Robotics and Computer–Integrated Manufacturing 68 (2021) 102092

Contents lists available at ScienceDirect

Robotics and Computer-Integrated Manufacturing

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

Digital twins for collaborative robots: A case study in

human-robot interaction
Ali Ahmad Malik a, *, Alexander Brem b
a
Siemens Gamesa Renewable Energy, Brande, Denmark
b
Chair of Entrepreneurship in Technology and Digitization, Endowed by the Daimler-Fond in the Stifterverband, University of Stuttgart, Stuttgart, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Human-robot collaboration (HRC) can expand the level of automation in areas that have conventionally been
Human-robot collaboration difficult to automate such as assembly. However, the need of adaptability and the dynamics of human presence
Cobot are keeping the full potential of human-robot collaborative systems difficult to achieve. This paper explores the
Collaborative robot
opportunities of using a digital twin to address the complexity of collaborative production systems through an
Digital twin, Demonstrator
Simulation
industrial case and a demonstrator. A digital twin, as a virtual counterpart of a physical human-robot assembly
system, is built as a ‘front-runner’ for validation and control throughout its design, build and operation. The forms
of digital twins along system’s life cycle, its building blocks and the potential advantages are presented and
discussed. Recommendations for future research and practice in the use of digital twins in the field of cobotics are
given.

1. Introduction repeatability of machines - the advantages of production flexibility,

product mix and reconfiguration can be achieved [9]. One approach of
In the recent decade, a kind of silent revolution has changed the way doing this is by using the new generation of industrial robots named
manufacturing is planned and performed. A new manufacturing para­ collaborative robots or cobots [10,11]. Cobots, integrated with
digm is emerging with shorter product life cycles, mass personalization advanced safety technologies [12], are designed for sharing their
and interactions between humans and robots in close proximity [1, 2]. workspace with humans thus achieving the ‘right’ amount of automa­
The manufacturing landscape is getting increasingly complex and pro­ tion [13]. However, as the level of collaboration grows, the systems tend
duction plants are becoming multiplexes of highly complex biome­ to become complex. Due to this reason, the industrial application of
chanical elements [3]. cobots in labor intensive processes (e.g. assembly) is still limited [14].
Similarly, the need for being adaptive, flexible and cost effective is A way to quickly and safely (re) design and integrate collaborative
growing. Humans have proven to be a flexible production resource but systems is by utilizing a virtual space [15]. Computer based virtual
the cost of human effort is continuously increasing, especially in the models of physical systems can be used to test and validate production
context of emerging markets, leading to new approaches of generating strategies before taking them into practice [3]. Although, this is also a
new products and services [4]. As a consequence, human effort is conventional approach of virtual modelling but the new ‘lifecycle’
effectively being replaced through automation [5]; however, conven­ approach is the concept of a digital twin (DT) [16] – an intelligent digital
tional automation solutions are not human friendly, and can’t coexist representation of a physical system – enabled by the advancement in
with humans [6]. Due to this reason, a large proportion of processes in a virtualization, sensing technologies and computing power [17].
manufacturing value chain are still human intensive [7]. Furthermore, This article is focused on the questions on what and how a digital
with the COVID-19 pandemic, the need of integrating human-friendly twin can contribute to address the complexity of a collaborative pro­
robotics is evident to address supply and demand challenges while duction system. Based on an industrial assembly case, this paper de­
maintaining social distancing in factories [8]. scribes the development of a digital twin and investigates its usefulness.
If the human and automation can be combined - creating a balance The case is an assembly problem because assembly is labor-intensive
between the flexibility of manual processes and the efficiency and work and has high potential for human-robot collaboration. The case

* Corresponding author at: Syddansk Universitet, Denmark.

E-mail address: [email protected] (A.A. Malik).

https://doi.org/10.1016/j.rcim.2020.102092
Received 19 September 2019; Received in revised form 8 November 2020; Accepted 9 November 2020
Available online 21 November 2020
0736-5845/© 2020 Elsevier Ltd. All rights reserved.
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

owning company is located in Denmark which is a high-wage economy elements was turbulent (e.g. an assembly line). Complicated
and is facing the challenges of globalization. Lastly, the case solution manufacturing systems, with a network connected components are the
aims to use human-robot collaboration where human and robot are result of the 3rd industrial revolution. Networked components are multi-
sharing the same space and time, meaning a high degree of component structures but the connection between the components is
collaboration. linear and straight forward. As the adoption of computers in
manufacturing activities surfaced, the behavior of the manufacturing
2. Theoretical background systems became largely predictable. An example of a complicated sys­
tem is an automotive assembly line of the modern era.
To strengthen the contribution of this study, the following section Whereas complex systems are a large network of components with
forms a theoretical foundation of the relevant research. The dimension many to many communications and with sophisticated information
of complexity in manufacturing systems is discussed and how digital processing that makes the behavior of a system difficult to predict.
twins have emerged to manage the complexity. Complex systems have a property of surprise, as in ‘I didn’t see that
coming’, and such a surprise can even be fatal [20].
2.1. Complexity in manufacturing systems With growing automation and many to many relationships, the
complexity of predicting the behavior is growing. It is even more
Systems’ theory defines a system [18] as: ‘two or more components that important if the ‘surprise’ can result in an accident, especially if the
system involves interaction with humans. There are other elements of
combine together to produce from one or more inputs one or more results that
could not be obtained from the components individually’. The components time, energy and effort that can result in economic wastes if the ‘sur­
prise’ brings a system to a failure.
in a system (as individual entities) contain and (their mutual interac­
tion) produce information. The quantity of this information proliferates
2.2. Complexity in human-robot collaborative systems
during the system’s life cycle, making it increasingly difficult at any
phase for an external observer to study or predict its future behavior.
Humans are complex. Not only as a group of humans, but even as
The level of difficulty faced by an observer (studying a system) can be
individuals, they exhibit a complex behavior which is often difficult to
referred to as system complexity. However, there is a little agreement on
predict [19]. The interaction of humans with another system (existing at
the term complexity and therefore, it is understood in various contexts.
any phase of knowing) can escalate the total information content and
Nevertheless, two important dimensions of a system complexity are its
information content and the predictability of its interaction-behavior. fabricate unpredictability, creating some sort of ‘complexity’. This is
evident from Perrow’s work, ‘Normal Accidents’ [21] to describe the
Another approach to understanding systems’ complexity is the
Cynefin diagram (Fig. 1) that describes systems in four ‘phases of inherent danger of complex systems and pointed to a common thread of
‘human-element’ when they are interacting with complex systems [20].
knowing’ i.e., simple, chaotic, complicated and complex [19]. The
differentiating property of each of these phases is their behavioral With the rise of Industry 4.0 technological trends, a growing ten­
dency in manufacturing automation is human-friendly robots or cobots
predictivity.
In a manufacturing context, the systems’ forms of knowing can be (Fig. 2). A cobot is a robot designed (both in its design and application)
to be safe when working alongside humans [22]. Since humans are
related to the four industrial revolutions. Starting from the 1st industrial
revolution, the manufacturing systems (e.g. mechanized power looms) prone to errors, human-robot systems do not always perform flawlessly.
From a system’s perspective, the complexity of an HRC variant oriented
were entirely predictable with obvious and transparent inputs, opera­
tions and outputs. The 2nd revolution gave rise to chaotic systems where production system can be described in four different dimensions i.e.,
many components, non-linearity, relationships between components
the complexity started to grow but the relationship between a system’s
and reconfiguration. An answer to the complex behavior of an HRC
system is to somehow (re) design the system in retrospect, predict the
future under maximum known variables and implement it.

