Computers in Industry 133 (2021) 103508

Contents lists available at ScienceDirect

Computers in Industry
journal homepage: www.elsevier.com/locate/compind

Autonomous, context-aware, adaptive Digital Twins—State of the art

and roadmap
Karl Hribernik a , Giacomo Cabri b , Federica Mandreoli b,∗ , Gregoris Mentzas c
a
BIBA - Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Hochschulring 20, 28359 Bremen, Germany
b
Dipartimento di Scienze Fisiche Informatiche e Matematiche Universita’ di Modena e Reggio Emilia, Italy
c
Information Management Unit (IMU), Institute of Communication and Computer Systems (ICCS), National Technical University of Athens (NTUA), Greece

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

Article history: Digital Twins are an important concept in the comprehensive digital representation of manufacturing
Received 20 January 2021 assets, products, and other resources, comprising their design and configuration, state, and behaviour.
Received in revised form 21 June 2021 Digital Twins provide information about and services based on their physical counterpart’s current condi-
Accepted 28 June 2021
tion, history and predicted future. They are the building blocks of a vision of future Digital Factories where
Available online 13 October 2021
stakeholders collaborate via the information Digital Twins provide about physical assets in the factory
and throughout the product lifecycle. Digital Twins may also contribute to more flexible and resilient Dig-
Keywords:
ital Factories. To achieve this, Digital Twins will need to evolve from today’s expert-centric tools towards
Digital Twins
Digital Factories
active entities which extend the capabilities of their physical counterparts. Required features include
Context-awareness sensing and processing their environment and situation, pro-actively communicating with each other,
Autonomy taking decisions towards their own or cooperative goals, and adapting themselves and their physical
Adaptivity counterparts to achieve those goals. Future Digital Twins will need to be context-aware, autonomous,
and adaptive. This paper aims to establish a roadmap for this evolution. It sets the scene by proposing a
working definition of Digital Twins and examines the state-of-the-art in the three topics in their relation
to DTs. It then elaborates potentials for each topic mapped against the working definition, to finally iden-
tify research gaps allowing for the definition of a roadmap towards the full realisation of autonomous,
context-aware, adaptive Digital Twins as building blocks of tomorrow’s Digital Factories.
© 2021 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. A working definition of Digital Twins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1. System state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. System design and configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3. System behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. State-of-the-art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.1. Context-Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2. Autonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3. Adaptivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4. Research and development gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

∗ Corresponding author.
E-mail addresses: [email protected] (K. Hribernik), [email protected] (G. Cabri), [email protected] (F. Mandreoli), [email protected]
(G. Mentzas).

https://doi.org/10.1016/j.compind.2021.103508
0166-3615/© 2021 Elsevier B.V. All rights reserved.
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

1. Introduction tus for research toward the implementation of context-aware,

autonomous, and adaptive DTs. Since the DT concept is still evolv-
Manufacturing companies are today confronted with increas- ing and is used to refer to a broad spectrum of approaches, the
ingly global and dynamic markets. They consequently need to next section proposes a working definition of DTs. Section 3 inves-
reshape their strategies to meet the 21 st Century’s challenges tigates the state-of-the-art of each of the topics context-awareness,
concerning competitiveness, productivity, and sustainability. autonomy and adaptivity. Since DT literature has hitherto only
Examples of these challenges include shorter product lifecycles, marginally been concerned with these topics, the state-of-the-art
ever more demanding customer requirements especially towards presented here includes work relevant to expanding the concept
quality and sustainability, the trend towards mass customisation, towards these capabilities. The fourth section looks at the poten-
as well as the necessity to establish new revenue streams via the tials for context-awareness, autonomy, and adaptability for DTs.
servitization of physical products. To meet these challenges, man- The firth section attempts to pinpoint research gaps which need to
ufacturing, supply chain management and logistics increasingly be addressed and proposes a roadmap. The last section draws con-
depend on interactions between all stakeholders in the product clusions from the above and highlights noteworthy trends in the
lifecycle. The heterogeneity of the resources involved, and the future development of context-aware, autonomous, and adaptive
autonomy of the actors bind resources and demand a high level DTs.
of coordination from all players, interoperating seamlessly with
different stakeholders, processes, products, and resources. In addi-
tion, demands for more flexible systems that can quickly react to 2. A working definition of Digital Twins
faults and changes throughout the product lifecycle mean that sys-
tems will need to adapt quickly and appropriately within these Grieves made the first recorded use of the term DT in 2003 in a
complex interaction environments. For example, manufacturing course on Product Lifecycle Management (PLM) (CoBuilder, 2018).
systems need to become more adaptable to react to fluctuations A DT was originally merely a digital representation of a physical
in demands, changes to product configurations or to the supply object as a CAD model (Grieves and Vickers, 2016). The concept
chain, and disturbances in production. IT systems in product devel- was expanded over time to mean an actionable simulation includ-
opment need to be able to easily incorporate information from ing not only the form of an object but also its behaviour (Grieves
stakeholders in the middle-of-life of the product lifecycle, such and Vickers, 2016), and was adopted by the aerospace industry
as usage behaviour, maintenance, and information from product- to describe mirroring information of space vehicles (Glaessgen
service systems. Reverse logistics systems can contribute to better and Stargel, 2012) as “twin” models (Rosen et al., 2015). Here,
sustainability by intelligently processing information from previ- DTs integrate sub-models of different systems and environmen-
ous lifecycle phases. tal interactions of a vehicle by combining stochastics, on-board
Digital Twins (DTs) and Digital Factories (DFs) are two concepts sensor data and historical data sets to mirror its operational phase
which promise to help address the challenges and requirements for health management, prognostics and diagnostics, used to esti-
outlined above. DTs are comprehensive, actionable, model-based mate remaining useful life or mission success rates (Glaessgen and
digital representations not only of the form of a product, system, Stargel, 2012; Reifsnider and Majumdar, 2013). There have also
or resource but also of its behaviour (Rosen et al., 2015), which been efforts to integrate physics-based models with DTs. Physics-
can include simulation or prediction. DTs can be used to plan, opti- based models offer a high degree of interpretability, reliability, and
mise, and simulate operations without the need for the physical predictive capability and are commonplace throughout engineer-
entity. As realistic digital representations, DTs can serve as com- ing. The integration of such models with DTs, that provide digital
prehensive sources of information about their counterparts. They models tailored to each unique physical asset, addresses the main
can be considered the building blocks of a vision of future DFs limitation of physics-based models, i.e., the assumption that the
in which product lifecycle stakeholders collaborate via the infor- model used is an accurate reflection of the underlying physical
mation provided by DTs. A DF built upon DTs viewed in this way system. Such physics-aware, or hybrid DTs (Chinesta et al., 2020)
can extend beyond company boundaries and offer the opportunity are quite useful in engineering applications in areas like aviation
for the business and its suppliers to collaborate on processes that (Kapteyn et al., 2020), smart grid (Tzanis et al., 2020) and battery
affect the entire supply chain (Bicocchi et al., 2019). By interfacing management (Sancarlos et al., 2021). The concept was transferred
DTs with each other, they can contribute to a seamless interoper- to production systems and smart manufacturing in 2013 (Lee et al.,
ation between products, processes, and resources. DTs also hold 2013; Lu et al., 2020). Subsequent authors extended the application
promise to contribute to more flexible and resilient DFs. To achieve area to products in general and other life cycle phases, most promi-
this, the DT concept will need to evolve beyond today’s expert- nently for design and production processes (Schleich et al., 2017).
centric tools for planning, optimisation and simulation activities Comparable concepts such as Product Avatars were developed in
(Hartmann et al., 2018) to encompass features which allow them parallel (Hribernik et al., 2005; Ríos et al., 2015).
to sense and process their environment and situation, pro-actively Usage of the term is inconsistent, for example regarding simula-
exchange information with each other, take decisions towards their tion. Whilst some authors equate DTs with simulations of systems
own or shared goals, and be able to adapt themselves and their (Glaessgen and Stargel, 2012; Sha to et al., 2012; Gabor et al., 2016),
physical counterparts to achieve those goals. That means, future others view them as models to be used as a basis for different
DTs should be context-aware, autonomous, and adaptive. kinds of simulations (Reifsnider and Majumdar, 2013; Tuegel, 2012;
As digital representations of physical assets, DTs are related to Gockel et al., 1813; Hochhalter et al., 2014). This is exacerbated
concepts such as agents, holons, cyber-physical systems (CPS) and by the lack of comprehensive and general reference models (Lu
product avatars. Nevertheless, there are few proposals suggesting et al., 2020; Kritzinger et al., 2018)), although models of varying
DTs may communicate with each other or their environment, or degrees of detail exist for specific use cases. RAMI4.0 has been
cooperate towards common goals as distributed, autonomous enti- shown to be a useful pattern for defining DTs in manufacturing.
ties. Furthermore, not all sources explicitly mandate a direct control In this context, DTs are increasingly being considered a part of
loop between DTs and their physical counterparts (Cimino et al., Cyber-physical System (CPS) architectures, realising twin models
2019). of assets and machines (Josifovska et al., 2019), the computational
This paper intends to contribute to the evolution of DTs by modules of the physical components of CPS (Alam and El Saddik,
investigating their potentials in tomorrow’s DF and give impe- 2017), or the RAMI4.0 asset administration shell (Anderl et al.,

