for workstation j’
with qualification procedure at rotation round ko with
overcrowding
xinq j ,ko Allocation of unqualified worker inq to workstation j’ with
qualification procedure at rotation round ko with
overcrowding
References
[1] World Bank Open Data. Population ages 65 and above; 2016. http://data.
worldbank.org/; [Accessed 06.12.2016].
[2] United Nations, Department of Economic and Social Affairs, Population
Division. World population prospects: the 2015 revision. Demographic
Profiles (ST/ESA/SER.A/380), Volume II; 2015.
[3] Weichel J, Buch M, Urban D, Frieling E. Sustainability and the ageing
workforce: considerations with regard to the German car manufacturing
industry. In: Shani AB, Kira M, Docherty P, editors. Creating sustainable work
systems: Developing social sustainability. London/New York: Routledge;
2009. p. 70–83.
[4] Weichel J, Stanic S, Diaz JAE, Frieling E. Job rotation – implications for old and
impaired assembly line workers. Occup Ergon 2010;9:67–74.
[5] Badura B, Ducki A, Schröder H, Klose J, Meyer M. Fehlzeiten-Report 2015:
Neue Wege für mehr Gesundheit – Qualitätsstandards für ein
zielgruppenspezifisches Gesundheitsmanagement. Berlin/Heidelberg:
Springer; 2015.
[6] Ward T, Loire P. Employment, skills and occupational trends in automotive
industry – Annex report; 2008. http://ec.europa.eu/social/
BlobServlet?docId=3049&langId=en; [Accessed 22.11.2016].
[7] Wirtschaftswoche. Demografischer Wandel Autofirmen bauen Produktion
altengerecht um; 2012. http://www.wiwo.de/erfolg/beruf/demografischer-
wandel-autofirmen-bauen-produktion-altengerecht-um/6827948.html;
[Accessed 03.06.2016].
[8] Rademacher H, Sinn-Bahrendt A, Schaub K, Landau K. An evaluation method
for age-related bottlenecks regarding musculoskeletal workload. Z für
Arbeitswiss 2006;60(4):230–44.
[9] ElMaraghy H, Schuh G, ElMaraghy W, Piller F, Schönsleben P, Tseng M, et al.
Product variety management. CIRP Ann – Manuf Technol 2013;62(2):629–52.
[10] Tayyari F, Smith J. Occupational ergonomics: principles and applications. In:
Manufacturing systems engineering series 3. 1st ed. London: Chapman & Hall;
1997.
[11] Lussier RN, Hendon JR. Human resource management: Functions,
applications, skill development. Los Angeles: SAGE Publications; 2013.
[12] Tharmmaphornphilas W, Norman BA. A methodology to create robust job
rotation schedules. Ann Oper Res 2007;155:339–60.
[13] Leider PC, Boschman JS, Frings-Dresen MHW, van der Molen HF. Effects of job
rotation on musculoskeletal complaints and related work exposures: a
systematic literature review. Ergon 2015;58(1):18–32.
[14] Loch CH, Sting FJ, Bauer N, Mauermann H. The Globe: how BMW is defusing
the demographic time bomb. Harv Bus Rev 2010;88(3):99–102.
[15] Stanić S. Fahrzeugendmontage – Herausforderung für den demografischen
Wandel. Schriftenreihe Personal- und Organisationsentwicklung 8. Kassel:
Univ. Pr. Univ.; 2010.
[16] Ellström P. The many meanings of occupational competence and
qualification. J Eur Ind Training 1997;21(6/7):266–73.
[17] Michalos G, Makris S, Chryssolouris G. The effect of job rotation during
assembly on the quality of final product. CIRP J Manuf Sci Technol
2013;6:187–97.
[18] Azizi N, Zolfaghari S, Liang M. Modeling job rotation in manufacturing
systems: the study of employee’s boredom and skill variations. Int J Prod Econ
2010;123:69–85.
[19] Ernst AT, Jiang H, Krishnamoorthy M, Sier D. Staff scheduling and rostering: a
[20] Drexl A, Mundschenk M. Long-term staffing based on qualification profiles.
Math Methods Oper Res 2008;68:21–47.
[21] International Ergonomics Association. Definition and Domains of Ergonomics;
2016. http://www.iea.cc/whats/index.html; [Accessed 05.04.2016].
[22] Schaub K, Caragnano G, Britzke B, Bruder R. The European assembly
worksheet. Theor Issues Ergon Sci 2013;14:616–39.