Fig. 1. Complexity and phases of knowing in the evolution of

manufacturing systems. Fig. 2. Human-robot collaborative workstation.

2
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

2.3. Digital twins to address complexity production variations.

With hyperscale computing, commoditization of sensor equipment,
A digital twin (DT) is a digital representation of both the components augmented/virtual reality [28] and data analytics platforms, it has
and dynamics of a physical system; enabled by the advancement in become possible to develop time dependent three-dimensional models
virtualization, sensing technologies and computing power [17]. It can of complex physical systems. It extends the usefulness of virtual models
mirror the real-time operating conditions of a physical system [23]. The developed in the design phase to the lifecycle of a system.
concept of DT emphasizes the utilization of a digital model of a physical
system and link each component of the virtual model to the corre­
2.4. Related work
sponding physical asset. By doing so, the digital twin becomes a mirror
and can be used as a ‘front-runner’ to the future behavior of its associ­
A multitude of definitions and understandings of digital twin
ated physical twin or system.
conception are available in the literature, however, the core idea hovers
The concept of DT was evolved from a hardware twin that was
around a synchronized multi-purpose simulation of a physical system
developed in NASA’s Apollo program consisted of two identical space
that (through sensor updates etc.) mirrors the life of its physical twin
vehicles. During the space mission, one vehicle remained on the ground
[29, 30]. Although the literature displays application of DT in various
while the second went up-to the space. The ground vehicle was
contexts such as aircraft design [31], architectural buildings [32],
continuously mirroring the flight conditions of the in-orbit vehicle to
healthcare [33], and farm management [34] etc.; but, an overwhelming
enable ground experts to better assist astronauts in orbit [24].
interest has been observed from manufacturing applications. Several
A different approach to the concept of DT as a ‘Conceptual Ideal for
contributions have been made in this regard such as digital twins for
PLM’ (PLM = Product Lifecyle Management) was presented at the
product design and service [35], digital twin shop floor [36], DTs for
University of Michigan [25]. The approach assumed that every system
factory design [37] and for production line performance [38] etc.
can be considered as a subset of two systems i.e. the physical system
Nevertheless, most research is available in the form of conceptual
existing in the physical world, and a virtual system existing in virtual
models and simulations but application to real-world scenarios is
space and containing all (necessary) information about the physical
limited.
system. The bi-directional relation between physical system and digital
The keynote paper [39] of the CIRP (Society of Production Engi­
system can make insights into a system’s performance improving
neering) pointed towards the importance of DTs for HRC production
product design, manufacturing and service throughout the system’s life
systems. It was presented that a DT can simplify combining and aligning
cycle [24].
the function, structure and behavior of an HRC cell with symbiotic
As already said, the basic idea of having an (informational) virtual
interplay of the humans in virtual models. Grieves [40] (among the
twin to understand the complexity of a physical system is not new. En­
earliest researchers in the field of DT [41]) has also presented the po­
gineering design and build problems have always been solved through
tential value of DTs for the development of the field of cobotics.
an informational twin. In earlier times, when developing a system, the
A digital twin integrating human-robot collaboration with level of
virtual twin was an image in human-mind [26], though limited in its
autonomy for container unloading was presented by [30]. The authors
ability to answer questions about system’s performance. Only in the
discussed the combination of adaptive automation and digital twin
mid-20th century, it became possible to draw a virtual model as a
enabling an operator to remotely control an autonomous system if
2-dimensional (2D) CAD (computer aided design) object and then as 3D
needed.
models, and later as dynamic simulations (Fig. 3). During development
The significance of digital twins for HRC assembly systems has been
of a system, these virtual models are created early as the first form
highlighted e.g. by [3] where the relevance of DT in relation to
containing maximum details of the proposed physical system. However,
human-robot collaboration was discussed. It was presented that DTs can
in practice, these virtual models get useless once the system has evolved
support HRC systems however, the study remained limited only to the
into its operational form. The reason is the effort required to make the
HRC challenges in operational phase without a lifecycle approach.
digital model intelligent to make sense of the operational behavior and
Various aspects of achieving a digital twin such as forms of DTs along
comparing with its algorithmic solution and assisting in day to day
system lifecycle, building blocks of a DT and demonstration of the

Fig. 3. Use of CAD model during system development [27].