2
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

Fig. 1. Schema of a Working Definition of Digital Twins.

(* Source: (Cimino et al., 2019)).

2.1. System state

One unique characteristic of DTs is the (real-time) reflection of

an active product’s, system’s, or product-service system’s (Stark
and Damerau, 2019) state based on sensor data (Bicocchi et al.,
2018; Glaessgen and Stargel, 2012; Reifsnider and Majumdar, 2013;
Lee et al., 2013; Lu et al., 2020), information from IT systems
(Glaessgen and Stargel, 2012; Reifsnider and Majumdar, 2013;
Lee et al., 2013), execution systems (Bicocchi et al., 2018), or
other sources of usage information, like maintenance histories
(Glaessgen and Stargel, 2012; Reifsnider and Majumdar, 2013). The
system state is structured by ultra-realistic, high-fidelity models
(see below) and thus requires a high level of digitalisation. DTs are
also mentioned in literature which do not yet have a physical coun-
terpart (e.g. a product under design), or their physical counterpart is
incomplete (e.g. in manufacture or assembly) (Boschert and Rosen,
2016; Qi and Tao, 2018). In these cases, the system state component
may contain data, for example, from test or manufacturing systems
(Lu et al., 2020), simulation or data analysis (Qi and Tao, 2018). This
way, the DT can already be used to simulate or predict the behaviour
of a system before it is built, similarly to Virtual Prototypes (Madni
et al., 2019). DTs which implement a control loop by reflecting the
Fig. 2. Relation of the properties adaptivity, autonomy and context-awareness.
virtual system state back onto the physical counterpart are also
described in literature (Bicocchi et al., 2018; El Saddik, 2018), and
can even be considered to be a characteristic which distinguishes
DTs from Digital Shadows (Cimino et al., 2019).
2018; Tantik and Anderl, 2017). Here, specific challenges towards
information, model, system and tool integration have not yet been
fully addressed (Koulamas and Kalogeras, 2018). STEP (Standard
for the Representation of Product Model Data, ISO 10303 Product 2.2. System design and configuration
Data Representation and Exchange) has also been suggested as a
basis for the specification of DTs (Anderl et al., 2018). Whilst these The System Design and Configuration part comprises mainly
works are solid groundwork towards a reference model integrated static information and models about the system, ranging from sim-
into the Industrie 4.0 context, more work is required before the role ple product models, over detailed design models such as CAx or
of DTs in smart manufacturing is fully understood and defined. STEP supporting the dimensioning and design of parts, bills of mate-
Due to these inconsistencies, the following proposes a work- rial, to configuration and layout data and other parameters (Rosen
ing definition of core characteristics of DTs common to most et al., 2015; CoBuilder, 2018; Lee et al., 2013; Sha to et al., 2012;
sources and aligned with the CIRP encyclopaedia definition (Stark Anderl et al., 2018; Stark and Damerau, 2019; Guo et al., 2019).
and Damerau, 2019) and informed by (Cimino et al., 2019). DTs This part can also consist of static process models where DTs of pro-
as understood here are comprehensive digital representations of cesses or systems are concerned (Bao et al., 2019). A DT is different
physical assets, comprising their design and configuration, state, from conventional design models in this dimension in that it rep-
and behaviour, as shown in the schema in Fig. 1 below. The three resents a specific instance of a product and a series (Madni et al.,
core components system design and configuration, system state, 2019). It should always reflect the characteristics of its physical
and system behaviour, are elaborated in the following subsections. counterpart. Consequently, data-driven parameter calibration and
This paper focuses its investigation on the three core components of model validation are necessary and design and configuration mod-
this working definition, considering DT services to be applications els need to evolve dynamically over the lifecycle of the counterpart
built upon this core. (Lu et al., 2019).