114 J. Hochdörffer et al. / Journal of Manufacturing Systems 46 (2018) 103–114
[23] Schaub K, Rademacher H, Caragnano G, Guth A, Bruder R. Ergonomics risk
assessments in automotive and electrical industry based on the dual European
concept of health and safety at work. Ergonomics is a lifestyle – Proceedings
of the 40th annual conference of the Nordic ergonomics society 2008:11–3.
[24] Carnahan BJ, Redfern MS, Norman B. Designing safe job rotation schedules
using optimization and heuristic search. Ergon 2000;43:543–60.
[25] Nanthavanij S, Kullpattaranirun T. A genetic algorithm approach to determine
minimax work assignments. Int J Ind Eng 2001;8:176–85.
[26] Kullpattaranirun T, Nanthavanij S. A heuristic genetic algorithm for solving
complex safety-based work assignment problems. Int J Ind Eng
2005;12:45–57.
[27] Aryanezhad MB, Kheirkhah AS, Deljoo V, Al-e-hashem M. Designing safe job
rotation schedules based upon workers’ skills. Int J Adv Manuf Technol
2009;41:193–9.
[28] Yoon S-Y, Ko J, Jung M. A model for developing job rotation schedules that
eliminate sequential high workloads and minimize between-worker
variability in cumulative daily workloads: application to automotive
assembly lines. Appl Ergon 2016;55:8–15.
[29] Asensio-Cuesta S, Diego-Mas JA, Canós-Darós L, Andrés-Romano C. A genetic
algorithm for the design of job rotation schedules considering ergonomic and
competence criteria. Int J Adv Manuf Technol 2012;60:1161–74.
[30] Costa AM, Miralles C. Job rotation in assembly line employing disabled
workers. Int J Prod Econ 2009;120:625–32.
[31] Bhadury J, Radovilsky Z. Job rotation using the multi-period assignment
model. Int J Prod Res 2006;44:4431–44.
[32] Otto A, Scholl A. Reducing ergonomic risks by job rotation scheduling. OR
Spectr 2013;35:711–33.
[33] Diego-Mas JA, Asensio-Cuesta S, Sanchez-Romero MA. A multi-criteria genetic
algorithm for the generation of job rotation schedules. Int J Ind Ergon
2009;39:23–33.
[34] Federal Anti-Discrimination Agency. Act Implementing European Directives
Putting Into Effect the Principle of Equal Treatment; 2015. http://www.
antidiskriminierungsstelle.de/SharedDocs/Downloads/DE/publikationen/
AGG/agg in englischer Sprache.
pdf;jsessionid=AA6BF964775A94F266F74F11D2F66039.2 cid350?
blob=publicationFile&v=2; [Accessed 22.11.2016].
[35] Domschke W, Drexl A, Klein R, Scholl A. Einführung in operations research.
9th ed. Berlin, Heidelberg: Springer Gabler; 2015.
[36] Lau HC. On the complexity of manpower shift scheduling. Comput Oper Res
1996;23:93–102.
[37] World Health Organisation. Global burden of mental disorders and the need
for a comprehensive, coordinated response from health and social sectors at
the country level: Executive Board 130/9; 2011. http://apps.who.int/gb/
ebwha/pdf files/EB130/B130 9-en.pdf; [Accessed 03.04.2016].
[38] Lodree E, Geiger C, Jiang X. Taxonomy for integrating scheduling theory and
human factors: review and research opportunities. Int J Ind Ergon
2009;39:39–51.
[39] Wongwien T, Nanthavanij S. Ergonomic workforce scheduling with
productivity and employee satisfaction consideration. Proceedings of the 4th
International Conference on Engineering, Project, and Production
Management (EPPM) 2013:1108–16.
Jan Hochdörffer holds a M.Sc. degree in Industrial Engineering and Management
from KIT. Currently, he is a Research Associate at the wbk Institute of Production
Science at KIT focusing on product, manufacturing network optimization, and pro-
duction system planning.
Marc Hedler recently graduated and likewise holds a M.Sc. degree in Industrial
Engineering and Management from KIT. He works as a Research Assistant at the wbk
Institute of Production Science at KIT in the field of production system planning.
Dr.-Ing. Gisela Lanza holds the Professorship “Production System and Quality Man-
agement” at KIT and is Head of the wbk Institute of Production Science. Her division
“Production Systems” mainly deals with the areas of global production strategies,
production system planning, and quality management in research, industrial appli-
cation, and teaching.