3
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

discussed concepts were also missing. digital twin be developed as early as possible during the idea generation
The life cycle context of production systems and their virtual and must evolve in parallel to the physical twin. As the development
modeling has been presented by Kovsturiak [42]. The study made rec­ evolves, both the DT and PT must continuously be updated as a reflec­
ommendations for effective utilization of discrete-event simulation for tion of each other.
design, operation and continuous improvement of complex As with all twins, there is a first born. In the case of digital twins of
manufacturing and logistics systems. Kibira [43] presented different production systems, the first born is the digital twin [40] because the
methods and tools for simulation application in life cycle of production idea, shape, functionality always precedes the actual realization of the
plants. It was argued that traditional simulation models built for concept physical form of the product . An HRC production system designed with
development and design are different from models for operational a DT approach can offer:
decision-making, but the distinction is getting blurred.
Two comprehensive works on digital twins have been [29, 23]. - Fast integration,
However, the studies do not cover the human-robot interaction aspect - Production reconfiguration, and
specifically. - Safety validation
To cover these gaps, the present study is contributing by taking a
holistic life cycle approach of digital twins for human-robot assembly To make DT and PT useful for each other, a DT assessment-model is
systems. The building blocks and various forms of DT are presented. A suggested. The assessment model is composed of sensing the data,
physical demonstrator of an industrial assembly case is developed evaluating the data, defining non-compliance, setting up a solution,
together with its DT to substantiate the presented concepts. simulating the solution, forming a strategy and implementing it. The
assessment model is usable for HRC system’s life cycle to optimize sys­
3. Digital twin driven HRC system tem performance.
The following sections describe phases, building blocks and appli­
From authors’ theorization, a digital twin for a human-machine cation of DT along the life cycle of an HRC production system.
system can be defined as, “A digital multidimensional data-connected
representation of the elements and dynamics of a collaborative bio­ 3.1. Forms and phases of an HRC digital twin
mechatronic complex physical system to simplify its designing, inte­
gration, operation and reconfiguration along the life cycle”. By integrating DT into systems engineering for HRC, the physical
A digital twin model for an HRC production system is shown in Fig. 4. system can have its informational twin all along its life cycle. The
The digital twin system is composed of two interconnected spaces i.e. informational twin or DT co-evolves with the physical system and takes
physical and digital. The digital space is a three-dimensional virtual different forms corresponding to level of details, connectivity with the
representation of the coexisting or collaborating human and robot while real-world data and its synchronization with the rest of the systems.
the physical space is the real production system composed of humans, Thereby a classification of DTs can be established (Fig. 5).
robots and other hardware. Each element in virtual space is displaying Two forms of DTs have been discussed in the literature [26] [29]; (a)
the design and operating behavior of a connected physical object in the DT that contains static and dynamic informational sets required to
physical space. describe and produce the physical entity; and (b) a DT where different
‘Systems engineering’ deals with the whole lifecycle of a (complex) information sets (monitoring, sensing, service etc.) are integrated for
system i.e. from analysis and design, through implementation and ubiquitous tracking. However, a DT can have additional forms, and,
operation to its redesign and reconfiguration [42]. The activities in with a system engineering approach, these can be better described
design, implementation and operation must be structured and iterative depending on the life cycle phase of its physical twin.
[44]. It is proposed that when developing an HRC production system, its

Fig. 4. Digital twin in an HRC system.

4
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 5. Forms and phases of a digital twin.

3.1.1. Digital twin – Design 3.1.4. Digital twin – Operation

When developing a new physical system, a digital twin is often The DT is extended for real-time communication with the physical
developed earlier than the physical twin. It can help to generate and system during its operation for behavioral analysis and performance
validate the initial design, behavior, layout etc. Although, at the design optimization. The interlinking of production planning and control da­
phase, the corresponding physical twin is not existing, but the DT-Design tabases can support dynamic scheduling of the production orders and
is still referring to a hypothetical future physical twin. Even without a allocation of workforce. The DT can simplify the reconfiguration or
real-time connectivity, the digital twin of the production system can repurposing of the production system under demand fluctuations.
make it possible to experiment several what-if scenarios for achieving In DT-Design the assessment is done manually. However, in DT-
faster, safer and better design. The parameters to achieve an effective Operation, the real-world data is synchronized with the DT and the
HRC production system may include selection of robot arm, workstation assessment cycle can become automated. Human effort is still needed
design, layout, fixtures and economic evaluations. when a solution to the given situation is generated and is presented to
the human user before allowing its implementation.
3.1.2. Digital twin – Development
Moving towards DT-Development, the results from the DT-Design 3.1.5. Digital twin – Maintenance
are utilized to develop elements of the physical twin corresponding to Maintenance is an integral part of most (if not all) production sys­
their counterparts in the digital twin. An HRC system may involve tems. Mixed reality technologies such as augmented reality (AR) or
development of the workstation, fixtures, feeding devices and other chatbots can be integrated with a DT to enable maintenance personnel
hardware. Bill of materials (BOM) and bill of processes (BOP) can be for better maintenance, fault detection and training.
generated and the physical system is developed accordingly.

3.1.3. Digital twin – Commissioning 3.2. Building blocks of a digital twin system
In this phase, initial connectivity between the physical and the dig­
ital twins can be established by connecting the DT to a real controller or A digital twin system is composed of several components such as a
PLCs to detect potential errors. This method is similar to virtual physical space, a digital space, data communications and the connec­
commissioning (VC) [45]. VC or hardware-in-the-loop simulations can tions. Some of them (the digital and the physical space) are primary
help in reduction of development time by virtual testing and integration elements, while the need of the secondary elements (data communica­
well before the real commissioning [46]. The real robot in PT can be live tions and the connections etc.) is dependent upon the expectations from
connected with the virtual robot in DT enabling the physical robot to a DT system.
perform the tasks as designed in the DT.
3.2.1. The digital space
As suggested by Tao [35] the DT is built in four layers i.e. geometry

5
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

(creation of 3D CAD objects), physics (kinematics of robots and human), created. The environment must be able to sync data with other
behavior (placement of CAD objects in the scene), and rule (assembly relevant software e.g. CAD modelers, statistical analysis tools, data
process sequence). Since each element of a digital twin corresponds to its loggers etc.
counterpart element in the physical twin the modelling of digital space
consists of identical activities as a physical system (Fig. 6). Although the a) DT of assembly components
digital space is an integral element of the DT-system however, a stand­
alone digital space without an interlinking with a physical twin is A holistic DT of an HRC system begins with the digital modeling of
incomplete. the components being assembled in the system. It may be desired to
It is important that a DT needs to simplify the complexity and create redesign the components for ease of automated assembly (Design for
value for its users. When modelling a DT, not all elements or aspects of Cobot Assembly) but that is out of scope of this study. This study
the PT need to be modelled because information retrieval has a cost and assumes no change in the design of the product, but virtual models of
is a tradeoff with time, energy and effort. For example, the rotation of a the assembly components are necessary to holistically model the DT
screwdriver can be modelled and tracked in the DT to know real-time processes.
torque, and speed. However, if in the given case, this information is
not creating any value then it is of no use and is only increasing the a) DT of production resources
complexity.
The digital space is composed of different elements such as: The digital model of each of these can be (1) supplied by the vendor;
(2) offered in the library for standard industrial equipment; (3)
a) DT Environment created via intensive scans of the part reconstructed into a CAD
model; or (4) designed and modeled by the designer (especially for
The digital twin environment is the space where the digital twin is the custom designed components and sub-systems). The digital ele­
created, visualized and interacted (e.g. on a computer screen). It is a ments are imported into the DT environment and are placed ac­
3D space where virtual models of production systems are imported or cording to the production requirements.
Digital models of robotic manipulators are often provided by the
robot manufacturer. The kinematics, joint limits and maximum joint
speeds may need to be defined. The dynamic behavior and con­
straints of other equipment such as, grippers are defined similarly.
Visually accurate 3D digital mannequins can be developed to opti­
mize human activities in computer. The digital human models are
often based on databases of human body shapes such as ANSUR II
(Anthropometric survey of US army personnel) database [48].

a) DT of processes and services

A static digital model of the production system is conceived from the

previous step. The next step is to define the dynamic behavior of
human and robot to execute the tasks. The performance of robotic
and human actions needs to be continuously tracked; therefore,
continuous simulation is used at this stage.