3
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

2.3. System behaviour between external and internal context (Alexopoulos et al., 2016).
The former relates to places, people and their roles, and things.
DTs include integrated “ultra-realistic, high-fidelity” (Glaessgen The latter describes goals, tasks, work contexts, business processes,
and Stargel, 2012; Sha to et al., 2012) models describing their phys- communication, and even emotional and physical states.
ical counterpart’s behaviour (Rosen et al., 2015; Glaessgen and Context modelling is used to organise context information,
Stargel, 2012; Reifsnider and Majumdar, 2013; Lee et al., 2013). allowing, for example, to define context types, relationships, and
To illustrate, predicting the remaining useful life of an aircraft dependencies between context information, and to control qual-
requires detailed modelling of its response to aerodynamic, high- ity and consistency. Context models also facilitate reasoning on
frequency sonic and dynamic loads, and thermal fluxes (Tuegel and context information. Different approaches can be taken to context
Ingraffea, 2011). In addition, simulation needs to model e.g., the modelling, including graphical and object-oriented, logic-based
history of the structure’s response to forces to analyse potential and semantic modelling (Lee and Martinez Lastra, 2013).
damage. This necessitates integrating many models: e.g. multi- Different technologies can be exploited to generate and com-
physics and damage modelling, models for dealing with aleatory municate context information in manufacturing environments.
and epistemic uncertainty derived from interval theory, and high- Context information describing environmental parameters can
resolution structural analysis (Tuegel and Ingraffea, 2011). Besides be sourced, for example, from sensors and embedded systems
multi-physics and numerical modelling (e.g. 3D modelling) and (Alexopoulos et al., 2016; Azouz and Pierreval, 2019), RFID
simulators (e.g. finite element methods), data driven analytics is (Alexopoulos et al., 2016), BLE (Bisio et al., 2018), NFC (Alexopoulos
becoming increasingly relevant (Tao et al., 2019a). This assumes et al., 2016), IoT (Alexopoulos et al., 2016; Bisio et al., 2018; Lee et al.,
that since data is a manifestation of both known and unknown 2016) and CPS (Alexopoulos et al., 2016). Context can also be more
physical factors, a data-driven model can account for a better vir- broadly understood to include not only the environment of sensors
tual model in a non-parametric setting and to anticipate events. and embedded devices, but also the users and their preferences
Various machine learning and artificial intelligence techniques are (Bisio et al., 2018). In manufacturing, context-aware applications
useful for the development of data-driven models, while a recent may also utilise information e.g. about the planning and execution
effort tries to consistently integrate the various approaches in the of manufacturing processes, which can be provided by IT systems
overall vision of DTs (Rasheed et al., 2019). such as manufacturing execution (MES) or enterprise resource
planning (ERP) systems (Alexopoulos et al., 2016). This may also
include a history of events related to products and resources (Azouz
3. State-of-the-art
and Pierreval, 2019). Since the system state component of a DT
reflects the status of a physical object as monitored e.g., by sensors,
Currently, no literature specifically addresses context-
IoT devices and other sources, DTs can act as providers of context
awareness, autonomy and adaptivity together in relation to
information to other DTs.
DTs. Literature specifically addressing the application of the
Context-awareness can contribute to the “smartness” of facto-
individual properties to DTs is also lacking. This paper conse-
ries and manufacturing environments (Radziwon et al., 2014). By
quently focuses on examining the three properties to identify their
bridging the gap between the virtual and physical worlds using IoT,
potentials for DTs. These three properties are often discussed in
objects can be made recognizable and smart, and context-related
conjunction with one another, as illustrated in Fig. 2. For example,
decisions can be supported in a smart factory environment (Bisio
context-awareness and adaptivity intersect where a system’s
et al., 2018). Examples of application areas in smart manufacturing
behaviour is modified based on information about its logical,
include energy management, failure management, manufacturing
physical, and/or external context. Likewise, autonomous systems
cycle time optimisation, shop-floor management, and manufac-
or entities may use context-awareness to take decisions based on
turing system reconfiguration. Context-aware energy management
information about their environment or situation. Autonomous
enables users to monitor and control their energy consumption in
entities can show adaptive behaviour when their decisions lead
a smart way, so they can control their utilities and equipment and
to the modification of the entity’s behaviour. This paper focuses
adapt their behaviour in order to avoid energy waste (Lee et al.,
on applying all three properties to DTs: DTs which are aware
2016). In mass production with high volumes and low profit mar-
of their own context, capable of autonomous decision-making
gins, context awareness can help optimize manufacturing cycle
and adapting themselves, their physical counterparts, and by
times. It has been shown to contribute to managing high product
extension, the systems within which they operate.
variety and batch size variation when handling many parts, e.g. in
white goods manufacturing (Alexopoulos et al., 2016). In failure
3.1. Context-Awareness management, context-awareness allows interventions to be more
effective because they can be focused on what happened and in
Context-awareness is the capability of customizing the which environment (Wieland et al., 2010). Shop-floor management
behaviour of a system or of a component depending on the con- can be improved by context-awareness by using sensors to expose
text in which it executes to make it more effective or relevant to processes in a clearer way, providing the right information, to the
the situation in which it is being used (Lucke et al., 2008). Context is right people, at the right time (Alexopoulos et al., 2016), support-
any information that can be used to characterise the situation of an ing worker’s decisions by providing useful information in the right
entity (Alexopoulos et al., 2016), such as a person, place, or physical context (Lee and Martinez Lastra, 2013).
or computational object (Lee and Martinez Lastra, 2013). A context- Making DTs context-aware will contribute to a more holistic
aware system can access and process information which describes representation of their physical counterparts including their environ-
context. What constitutes context information can been classified ment and execution context (Borangiu et al., 2019), which is helpful
in different ways. A prominent classification distinguishes pri- for e.g. HRC applications, where a robot’s position needs to be con-
mary (time, identity, location and activity) and secondary context sidered in relation to humans. Context-awareness in DT literature
which can be derived from the primary ones (Alexopoulos et al., is found mainly in CPS-related research relating to the system state
2016). These basic classes are not always appropriate for industrial component. An early approach to context-aware DTs in CPS includes
applications, which can benefit from a more specific classification context information in the representation of the system state of a
such as user, environment and production contexts (Rosenberger production plant, via a separate context-aware application (Park
and Gerhard, 2018). Another helpful classification distinguishes et al., 2020). In this case, the DT itself is not fully context aware.