3.2.2. The physical space

A DT always refers to a physical space. Although at the design phase,
for example, there is no physical space available, but the DT is still
referring to a hypothetical future physical space which is driving the
evolution of the digital space. The physical twin is conceived once the
design validation is completed in the virtual twin.

3.2.3. DT connections
The connection between the physical space and the digital space is
established. The connection or the data syncing can both be manual or
automated. However, at the operational phase the data communication
must be automated.

3.2.4. DT data
It is the data repository to save logs and process related data
generated by the digital twin, the physical twin and their interlinking.
There can be more data than needed and it is important to decide what
data should be recorded that can help in optimizing and improving the
performance of the system.

3.3. Usability of a digital twin for HRC production systems

The following section provides an overview of the domains where

the digital twin can help in systems engineering of an HRC production
Fig. 6. The structured process of designing a collaborative assembly sys­ system. The usefulness is described in three phases of design, integration
tem [47]. and operation.

6
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

3.3.1. The design phase key positions (pick and place) for the robot and generating robot
Selection and optimal placement of production resources is impor­ trajectory. A final assembly plan can be developed depending upon
tant in a production setup to avoid wastes. A deficient workplace design task sequence, task time and avoiding any idle time.
also causes physical stress, reduces productivity [49]. The co-existence
of humans and robots imposes further safety implications when plan­ a) Robot control program
ning the layout. In this regard, the following experiments can help to
achieve a safe and waste free layout: A great deal of human effort in robotic applications is associated with
simplifying the programming of robots [53]. Even with the devel­
a) Reach and placement evaluation opment towards ‘intuitive user-interfaces’ – often integrated in
cobots - the programming can be a time consuming and tedious task.
The selected robot (under its joint limits) has a certain allowable With the use of a DT, the robot program is intuitively generated in
reach distance. The placement locations of the robot as well as the the DT environment. The generated robot program is transferred to
equipment will determine if the robot can safely reach the desired the connected robot that starts working as a reflection of the robot in
locations within its workspace. These tests can be performed in a the digital twin.
simulated environment. Similarly, the reachability of human arms An online connection between physical and virtual robot can
(depending upon its body measurements) without bending the body dynamically transmit robot manipulation and actions made in
can be examined. The goal is to have minimum cycle time, minimum physical space to the robot in the virtual space and vice versa (Fig. 7).
collisions and safer working conditions. Thus, avoiding any need for additional programing. Since the DT
performs robotic actions in a robot understandable language, the
a) Collision tests robot program from the physical robot can also be exported to DT to
for any verifications.
In an HRC, where humans are interacting with cobots and frequent
production changes may be desired, altering the human-robot tra­ a) Data logs and performance analytics
jectories to move from one point to other. Collisions are likely to
occur in such a dynamic environment. Although cobots are designed Real-time performance metrics, optimization analytics and alerts for
for workspace where a collision between human and robot is likely to system performance can be recorded in a data repository for per­
happen, however frequent collisions reduce productivity. Efforts formance evaluations. Data connectivity in the operation phase is
must be made to design the robot trajectories encountering minimum needed to feed in real world data and making evaluations under
collisions with the human. continuously changing variables.

a) Ergonomic analysis 4. An industrial case

Manual work is associated with various human performance con­ To demonstrate the design and development of a digital twin enabled
cerns including work postures, load on human body during weight HRC assembly system, a case from a manufacturing company is used
lift and task frequency. A large variety of these ergonomic issues of (Fig. 8). The case is about production of battery packs used with elec­
production systems can be simulated and tested early in the design tronic linear actuators. Electronic linear actuators, for example, are in
phase [50]. hospital beds for automated linear movement allowing patients with
limited mobility to change their position without needing human
3.3.2. The integration phase assistance. The manufacturing company has been facing challenges of
By having a mix of physical and digital resources a mixed reality high labor costs and therefore wanted to invest in flexible automation
environment can be enabled. For example, a physical robot can be solutions.
connected with a virtual robot (available in the DT model) to study its
behavior. When connected, the virtual robot will follow the trajectory of 4.1. Problem description
the physical robot. If the physical robot is in an empty area, the virtual
model can be populated with tools, equipment and humans to create the The battery-pack consisted of eight unique parts; corresponding to
future scene of the physical space. Thus, a task executed at the physical eight manual assembly tasks (each component corresponds to one task).
robot enables safe analysis in the virtual space. Similarly, the signals and The conventional mode of production was manual, and tasks were
cycle times can be tested with the use of PLCs and defined PLC logics. divided in between two human operators taking 60 seconds each. The
HRC cell is supposed to reduce the person hours but maintaining the
3.3.3. The operation phase same production rate.
Each assembly task is evaluated for its ease of HRC automation. Task
a) Dynamic task allocation evaluation for cobot automation is different from conventional robotic
automation as additional parameters, particularly for safety implica­
To complete the assembly, humans and robots both have some tions, need to be considered. A complexity-based task allocation method
unique as well as common capabilities. Choosing the right resource [54] can be used to decompose each task into its attributes, and
to perform a task can be based on the skills of the humans and the assigning an automation potential score thus identifying the tasks car­
robots [9, 51]. Various other factors can also influence the suitability rying higher automation potential, assigning the right resource and
of a task for the robot e.g. physical and geometrical properties of the balancing the assembly process.
components, assembly characteristics, safety implications, feeding As shown in Table 1, a total of two assembly tasks and two supple­
and joining method involved. Method to quantify the automation mentary (pick and place tasks) are selected to be designated to the robot.
complexity of assembly tasks has been discussed by [52] which takes The remaining six tasks (requiring precise adjustments) are kept
care of assembly attributes contributing to automation complexity manual.
deducing a score as a representation for automation potential can be
generated. A decision-making arena can be formed where the tasks 4.2. Building the digital twin
can be classified as robotic, manual or either robotic or manual. The
DT model can estimate the task time for each task considering the The digital twin (digital space in the DT model) is a three-

7
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 7. The robot program in physical and digital twin.