4
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

DTs can contribute to processing context information generated by 2009). This makes autonomy an interesting approach for systems
CPS, since CPS generally focus on the physical layer of embed- about which little is known in advance.
ded systems (Leng et al., 2019). They have also been proposed to Smaller organisational units need to receive and process infor-
improve the identification of relevant contexts in smart manufactur- mation about their environment to take their own decisions
ing systems (Preuveneers et al., 2018). DTs should also act as context (Scholz-Reiter et al., 2021; Scholz-Reiter, 2004). Realising auton-
information sources for other networked DTs, allowing them to pro- omy thus requires linking the physical to the virtual world hosting
cess information about their environment, an important ingredient IT capable of autonomous decision-making. Technologies such as
for distributed, autonomous decision-making. DTs should also be smart or intelligent processes, IoT, CPS, and indeed DTs, support
able to access appropriate data sources based on context changes. this link (Maes, 1993).
A physical counterpart’s context may require changes to its Autonomous decision-making is often modelled on biological
configuration, affecting the DT’s system design component. For systems, for example, how humans or animals perform control
example, a machine tool might need to be configured to carry out tasks, recognise patterns and make decisions (Passino, 1995).
a specific task (changing moulds or tools). The DT should be able to Examples of these methods range from simple due date (Windt
understand the context to reflect this in its design models and configu- et al., 2010), queue-length estimation and rule-based approaches
ration appropriately. For example, a DT architecture for cloud-based (Gebhardt et al., 2011) through fuzzy control, fuzzy supervisory
cyber-physical systems supports context-aware self-reconfiguration control, fuzzy learning control, knowledge-based control, neural
of manufacturing systems to increase the resilience of manufactur- networks and genetic algorithms (Passino, 1995). Many methods
ing systems, applying fuzzy rules and Bayes networks to the system of decision-making in autonomous systems rely on the interac-
context (Alam and El Saddik, 2017). This contributes to production tion, cooperation or competition of individual units (Hülsmann and
system adaptivity. Windt, 2007). The dynamics of such an interaction “can lead to emer-
To realise context-awareness, DTs must incorporate context gent structure or emergent functionality” (Maes, 1993). Emergent
models into their system behaviour component, including models behaviour may be Predicted Desirable (PD), Unpredicted Desirable
of system tasks, processes, objectives, and operations. They must (UD), Predicted Undesirable (PU), or Unpredicted Undesirable (UU)
access knowledge about their context, or include models reflect- (Grieves and Vickers, 2016). Desirable behaviour is, for example, an
ing their operational context. Graph-based models can represent increase in the robustness of the overall system due to distributed
components involved in smart manufacturing processes along with and flexible handling of dynamics and complexity (Hülsmann and
their relations, constraints and paths (Zheng and Sivabalan, 2020). Windt, 2007). DTs can help manage emergent behaviour, guarantee
This allows the DT to apply context-aware cognition to its tasks. obtaining PD, eliminate PU, and decrease UU behaviour (Grieves
A context model which includes the states of all objects, e.g., and Vickers, 2016). Methods supporting this kind of decision-
machines, tools, production processes, material, in a production making include swarm intelligence approaches (Mullen et al., 2009)
facility has been integrated into DTs which constitute a “social fac- such as ant algorithms (Gebhardt et al., 2011) and bee foraging algo-
tory”, that is a smart factory which focusses on the interactions rithms (Windt et al., 2010), Distributed Logistics Routing Protocol
between people and the production environment (Kassner et al., (DLRP) (Scholz-Reiter et al., 2008), Link-state internet routing pro-
2017). tocol (LSIRP), asynchronous teams (Talukdar et al., 1998), holonic
manufacturing algorithms (Windt et al., 2010), or auction algo-
rithms for decentralized parallel machine scheduling (Attanasio
3.2. Autonomy et al., 2006).
Autonomous systems not only need to be able to react to
A dictionary definition of autonomy is “freedom from external changes in their environment, but also to act proactively with
control or influence; independence” (autonomy, 2021). More specifi- respect to their own purposes. A proactive and anticipatory sys-
cally, it is “the ability to act with some measure of independence, and to tem contains “a predictive model of itself and/or its environment,
assume responsibility for one’s own resources and behaviour” (Steiner, which allows it to change state at an instant in accord with the
2008). From a technical perspective, “[a]gent X is autonomous from model’s predictions pertaining to a later instant” (Nadin, 2017; Louie,
agent Y to the degree that X pursues its goals without input from Y” 2010). Proactive computing is currently understood as an evolution
(Müller, 2012). More autonomy implies less control (Müller, 2012), away from interactive computing, towards human-(un)supervised
qualifying “freedom from external control” above. Entities may have pervasive computing scenarios (Tennenhouse, 2000). Proactivity
varying degrees of autonomy and be subject to inverse degrees of extends reactivity, referred to in literature as sense-and-response
external control. (Elwany and Gebraeel, 2008) or detect-and-act (Tao et al., 2014) – to
Autonomy can help handle disturbances, changing market a new model of situational awareness with four phases: Detect, Pre-
demands or other unforeseen events by shifting decision-making dict, Decide, Act (Engel et al., 2012). “Detect” deals with monitoring
to smaller organizational units, which can contribute to increasing and detecting current indicators. “Predict” uses those indicators to
the system’s flexibility and adaptability to changes or disturbances forecast when the system will leave an admissible state if noth-
(Hülsmann and Grapp, 2005; Antsaklis et al., 2021). Resources may ing changes. “Decide” takes real-time decisions about how best
be autonomously reconfigured, restructured or rescheduled to con- to eliminate or mitigate the problem and remain in an admissible
tinue production with minimal downtime (Park and Tran, 2011). state. The “Act” deals with the actual implementation of that action.
Autonomy also helps tackle very unstructured or complex envi- Proactive event-driven architectures combine advanced event pro-
ronments which cannot be appropriately anticipated (Pachter and cessing with dynamic forecasting capabilities leveraged towards
Chandler, 1998) since shifting decision-making to smaller units online optimisation and decision-making (Lindberg, 1990). The
breaks the total complexity of the system down to partial complex- decisions are made in real time and require swift and immediate
ity (Hülsmann and Grapp, 2005). Decentralizing decision-making processing of big data, that is, extremely large amounts of noisy
allows for a problem decomposition which reduces computational data flooding in from various locations, as well as historical data
effort (Schuldt et al., 2011). In very complex systems, linear plan- (Bousdekis et al., 2015, 2017; Bousdekis et al., 2019; Petersen et al.,
ning and scheduling can be difficult, and optimization can take 2016).
considerable effort (Gebhardt et al., 2011). Autonomous systems DT characteristics lend themselves to implementing autonomy
alleviate these difficulties by reducing the need for a-priori plan- and are already considered core enablers of autonomous systems
ning and optimisation by local decision-making (Hülsmann et al., (Cronrath et al., 2019; Rassolkin et al., 2019). However, current lit-