dimensional virtual environment composed in Tecnomatix Process if the sub-assembly is available in the fixture (Proximity sensor_1),
Simulate (TPS) software (Fig. 9). The proposed HRC assembly station switch is pressed by the human (SWITCH_1), and the robot is at home
comprises of a robot manipulator and a human operator that jointly position (determined through a joint value sensor). If the logic is ful­
complete the assembly in a collaborative fashion. The suggested robot is filled, the robotic task is executed. A logic driven simulation offers
a Universal Robot UR-5 e-series. The robot has 6 degrees of freedom, a extended control over the process and additional logics can be defined to
payload capacity of 5 kg and a reach of 850 mm. The robot is equipped create what-if scenarios.
with a parallel-fingers SCHUNK gripper EGP 64-N-N-B with a finger Each robotic and manual task is performed in the DT environment. A
length of 40 mm. The gripper in its default design is not enough to logic is defined that ensures the completion of each task and initiates the
handle large parts (>120mm) used in the case study. Therefore, next task thus forming an event driven simulation. Finally, a process
extended fingers are designed to be produced with additive plan is generated that shows the cycle time for each task.
manufacturing.
Computer aided design (CAD) model of the cobot is received from
the robot manufacturer. It is imported into the simulation environment 4.3. Tests and analysis
and its upper/lower joints limits and acceleration are defined as per
robot specifications. Other hardware resources (table, trays, fixtures Once a digital twin model is achieved for the proposed physical HRC
etc.) and the assembly components are created in Autodesk Inventor as assembly system, the tests are performed to deal with different issues of
3D CAD models; and are imported into the TPS environment. TPS can system design. An effective HRC layout must ensure maximum utiliza­
import CAD data in JT (Jupiter Tessellation) format. A digital human tion of resources, minimum movements, least travel time from one
mannequin of the human worker is integrated to conduct ergonomic location to the next and the safety of human operators.
evaluations. Since the case-factory is located in Denmark, an average By performing the assembly tasks in DT, an accurate estimation of
size of human population is selected from the Danish statistics report the task times can be generated (Fig. 11). The task times helped to refine
2017 [55]. The integrated human mannequin corresponds to a female the process balancing and get it aligned with the takt time. The reach test
with a height of 182 cm, body-mass index (BMI) 24.5 kg/m2 and waist to of the robot arm - with respect to its placement location - displayed the
hip ratio of 0.85. points where the robot can reach, identifying placement locations for
To detect the presence or position of assembly components in the components and fixtures. A human work-envelope (given the body di­
assembly station, virtual sensors are integrated to emulate the physical mensions) identified the space that human can safely reach. A vision-
sensors. The sensors used in the DT are listed in Table 2 while their lo­ window enabled the scene to be viewed through the line of sight of
cations are shown in Fig. 10. the human mannequin. Considering the findings, a layout is developed
Task transition logics are defined for each robotic task that examine and validated in the DT environment. Robot swept volume (Fig. 12)
the conditions to be fulfilled when activating the robot (Fig. 10). For identified the space occupied by the robot during its operation. Mini­
example, the Task 3 (pick and place of batteries by the robot) is executed mum human activity is ensured within the robot swept volume.
To define human actions, a motion capture device (Microsoft Kinect)

8
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 8. Assembly components considered for HRC automation and the manual cell being used in the case company.

After an optimal design of robot and human tasks is achieved in the

Table 1
DT - moving towards system integration - the robot program was tested
Assembly tasks assignment to human and robot.
with the real robot. The robot program generated from the DT model
Sr Task Manual Precedent Automation Task Suggested was downloaded into a real robot. A live connection was developed
No name time score time resource
between the DT robot and the real robot. The physical robot followed the
1 Base 24 s 0 35% 24 s Human defined trajectory as in the DT but was placed in an empty space how­
plate ever, the DT robot was surrounded with equipment and virtual human.
2 PCB 12 s 1 85% 12 s Human
3 Batteries 11 s 2 60% 12 s Robot
The collisions were identified in DT that pointed areas needing layout
4 Wires 17 s 3 10% 17 s Human and/or robot path optimization (Fig. 13).
5 Foam 15 s 4 15% 15 s Human
6 Pick & 90% 10 s Robot
place 4.4. Building the physical twin
7 Stickers 17 s 0 75% 17 s Human
8 Top 2,5 s 7 75% 2,5 s Human
cover After generating the necessary design information, the next step was
9 Screws 21 s 8 90% 21 s Robot to generate bill of materials from the DT. The physical twin is built as a
10 Pick & 9 90% 12 s Robot modular fixturing table (Fig. 14). For each robot task, a gripping device
deliver
is needed to pick the parts securely and to move to the point of action or
delivery. Suction device is designed for gripping the components with
is used to record accurate and realistic human motions for the given flattish surfaces and without gripping features available for parallel-
tasks. It integrates the mechanics of gaming technology into the DT fingers gripper. An electric screwdriver also needs to be handled by
environment to have high degree of connectedness between human and the robot for screw driving tasks. Mountings to hold these devices are
the software. The human-operator performed the assigned tasks in front designed and produced with additive manufacturing to enable the robot
of the camera. The motions are real-time connected to the DT and each to grasp the devices during operation.
manual task was recorded and integrated with the robotic tasks. It Two activation switches are enabled for the operator to activate the
avoids the need to manually simulate the manual tasks. robot to perform a task. A part presentation strategy is also needed that
A virtual robot controller is used in the DT model that translates defines the way the components are presented to the robot. In this case,
robot movements into robot understandable language. A general- no camera is used rather fixtures are utilized to lock the pick and place
purpose robot controller in the simulation can be used to generate locations.
robot trajectory, however accurate robot trajectory must be generated The fixture required to hold the components during HRC assembly
by using manufacturer-specific robot controller. Virtual controller for are produced with additive manufacturing. Fixtures are needed both for
UR5 robot is integrated with the DT model to generate accurate robot the manual and the robotic tasks. They fix and lock the position of
path in robot codes readable by the real robot. components to repetitively performing assembly tasks and helped the
robot to repeatedly pick and place the components from the same

9
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 9. Digital twin model of the HRC system.

(c) Productivity: the number of completed assemblies

Table 2
Sensors usable in an event driven DT of the HRC assembly cell.
A total of 30 assemblies were completed and the above cited event
Sr Sensor type Description logs were recorded. During experimentation, the robot and human
Nr.
collided 8 times. A significant waiting time (>10 sec) was observed in
1 Proximity This binary sensor is activated when an object enters with a total of 10 times that caused a variation in assembly cycle time (targeted
specified distance from a selected resource at 144 seconds). The data was used in the digital twin stochastic simu­
2 Photoelectric The sensor is activated when an object crosses the path of a
defined beam emitted from the sensor
lation that computed the production of 300 assemblies during an eight
3 Property This binary property sensor is activated when an object hours’ shift (Fig. 16).
with a certain property enters in a specified range To refine the process, the location of the fixtures was changed and
4 Joint value Sensor to obtain online feedback of a specified joint the cobot trajectory was optimized by creating a virtual wall at the end
5 Joint A sensor linked to the joint of a device or robot doe its
of the fixtures. Thus, making the robot to not to move beyond fixtures
distance detection range
and enter the human workspace (Fig. 17). Furthermore, to the assembly
station was divided into two zones for robot manipulation. The robot
location. Two fixtures are enabled thereby the human and the robot can moves only in the direction of the human when it enters the human
work in parallel. workspace to enable the human to see the robot and not to collide with
The utilization of the HRC system is shown in Fig. 15 and is visible it. Beyond human workspace, the robot can move in either direction.
that both the robot and the human remain active during the cycle and it The robot program was downloaded and integrated with the robot. After
takes nearly 144 sec to complete the assembly. However, both the the optimizations, another experimentation with 30 assemblies was
human and the robot can work alternately on different assemblies and performed. This helped to reduce the collisions and waiting time events.
this is why two assembly fixtures are enabled at the assembly station. The optimization caused to reduce the collision events to 1 and reduced
Furthermore, the long cycle time can be reduced by using better safety the waiting time events to 3. With the given data, the system was able to
technologies and using the robot at higher speed. produce 314 assemblies which is slightly less than the manual assem­
blies. The reason is attributed towards the slow speed of the robot that
can be optimized by having advanced safety features.
4.5. Data logging
5. Discussion and further research
The cobot is connected with a cloud data repository through an
internet-based router for data logging. The performance of the HRC The main contributions of this paper are as follows. (1) It presents the
assembly system in the case example is based upon the idle time (both need and usefulness of digital twins in designing, developing and
for the operator and the robot), safety of the human operator and the operating human-robot production systems. (2) A demonstrator is pre­
number of completed cycles. Three variables are recorded in the robot sented of an HRC digital twin based on an industrial case; Several results
program which are: from design to operation of an HRC system are discussed.
In the design phase the digital twin was used to design and select the
(a) Safety of the human operator: determined by the activation of resources of the assembly system corresponding to the production re­
force-torque sensor in the robot arm. The force torque sensor quirements and its relationship with the rest of the system. The dynamic
enables the robot to move back; or stop its movement under a simulation enabled quantitative assessment and business vale proposi­
predefined force (10N in the case example) tion of the proposed solution. With advances in information and
(b) Cycle time: when the robot is waiting for the operator to activate communication technologies, the DT can continuously be evolved from
it to perform the tasks. There are two switches (one for each as­ manual data syncing to automated and real-time data syncing thus
sembly fixture) available for the operator to activate the robot for enabling greater usefulness at system level.
each task When it comes to operational phase the usefulness of DT can be in the

10
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 10. Assembly tasks in the logic driven simulation.

Fig. 11. (a) Cycle time estimation; (b) Blue boxes denote areas where the robot can be placed given the reach location while red areas are unavailable locations; (c)
The grasp envelop of the human operator showing the areas where human can safely reach without bending the body.

11
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 12. Robot swept volume.

Fig. 13. Online optimization of robot trajectories.

form of (a) dynamic task distribution by task complexity rating and (continued )
event-driven simulation; (b) Intuitive robot programming to avoid tools and techniques. The
manual programming efforts; (c) human-safety assessments; (d) gener­ integration of a dynamic
ating data logs for critical actions by integration of sensors; (e) digital human model makes
it easier to generate adaptive
embedding artificial intelligence to get the system self-learned and make human-robot trajectories.
decisions according to its past experiences. Dynamic task As the task allocation Comparison of dynamic
Achieving a live connection between the DT and the physical system allocation problem arises when both and static task allocation
can make the DT to act in the real world. However, sensor input delays the human and the robot can Integration of artificial
accomplish a task. By intelligence for automatic
and action delays can cause problems. To ensure safety through DT in
defining the tasks’ rating, it task allocation
real-time, the adaptability of the robot in fraction of a second is needed becomes possible for the Automation of tasks with
to adapt to human movements. simulation to allocate the higher complexity
The following table summarizes the key learnings from this HRC in tasks between human and
practice: robot depending upon their
availability and ability.
Challenge Solution and success Future research Robot control The robot program is one big Automated control
factors opportunities examples program advantage of using digital program
Layout optimization Layout optimization of a Integration of new (AI) twins. The program is Application with new
with digital twin physical HRC system tools communicable in both standards like 5G
through a digital twin is Adaptivity through real-
achievable with available life tests (continued on next page)

(continued on next column)

12
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 14. Physical twin developed in the case.

Fig. 15. Utilization of the robot and the human operator.

(continued ) (continued )
directions between digital Remote control of the able to communicate adaption
and physical twin. robot system intuitively with the robot e. Speechless communication
Online collision By having an online linking Predictive systems control g. hand gestures, and smart
optimization between physical robot in an Big data analytics for faster watch.
empty space and a virtual optimization Enabling flexibility The digital twin can be used Teams of robots and
robot populated with virtual and high-mix low- for assembly process humans
equipment in a virtual volume production balancing. Since the key Integration of low-volume
environment helped to make locations are saved in the technologies such as 3D
safe estimations of any DT, for any change in task printing
possible collisions during execution, the robot needs to Development of machine
operation adapt to a new path and learning tools
Human-robot Several forms of human- Test of new interfaces such reach the defined locations.
interface robot interfaces can be as smart glasses, and smart Human ergonomic Digital human models can be Digital human models in
developed using a DT. The watch analysis used to perform human tasks different cultural contexts
purpose is to make human Automated language and evaluate the Ethical concerns
(continued on next column) (continued on next page)

13
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

Fig. 16. Data logging from the cobot.

Fig. 17. Stochastic analysis of the HRC assembly before and after experimentation and optimization.