5
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

erature describes DTs as expert-centric tools without autonomous and control of a manufacturing system are too fast, frequent,
capabilities (Hartmann et al., 2018), and their evolution towards complex, or unpredictable for human intervention at run-time,
active components which extend the characteristics of their phys- adaptivity based on MAS (multi-agent systems) can help man-
ical counterparts is not yet complete (Saracco, 2019; Dambrot, age industrial control systems with the required speed (Sanderson
2021). The system state component links the physical and virtual et al., 2015). Small-scale adaptivity realised by context-aware dis-
worlds, which is a basic requirement for realising autonomous enti- tributed, autonomous CPS which monitor themselves can also
ties. A full control loop is required for the DT to autonomously help ensure and optimize production quality. By storing informa-
control its physical counterpart. Here, principles of autonomous tion about quality and process execution in each product’s own
control, both reactively and proactively, may be applied. Agent, CPS-enabled memory, deviations can be detected directly in the
holon or CPS principles may inspire the implementation of DT production line and corrected accordingly (Bergweiler, 2016).
autonomy. First considerations towards the convergence of DT and In smart manufacturing, the detection of sensor or actuator out-
CPS, formalised in the RAMI 4.0 Asset Administration Shell, exem- liers that deviate from the expected parameters as modelled may
plify this (Wagner et al., 2017). either indicate a fault in the sensor/actuator or a change in the
With regards to system design, based on simulations and analy- process. Once a decision is made between these two options, mod-
sis in conjunction with real-time and historical data about physical els may need to be adapted to reflect the change. A rapid (e.g.
assets, DTs could autonomously take decisions about the product discarding historical data before the change) or gradual (e.g. expo-
design and configuration and influence future product design itera- nential forgetting) approach can be used to facilitate self-adaptive
tions (Lützenberger et al., 2016; Klein et al., 2019]. The configuration modelling to address this problem (He and Wang, 2018). Adaptiv-
of the physical counterpart can also be influenced by the DT in this ity can also contribute to dynamically organising supply chains in
way (Wuest et al., 2013). In predictive maintenance, the DT may multi-organisational collaborations and thus optimising produc-
take autonomous decisions with regards to maintenance, affecting tion (Radziwon et al., 2014), or dynamically on multiple levels for
the physical configuration of the asset (Hribernik et al., 2018). production logistics synchronisation, using Cloud Manufacturing
The advanced simulation and analytical capabilities offered in combination with IoT to dynamically synchronise logistics plans
by DTs’ system behaviour component, fed with high-resolution and collaborations (Qu et al., 2016).
real-time system state data and extended by context-awareness Changing systems to meet the needs of workers is another sig-
allows more accurate predictive capabilities, which can improve nificant area of application, and is fundamental to efficient and
autonomous decision-making in industrial scenarios (Rosen et al., effective smart manufacturing systems (Peruzzini and Pellicciari,
2015). Furthermore, context-awareness could allow DTs to 2017). For example, aging workers may have different profiles of
autonomously select, integrate, and execute appropriate physical, physical and cognitive limitations, restricting their ability to work
simulation, and analytical models. For example, different physical at standardised workplaces. A methodology has been proposed to
models are required for a spacecraft travelling in a planet’s atmo- self-adapt manufacturing systems to their needs (Peruzzini and
sphere or in space (orbital mechanics models). Pellicciari, 2017). User interfaces and information provision in man-
ufacturing can also be adapted to workers’ needs. Humans tend to
3.3. Adaptivity behave in nondeterministic, flexible ways and adapt themselves to
changing environments (Joo and Shin, 2019). This means that HMI
Adaptivity is the capability to modify a system or component’s (human–machine interaction) should be designed to provide them
behaviour depending on unpredicted situations to achieve one or with the appropriate level of information depending on different
more goals. It describes how well a machine or group of machines system states (de Visser and Parasuraman, 2011; Inagaki, 2003).
can adapt to change, which can be measured by comparing the An example is preventing mode confusion, where humans work
weighted flexibilities of two situations occurring at different points with mental models about a process that conflicts with those of
in time (Mandelbaum, 1978). Adaptivity is an objective of Industrie involved machines, or receive confusing feedback from their user
4.0 and Cloud Manufacturing (Qu et al., 2016) and a core charac- interfaces (Joo and Shin, 2019).
teristic of smart factories, defined as a “manufacturing solution that DT adaptivity may refer to the capability of the digital repre-
provides such flexible and adaptive production processes that will solve sentation, the physical counterpart, or both, to adapt. Regarding
problems arising on a production facility with dynamic and rapidly system state, literature highlights DTs’ potential to optimise or mod-
changing boundary conditions in a world of increasing complexity” ify their counterparts’ control parameters based on awareness of their
(Radziwon et al., 2014). current operation conditions (Tao et al., 2019a). For example, DTs
Adaptivity involves monitoring the state and/or environment using adaptive algorithms can change production system control con-
of the system or component whose behaviour is to be changed. figurations ensuring safe operations when faults occur (He et al.,
Relevant information can be sourced in real-time, e.g. from IoT, 2019). Context changes can trigger the DT to modify data sources
Industrial IoT systems or CPS (Qu et al., 2016; Bergweiler, 2016). for an optimal reflection of the physical asset. For example, changes
Suitable decision-making mechanisms are required to enact adap- to the stakeholder network of a Digital Factory may require the DT
tivity once a system state is detected that requires modification. to access new enterprise systems or disconnect no longer relevant
Examples include case-based reasoning (Peruzzini and Pellicciari, ones. The DT should thus autonomously recognise the need, discover
2017) and decentral decision-making based on autonomous enti- appropriate data sources, and (disc)connect those sources. Another
ties as described above (Gonçalves and Peschl, 2018; Sanderson interesting example is represented by robotic work cells. As the
et al., 2015). tool of iterative refinement is calibration, its effective and system-
Adaptivity has been applied to different manufacturing prob- atic execution has to be ensured (Erdős et al., 2020). Since data can
lems. For example, it can help manage increasing complexity and be collected continuously from the real work cell, relevant infor-
unpredictability. By monitoring and identifying machine status in mation can be stored historically. When the DT model parameters
real time, maintenance can be appropriately triggered and produc- are out of the feasible tolerance region surrounding the real mea-
tion rescheduled via dynamic scheduling mechanisms (Mourtzis surements on the physical system, they need to be adjusted. The
et al., 2014). Decentralized, autonomous approaches have been DT should thus autonomously recognize such a divergent situation
shown to enable the dynamic reconfiguration of manufacturing and automatically recalculate the parameter values.
systems, increasing flexibility, adaptability to change, and fault Relating to system design, DTs can be considered adaptive mod-
tolerance (Gonçalves and Peschl, 2018). Where decision-making els of complex physical systems (Rasheed et al., 2019). Context-aware