(continued ) (continued )
biomechanical loads, fatigue Automated ergonomic interesting area to be workplace based on
and stress levels. This is an adaption to the human explored further in relation individual characteristics
(continued on next column) to a digital twin

14
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

[6] Malik A.A., Masood T., Bilberg A.. Virtual reality in manufacturing: immersive and
collaborative artificial-reality in design of human-robot workspace. International
Journal of Computer Integrated Manufacturing 2019:1–16.
As always, this kind of research has also several limitations. How­ [7] T. Masood, R. Weston, An integrated modelling approach in support of next
ever, these limitations also offer several insights for further research, generation reconfigurable manufacturing systems, International Journal of
which is outlined in the following. Computer Aided Engineering and Technology 3 (2011) 372–398.
[8] A.A. Malik, T. Masood, R. Kousar, Re-purposing factories with robotics in the face
First, this research was done in a single case study setup. The case of COVID-19, Science Robotics (2020) 5.
was chosen based on several criteria; however, more cases are needed. [9] R. Müller, M. Vette, A. Geenen, Skill-based Dynamic Task Allocation in Human-
For this, we call for more research on use cases in the combination with Robot-Cooperation with the Example of Welding Application, Procedia
Manufacturing 11 (2017) 13–21.
man, machine and work in a digital-twin setup. Also, cultural influences [10] R. Meziane, M.J.-D. Otis, H. Ezzaidi, Human-robot collaboration while sharing
might also have an impact on HRC combinations, which were not production activities in dynamic environment: SPADER system, Robotics and
explicitly considered in this setup. For instance, mimics and gestures Computer-Integrated Manufacturing 48 (2017) 243–253.
[11] A.A. Malik, T. Masood, R. Kousar, Reconfiguring and ramping-up ventilator
have a very different meaning in different cultural environments. If this production in the face of COVID-19: can robots help? Journal of Manufacturing
is considered, HRC can be even more efficient. So, similar studies in Systems (2020) https://doi.org/10.1016/j.jmsy.2020.09.008.
other countries, e.g. in Asia, are encouraged. Finally, also more longi­ [12] L. Gualtieri, E. Rauch, R. Vidoni, Emerging research fields in safety and ergonomics
in industrial collaborative robotics: A systematic literature review, Robotics and
tudinal studies would offer interesting insights in the dynamics of HRC
Computer-Integrated Manufacturing 67 (2020), 101998.
in a digital twin setup. [13] R. Müller, M. Vette, M. Scholer, Robot Workmate: A Trustworthy Coworker for the
Continuous Automotive Assembly Line and its Implementation, Procedia CIRP 44
(2016) 263–268.
6. Conclusion
[14] Salunkhe O., Stensӧta O., Åkerman M., Berglund Å.F., Alveflo P.-A.. Assembly 4.0:
Wheel Hub Nut Assembly Using a Cobot n.d.
Cobots are new choices to enable HRC systems offering human [15] S. Choi, K. Jung, S. Do Noh, Virtual reality applications in manufacturing
flexibility and machine endurance in a combination. Robots are getting industries: Past research, present findings, and future directions, Concurrent
Engineering 23 (2015) 40–63.
closer to the humans and the boundaries between human and robot [16] J. Cheng, H. Zhang, F. Tao, C.-F. Juang, DT-II: Digital twin enhanced Industrial
workspace are getting blurred.The usefulness of an HRC assembly sys­ Internet reference framework towards smart manufacturing, Robotics and
tem is based on its promise of being agile, adaptable and safe for its Computer-Integrated Manufacturing 62 (2020), 101881.
[17] R. Stark, C. Fresemann, K. Lindow, Development and operation of Digital Twins for
human members. The challenges and possible solutions of human-robot technical systems and services, CIRP Annals 68 (2019) 129–132.
interaction in production environments are not new however a seamless [18] K.J. Schlager, Systems engineering-key to modern development, IRE Transactions
integration of an HRC system requires that the system is designed, on Engineering Management (1956) 64–66.
[19] D. Snowden, Complex acts of knowing: paradox and descriptive self-awareness,
validated and controlled with innovative approaches. The idea of a Journal of Knowledge Management 6 (2002) 100–111.
digital twin as a ‘front runner’ can lead the optimized behavior of a [20] M. Grieves, J. Vickers, Digital twin: Mitigating unpredictable, undesirable
physical system by creating its time dependent accurate virtual model emergent behavior in complex systems. Transdisciplinary Perspectives on Complex
Systems, Springer, 2017, pp. 85–113.
and simulating it. With each change introduced in production parame­
[21] C. Perrow, Normal accidents: Living with high risk technologies-Updated edition,
ters of the physical system, new variables are simulated to predict cor­ Princeton University Press, 2011.
responding future behavior and highlight required optimizations. The [22] Z. Bi, M. Luo, Z. Miao, B. Zhang, W. Zhang, L. Wang, Safety assurance mechanisms
of collaborative robotic systems in manufacturing, Robotics and Computer-
behavior can be visualized and assessed without the risk of any financial
Integrated Manufacturing n.d. 67 (2020), 102022.
loss or human injury that may otherwise be present in the real [23] Y. Lu, C. Liu, I. Kevin, K. Wang, H. Huang, X. Xu, Digital Twin-driven smart
production. manufacturing: Connotation, reference model, applications and research issues,
Robotics and Computer-Integrated Manufacturing 61 (2020), 101837.
[24] B. Schleich, N. Anwer, L. Mathieu, S. Wartzack, Shaping the digital twin for design
CRediT authorship contribution statement and production engineering, CIRP Annals 66 (2017) 141–144.
[25] F.-J. Kahlen, S. Flumerfelt, A. Alves, Transdisciplinary Perspectives on Complex
Ali Ahmad Malik: Conceptualization, Methodology, Validation, Systems: New Findings and Approaches, Springer, 2016.
[26] M. Grieves, J. Vickers, Digital twin: Mitigating unpredictable, undesirable
Writing - original draft, Visualization. Alexander Brem: Resources, emergent behavior in complex systems. Transdisciplinary perspectives on complex
Supervision, Writing - review & editing. systems, Springer, 2017, pp. 85–113.
[27] P. Hehenberger, D. Bradley, Mechatronic Futures: Challenges and Solutions for
Mechatronic Systems and their Designers, Springer, 2016.
Declaration of Competing Interest [28] T. Masood, J. Egger, Augmented reality in support of Industry
4.0—Implementation challenges and success factors, Robotics and Computer-
The authors declare that they have no known competing financial Integrated Manufacturing 58 (2019) 181–195.
[29] Q. Qi, F. Tao, T. Hu, N. Anwer, A. Liu, Y. Wei, et al., Enabling technologies and
interests or personal relationships that could have appeared to influence tools for digital twin, Journal of Manufacturing Systems (2019).
the work reported in this paper. [30] J. Wilhelm, T. Beinke, M. Freitag, Improving Human-Machine Interaction with a
Digital Twin, in: International Conference on Dynamics in Logistics, 2020,
pp. 527–540.
Supplementary materials [31] E.J. Tuegel, A.R. Ingraffea, T.G. Eason, S.M. Spottswood, Reengineering aircraft
structural life prediction using a digital twin, International Journal of Aerospace
Supplementary material associated with this article can be found, in Engineering (2011) 2011.
[32] B. Naticchia, L. Messi, M. Pirani, A. Bonci, A. Carbonari, L. Tolve, Holonic System
the online version, at doi:10.1016/j.rcim.2020.102092. for Real-Time Emergency Management in Buildings, in: ISARC. Proceedings of the
International Symposium on Automation and Robotics in Construction 36, 2019,
References pp. 453–460.
[33] Y. Liu, L. Zhang, Y. Yang, L. Zhou, L. Ren, F. Wang, et al., A novel cloud-based
framework for the elderly healthcare services using digital twin, IEEE Access 7
[1] Rauschnabel P.A., Brem A., Ro Y.. Augmented Reality Smart Glasses: Definition,
(2019) 49088–49101.
Conceptual Insights, and Managerial Importance. 2015.
[34] C. Verdouw, J. Kruize, Digital twins in farm management: illustrations from the
[2] Mourtzis D. Simulation in the design and operation of manufacturing systems: state
FIWARE accelerators SmartAgriFood and Fractals, in: 7th Asian-Australasian
of the art and new trends. International Journal of Production Research 2019:1–23.
Conference on Precision Agriculture Digital, 2017, pp. 1–5.
[3] A. Bilberg, A.A. Malik, Digital twin driven human-robot collaborative assembly,
[35] Tao F., Cheng J., Qi Q., Zhang M., Zhang H., Sui F.. Digital twin-driven product
CIRP Annals - Manufacturing Technology 68 (2019) 499–502.
design, manufacturing and service with big data. The International Journal of
[4] A. Brem, P. Wolfram, Research and development from the bottom up-introduction
Advanced Manufacturing Technology 2017:1–14.
of terminologies for new product development in emerging markets, Journal of
[36] F. Tao, M. Zhang, Digital Twin Shop-Floor: A New Shop-Floor Paradigm Towards
Innovation and Entrepreneurship 3 (2014) 9.
Smart Manufacturing, IEEE Access 5 (2017) 20418–20427.
[5] D. Romero, T. Wuest, J. Stahre, D. Gorecky, Social factory architecture: social
[37] J. Guo, N. Zhao, L. Sun, S. Zhang, Modular based flexible digital twin for factory
networking services and production scenarios through the social internet of things,
design, Journal of Ambient Intelligence and Humanized Computing 10 (2019)
services and people for the social operator 4.0, in: IFIP international conference on
1189–1200.
advances in production management systems, 2017, pp. 265–273.