6
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

changes to the physical counterpart’s configuration and conse- Data Space reference architecture (Otto et al. (2018)), or Indus-
quently its DT may be required, requiring the DT to select, modify trial Ontologies Foundry (Kulvatunyou et al., 2018) are promising
and configure models appropriately in accordance with the situa- avenues of investigation. Underlying these interoperability issues
tion they represent whenever changes occur. Modifications can be is the need for work on standards and protocols for DT informa-
human-supervised or automatic. The changes can concern different tion exchange. The breadth of information, model and technology
aspects of the production cycle, e.g. products, services, stakehold- heterogeneity and the use of DTs in different sectors makes stan-
ers, and the general context. Here, a technical barrier is adapting dardization for DT interoperability extremely challenging.
models to avoid discrepancies over time, e.g. due to degradation. The next line deals with modelling. The model-based nature
Model updating schemes have been put forward to facilitate adap- of DTs means that context models need to be incorporated to use
tive models in DTs (Lu et al., 2019; Wang et al., 2019). context in combination with other models on the level of system
An adaptive physics-based performance model has been used in behaviour. This is a significant step towards enabling autonomous,
a DT of aero-engines to generate fault signatures that can be com- adaptive DTs. DTs capable of understanding and acting upon con-
pared with measured flight data to assess the engine conditions text will be able to identify, connect to and collaborate with other
(Zaccaria et al., 2018). DTs can contribute to the dynamic adapt- DTs sharing their context. Several DTs in a common manufacturing
ability of manufacturing execution systems, performing real-time context within a production line could join together to form an
optimization of products and production processes (Zhang et al., autonomous, adaptive virtual production line. Research towards
2017). Machine learning has been put forward as a way for DTs understanding how individual DTs may connect and collaborate
to adjust system behaviour with minimal need for supervision is required. Agents, holons and CPS are furthermore good exam-
(Madni et al., 2019). Context may influence the parameters of or ples of how autonomous decision-making can be modelled and
necessitate new physical, simulation and analytical models. On one implemented based on digital representations of physical entities
level, adaptation influences the physical counterpart’s configura- (Giret and Botti, 2004). Their relation to DTs has however not yet
tion, depending on the goals defined by the models. The models been studied deeply (Rosen et al., 2015); these paradigms can also
do not change, but trigger adaptation. On a higher level, context contribute to understanding general issues such as the required
change may require model modification: models representing the granularity of mapping of DTs to physical counterparts and managing
physical counterpart should be modified or replaced. Self-ware emergent system behaviour. Due to the complexity of the system of
models can be used which understand when and how to modify systems created by networking DTs, promoting PD and UD whilst
themselves (Schaumeier et al., 2012). mitigating undesirable behaviour will be challenging. Again, inves-
Adaptive interaction with DTs is also considered in literature, with tigating and adapting methods successfully applied in the fields of MAS
adaptive user interfaces proposed which learn the preferences and and HMS can contribute to solving this problem. Concepts such as
priorities of their users in different contexts and change accordingly resource, operational and product holons (Van Brussel, 2014; Leitão
(Madni et al., 2019, 1982). et al., 2003) and product agents (Främling et al., 2003) need to be
looked at in this light. Agent-based and Holonic Manufacturing Sys-
tems (HMS) have successfully shown how digital representations
4. Research and development gaps of physical entities can autonomously cooperate in a manufactur-
ing context towards common goals (Giret and Botti, 2004; Botti
This section identifies the research and development gaps to be and Giret, 2008; Wang and Haghighi, 2016) and can help under-
bridged to realise the potentials of DTs summarised in Table 1. stand how systems of networked DTs may be organised e.g. as agent
Fig. 3 below shows a roadmap of identified research challenges heterarchies/holarchies.
towards context-aware, autonomous, adaptive DTs. It is read from An issue regarding adaptivity is the definition of mechanisms
top to bottom, left to right. It shows one line for each issue and a for the translation of business goals into actionable plans that the
separate one for general issues. Research gaps are shown at the physical counterpart can carry out and can adapt in response to
intersection of these lines with those of the identified research unexpected situations. Methods and corresponding models are
topics and are as described in the following. required to achieve this translation, to realise the potentials of DTs
The first line deals with interoperability. To enable the poten- to act autonomously and adaptively within the context of higher-
tials of context-aware DTs, context information must be integrated level business goals. A further open issue is modelling interaction
into the system state component. A general issue here is matu- between networked DTs for cooperation and collaboration. Agent
ration research towards its harmonization with the IoT paradigm. interactions are complex to manage, so simpler models might be
The role of DTs in IoT needs to be better understood, as well as more appropriate for DTs. DTs represent physical systems which
which IoT protocols are appropriate for DTs. This is a prerequisite may change over time. The models used to represent them need to
e.g. for including third party data sources into DTs. Further matu- reflect these changes. Appropriate evolving, dynamic and traceable
ration of OPC-UA as a core protocol of Industry 4.0 may contribute models are required. Research in this field exists but more work is
to solving this issue (Jaekel et al., 2020). The establishment of DFs necessary (Lu et al., 2019). Another general issue is that reference
built on networked DTs will require a comprehensive approach to models for DTs are lacking and a comprehensive framework is
inter- and intra-factory, supply chain and lifecycle-wide interoper- required to facilitate full-stack solutions and avoid fragmentation.
ability. Approaches in the intelligent CPS field to monitor physical A final general research gap regarding modelling relates to the
processes, create a virtual copy of the physical world and make abovementioned standards and protocols and reference models.
decentralized decisions (Lu, 2017) can be exploited for this pur- The work towards full integration of the DT concept with RAMI 4.0
pose. However, such approaches need to be revisited and extended wrt. AAS and its convergence with CPS must progress to reinforce
to be useful for DTs interoperability. Indeed, whilst the CPS con- its potential in smart manufacturing and Digital Factories.
cept focusses more on sensors and actuators and the integration The third line of research deals with DT human interaction. At
and collaboration of computing, communication and control, DTs the intersection with context-awareness, research is also required
emphasise models and data (Tao et al., 2019b), and go beyond as how to appropriately represent humans in context models used
CPS in their incorporation of the system behaviour element, with by DTs. Borders between DT and human autonomy must be investi-
simulation and prediction outside of the scope of the core CPS gated to ensure DTs have as much autonomy as possible without
concept. Therefore, combining DTs with semantic interoperability sacrificing human agency. This involves investigating workplace
approaches such as mediation (Shani et al., 2017), the Industrial safety, psychological factors of workers’ motivation, and the lim-

7
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

Table 1
Identified Potentials.

System State System Design System Behaviour

Context-Awareness • Identify relevant contexts • Context-aware configuration of • Integrate rich context models
• Process context information physical counterparts • Include physical, logical, and external
• Incorporate context information • Understand context and reflect it in context models
• Provide context information design models • Include models of tasks, objectives, and
• Context-aware data sources selection • Support design and configuration operations
• Include third party data sources adaptability

Autonomy • Context support for autonomous control • Feedback loop to design • Advanced, multi-model simulations in
• Proactive/reactive autonomous control • Feedback loop to configuration autonomous decision-making
• Autonomous networked entities • Feedback loop to maintenance • Context-aware, autonomous selection,
integration, and execution of physical,
simulation and analytical models

Adaptivity • Context-aware data source adaptation • Adaptive models of complex systems • Real-time system adjustment and
• Control parameter adaptivity • Design and configuration model optimisation
• Control configuration adaptivity adaptivity • Context-aware model adaptivity
• Context-aware behaviour model
adaptivity
• Self-aware, self-modifying models

Fig. 3. Roadmap of Research Challenges.

itations of AI decision-making, etc. To effectively interact with starting point to address these issues. Specifically, solutions at the
humans in production, DTs need to be able to understand and pre- interface level represent the manufacturing space through Virtual
dict the behaviour of humans working alongside them in different and Augmented Reality based on multitouch and voice control
roles. Touching on adaptivity, Human-to-DT Interaction Models are (Gorecky et al., 2014). More visionary concepts discuss cognitive
required which consider different humans’ profiles and how and intelligent assistants using voice or even brain wave interfaces
by what medium interaction is appropriate. CPS too aim to include (Schirner et al., 2013). With respect to the control strategy, dis-
humans in the control loop of autonomous systems (Fantini et al., tributed manufacturing scheduling control paradigms have been
2016) and the solutions proposed in this field represents a good proposed which integrate the human decision-maker’s preferences

8
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

into the control process (Gaham et al., 2015). Finally, DT models, Acknowledgements
simulations and predictions need to be explainable and, in some
cases, certified for widespread industrial adoption, especially in This work is partly funded by the European Commission project
safety and security-critical human-machine interaction applica- H2020 UPTIME “Unified Predictive Maintenance System” (768634),
tions. the European Commission Marie Skłodowska-Curie Research and
The final research line deals with issues of data process- Innovation Staff Exchange (MSCA RISE) project FIRST “vF Interop-
ing for real-time, proactive decision-making. DTs will need eration suppoRting buSiness innovation” (734599), and the H2020
to acquire context information in real-time to enable real-time project DIH4CPS “Fostering DIHs for Embedding Interoperability in
decision-making and self-aware, self-modifying models capable. Cyber-Physical Systems of European SMEs” (872548).
Developments e.g. in ultra-low latency 5 G context-aware appli-
cations may inform these developments (Nunna et al., 2015).
Real-time, proactive decision-making itself requires methods for
References
making time-constrained near-optimal decisions. This might dictate
the use of approximation techniques and could entail autonomic Alam, K.M., El Saddik, A., 2017. C2PS: a Digital Twin architecture reference model
actions, rather than only providing recommendations for humans. for the cloud-based cyber-physical systems. IEEE Access 5, 2050–2062, http://
Furthermore, massive volumes of historical data and streaming dx.doi.org/10.1109/ACCESS.2017.2657006.
Alexopoulos, K., Makris, S., Xanthakis, V., Sipsas, K., Chryssolouris, G., 2016. A concept
data need to be handled and analysed in real-time to forecast for context-aware computing in manufacturing: the white goods case. Int. J.
events, which will require new methods. Most current systems Comput. Integr. Manuf. 29, 839–849, http://dx.doi.org/10.1080/0951192X.2015.
cannot handle big data in real-time because of scalability prob- 1130257.
Anderl, R., Haag, S., Schützer, K., Zancul, E., 2018. Digital twin technology – an
lems, the need to cleanse noisy data offline, or difficulties in fusing approach for Industrie 4.0 vertical and horizontal lifecycle integration. It - Inf.
heterogeneous data types and sources online. Most analyses are Technol. 60, 125–132, http://dx.doi.org/10.1515/itit-2017-0038.
today done on offline data, which means new approaches need to Antsaklis, P.J., Passino, K.M., Wang, S.J., 1991. An introduction to autonomous control
systems. IEEE Control Syst. 11, 5–13, http://dx.doi.org/10.1109/37.88585.
be developed. In contrast, data scarcity can result in problems of Attanasio, A., Ghiani, G., Grandinetti, L., Guerriero, F., 2006. Auction algorithms for
epistemic uncertainty, where a lack of knowledge can lead to diffi- decentralized parallel machine scheduling. Parallel Comput. 32, 701–709, http://
culties in simulating or predicting system behaviour. Approaches to dx.doi.org/10.1016/J.PARCO.2006.03.002.
autonomy | Definition of autonomy in English by Lexico Dictionaries, (n.d.). https://
tackling epistemic uncertainty include incorporating expert knowl-
www.lexico.com/en/definition/autonomy (Accessed 9 July 2019).
edge, and applying interval theory (Wang et al., 2018). Methods Azouz, N., Pierreval, H., 2019. Adaptive smart card-based pull control systems
of one and few-shot learning may also be applied to tackle this in context-aware manufacturing systems: training a neural network through
multi-objective simulation optimization. Appl. Soft Comput. J. 75, 46–57, http://
problem (Zhang et al., 2019).
dx.doi.org/10.1016/j.asoc.2018.10.051.
Bao, J., Guo, D., Li, J., Zhang, J., 2019. The modelling and operations for the digital twin
in the context of manufacturing. Enterp. Model. Inf. Syst. Archit. 13, 534–556,
5. Conclusions http://dx.doi.org/10.1080/17517575.2018.1526324.
Bergweiler, S., 2016. Smart factory systems-fostering cloud-based manufacturing
Context-aware, autonomous, adaptive DTs have the potential based on self-monitoring cyber-physical systems. J. Adv. Syst. Meas. 9, 91–101
(Accessed 19 November 2019) https://www.researchgate.net/profile/Simon
to constitute the building blocks of tomorrow’s DFs by supporting Bergweiler/publication/307908127 Smart Factory Systems - Fostering Cloud-
complex interactions via a multi-layered integration of the infor- based Manufacturing based on Self-Monitoring Cyber-Physical Systems/
mation related to various activities throughout the factory and the links/57d170fb08ae601b39a1c8aa.pdf.
Bicocchi, N., Cabri, G., Mandreoli, F., Mecella, M., 2018. Dealing with data and soft-
product lifecycle. Specifically, the vision of a Digital Factory archi-
ware interoperability issues in digital factories. In: Peruzzini, M., Pellicciari, M.,
tecture (Cardin, 2019) where DTs wrap physical entities involved Bil, C., Stjepandić, J., Wognum, N. (Eds.), Transdiscipl. Eng. Methods Soc. Innov.
and interact in an intelligent dataspace of diverse and distributed Ind. 4.0 - Proc. 25th ISPE Inc. Int. Conf. Transdiscipl. Eng. July 3–6, 2018. IOS Press,
pp. 13–22, http://dx.doi.org/10.3233/978-1-61499-898-3-13.
data sources can significantly reduce the complexity in the coordi-
Bicocchi, Nicola, Cabri, Giacomo, Mandreoli, Federica, Mecella, Massimo, 2019.
nation and planning of supply chains, manufacturing systems and Dynamic digital factories for agile supply chains: An architectural approach.
product lifecycles. JOURNAL OF INDUSTRIAL INFORMATION INTEGRATION 15, 110–121.
The potentials of these DTs outlined above can enable the aggre- Bisio, I., Garibotto, C., Grattarola, A., Lavagetto, F., Sciarrone, A., 2018. Exploit-
ing context-aware capabilities over the internet of things for industry 4.0
gating analyses of data exchanged throughout the product lifecycle, applications. IEEE Netw. 32 (3), 101–107, http://dx.doi.org/10.1109/MNET.2018.
succinctly reflecting an up-to-date model of their operational 1700355.
states. In appropriately reflecting and predicting system behaviour Borangiu, T., Trentesaux, D., Thomas, A., Leitão, P., Barata, J., 2019. Digital transfor-
mation of manufacturing through cloud services and resource virtualization.
and acting autonomously, they can proactively adapt manufactur- Comput. Ind. 108, 150–162, http://dx.doi.org/10.1016/j.compind.2019.01.006.
ing, supply chain and product lifecycles processes in anticipation Boschert, S., Rosen, R., 2016. Digital Twin—the simulation aspect. In: Hehenberger,
of changes, probable faults, and fluctuations. This would contribute P., Bradley, D. (Eds.), Mechatron. Futur. Springer International Publishing, Cham,
pp. 59–74, http://dx.doi.org/10.1007/978-3-319-32156-1 5.
to more resilience of the overall system, ready to both capture new Botti, V., Giret, A., 2008. Holonic manufacturing systems. In: ANEMONA A Mulit-
business opportunities and successfully deal with sudden issues Agent Methodol. Holonic Manuf. Syst. Springer, London, London, pp. 7–20,
that may negatively impact processes. http://dx.doi.org/10.1007/978-1-84800-310-1 2.
Bousdekis, A., Magoutas, B., Apostolou, D., Mentzas, G., 2015. A proactive decision
As described in the potentials, a Digital Factory architecture
making framework for condition-based maintenance. Ind. Manag. Data Syst.
built from context-aware, autonomous, adaptive DTs which could 115, 1225–1250, http://dx.doi.org/10.1108/IMDS-03-2015-0071.
autonomously cooperate or compete to achieve common goals, Bousdekis, A., Papageorgiou, N., Magoutas, B., Apostolou, D., Mentzas, G., 2017. A
Framework for Integrated Proactive Maintenance Decision Making and Supplier
realising many of the benefits of context-awareness, autonomy and
Selection. Springer, Cham, pp. 416–424, http://dx.doi.org/10.1007/978-3-319-
adaptivity. However, the research gaps outlined above need to be 66923-6 49.
addressed in a systematic way to achieve the appropriate techno- Bousdekis, A., Mentzas, G., Hribernik, K., Lewandowski, M., von Stietencron, M.,
logical maturity to achieve this goal. The following, final section Thoben, K.-D.K.D., 2019. A unified architecture for proactive maintenance in
manufacturing enterprises. In: Proc. I-ESA Conf., Springer International Publish-
proposes a roadmap of future research and development work to ing, pp. 307–317, http://dx.doi.org/10.1007/978-3-030-13693-2 26.
reach that maturity. Cardin, O., 2019. Classification of cyber-physical production systems applications:
proposition of an analysis framework. Comput. Ind. 104, 11–21, http://dx.doi.
org/10.1016/j.compind.2018.10.002.
Declaration of Competing Interest Chinesta, F., Cueto, E., Abisset-Chavanne, E., Duval, J.L., El Khaldi, F., 2020. Virtual,
digital and hybrid twins: a new paradigm in data-based engineering and engi-
neered data. Arch. Comput. Methods Eng. 27, 105–134, http://dx.doi.org/10.
The authors report no declarations of interest. 1007/s11831-018-9301-4.