15
A.A. Malik and A. Brem Robotics and Computer-Integrated Manufacturing 68 (2021) 102092

[38] M. Fera, A. Greco, M. Caterino, S. Gerbino, F. Caputo, R. Macchiaroli, et al., [47] D. Chavarria-Barrientos, L.E. Villagomez, J. Miranda, A. Molina, R. Batres,
Towards Digital Twin Implementation for Assessing Production Line Performance I. Gutierrez, et al., A Methodology to Support Manufacturing System Design Using
and Balancing, Sensors 20 (2020) 97. Digital Models and Simulations: An Automotive Supplier Case Study, IFAC-
[39] L. Wang, R. Gao, J. Váncza, J. Krüger, X. Wang, S. Makris, et al., Symbiotic human- PapersOnLine 51 (2018) 1598–1603.
robot collaborative assembly, CIRP Annals - Manufacturing Technology 68 (2019) [48] Gordon C.C., Blackwell C.L., Bradtmiller B., Parham J.L., Barrientos P., Paquette S.
701–726. P., et al. 2012 anthropometric survey of US army personnel: methods and summary
[40] M.W. Grieves, et al., Virtually Intelligent Product Systems: Digital and Physical statistics. 2014.
Twins, in: S. Flumerfelt, et al. (Eds.), Complex Systems Engineering: Theory and [49] M. Peruzzini, M. Pellicciari, M. Gadaleta, A comparative study on computer-
Practice 256, 2019, pp. 175–200. integrated set-ups to design human-centred manufacturing systems, Robotics and
[41] Grieves M.. Digital twin: manufacturing excellence through virtual factory Computer-Integrated Manufacturing 55 (2019) 265–278.
replication. White Paper 2014. [50] D.B. Chaffin, Digital human modeling for workspace design, Reviews of Human
[42] J. Kovsturiak, M. Gregor, Simulation in production system life cycle, Computers in Factors and Ergonomics 4 (2008) 41–74.
Industry 38 (1999) 159–172. [51] F. Ranz, V. Hummel, W. Sihn, Capability-based task allocation in human-robot
[43] D. Kibira, D. Kibira, S. Kumaraguru, K. Morris, Methods and tools for performance collaboration, Procedia Manufacturing 9 (2017) 182–189.
assurance of smart manufacturing systems, US Department of Commerce, National [52] S. Samy, H. ElMaraghy, A model for measuring products assembly complexity,
Institute of Standards and Technology, 2015. International Journal of Computer Integrated Manufacturing 23 (2010)
[44] A.A. Malik, A. Bilberg, Framework to implement collaborative robots in manual 1015–1027.
assembly: a lean automation approach, in: B. Katalinic (Ed.), Proceedings of the [53] G.F. Rossano, C. Martinez, M. Hedelind, S. Murphy, T.A. Fuhlbrigge, Easy robot
28th DAAAM International Symposium, DAAAM International, 2017, path programming concepts: An industrial perspective on path creation, in:
pp. 1726–9679. Published by ISSN. Robotics (ISR), 2013 44th International Symposium on, 2013, pp. 1–6.
[45] T. Lechler, E. Fischer, M. Metzner, A. Mayr, J. Franke, Virtual Commissioning- [54] A.A. Malik, A. Bilberg, Complexity-based task allocation in human-robot
Scientific review and exploratory use cases in advanced production systems, in: collaborative assembly, Industrial Robot: The International Journal of Robotics
26th CIRP Design Conference 81, 2019, pp. 1125–1130. Research and Application 46 (2019) 471–480.
[46] C.G. Lee, S.C. Park, Survey on the virtual commissioning of manufacturing systems, [55] S. Denmark, Population in Denmark, Population and Elections (2017).
Journal of Computational Design and Engineering 1 (2014) 213–222.

16