9
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

Cimino, C., Negri, E., Fumagalli, L., 2019. Review of digital twin applications in manu- Hribernik, K., Rabe, L., Schumacher, J., Thoben, K.-D., 2005. A concept for product-
facturing. Comput. Ind. 113, http://dx.doi.org/10.1016/j.compind.2019.103130. instance-centric information management. IEEE Int. Technol. Manag. Conf. ICE
CoBuilder, 2018. The ‘digital Twin’ – A Bridge Between the Physical and the Dig- 2005, 2016, http://dx.doi.org/10.1109/ITMC.2005.7461301.
ital World. https://cobuilder.com/en/the-digital-twin-a-bridge-between-the- Hribernik, K., von Stietencron, M., Bousdekis, A., Bredehorst, B., Mentzas, G., Thoben,
physical-and-the-digital-world/. K.-D., 2018. Towards a unified predictive maintenance system - a use case in
Cronrath, C., Aderiani, A.R., Lennartson, B., 2019. Enhancing digital twins through production logistics in aeronautics. Procedia Manuf. 16, 131–138, http://dx.doi.
reinforcement learning. IEEE Int. Conf. Autom. Sci. Eng., IEEE Computer Society, org/10.1016/j.promfg.2018.10.168.
293–298, http://dx.doi.org/10.1109/COASE.2019.8842888. Hülsmann, M., Grapp, J., 2005. Autonomous cooperation in international- supply-
S.M. Dambrot, Symbiotic Autonomous Systems, Digital Twins and Artificial networks the need for a shift from centralized planning to decentralized decision
Intelligence: Emergence and Evolution, n.d., http://mondodigitale.aicanet.net/ Making in logistic processes. In: Pawar, K.S., Lalwani, C.S., de Carvalho, J.C., Muf-
2019-1/articoli/03 MD80 Symbiotic Autonomous Digital Twins and Artificial fatto, M. (Eds.), Proc. 10th Int. Symp. Logist. (10th ISL), Loughborough. , pp.
Intelligence.pdf (Accessed 3 July 2020). 243–249.
de Visser, E., Parasuraman, R., 2011. Adaptive aiding of human-robot team- Hülsmann, M., Windt, K., 2007. Understanding Autonomous Cooperation and
ing. J. Cogn. Eng. Decis. Mak. 5, 209–231, http://dx.doi.org/10.1177/ Control in Logistics: The Impact of Autonomy on Management, Infor-
1555343411410160. mation, Communication and Material Flow. https://books.google.de/
El Saddik, A., 2018. Digital Twins: the convergence of multimedia technologies. IEEE books?hl=en&lr=&id=7fAusY75q88C&oi=fnd&pg=PA1&dq=autonomous+
Multimed. 25, 87–92, http://dx.doi.org/10.1109/MMUL.2018.023121167. cooperation&ots=qNs EVn3vt&sig=y5Mfdu1TfvcQIFayB3N5Mbacv0A.
Elwany, A.H., Gebraeel, N.Z., 2008. Sensor-driven prognostic models for equip- Hülsmann, M., Kopfer, H., Cordes, P., Bloos, M., 2009. Collaborative transportation
ment replacement and spare parts inventory. IIE Trans. (Institute Ind. Eng. 40, planning in complex adaptive logistics systems: a complexity science-based
629–639, http://dx.doi.org/10.1080/07408170701730818. analysis of decision-making problems of “groupage systems” in: complex 2009.
Engel, Y., Etzion, O., Feldman, Z., 2012. A basic model for proactive event-driven In: Complex Sci. Springer, Berlin, pp. 1160–1166, http://dx.doi.org/10.1007/978-
computing. In: Proc. 6th ACM Int. Conf. Distrib. Event-Based Syst. - DEBS’ 12. 3-642-02466-5 116.
ACM Press, New York, New York, USA, pp. 107–118, http://dx.doi.org/10.1145/ Inagaki, T., 2003. Hanbook of cognitive task design, 8 adaptive automation:
2335484.2335496. sharing and trading of control. In: Hollnagel, E. (Ed.), Handb. Cogn. Task Des.
Erd"os, G., Paniti, I., Tipary, B., 2020. Transformation of robotic workcells to digital Lawrence Erlbaum Associates, Mahwah, New Jersey, pp. 147–169 (Accessed
twins. CIRP Ann. 69, 149–152, http://dx.doi.org/10.1016/j.cirp.2020.03.003. 14 May 2020) https://books.google.de/books?hl=en&lr=&id=dElPH0ruR-
Fantini, P., Tavola, G., Taisch, M., Barbosa, J., Leitao, P., Liu, Y., Sayed, M.S., Lohse, sC&oi=fnd&pg=PA147&dq=Adaptive+automation:+Sharing+and+trading+of
N., 2016. Exploring the integration of the human as a flexibility factor in CPS +control&ots=EHnWp30PFE&sig=s16dM3vldm9XMaTp84R5aEbcv8U.
enabled manufacturing environments: methodology and results. IECON Proc. Jaekel, F.W., Wolff, T., Happersberger, V., Knothe, T., 2020. Ensure OPC-UA interfaces
(Industrial Electron. Conf.), http://dx.doi.org/10.1109/IECON.2016.7793579. for digital plug-and-produce. In: Lect. Notes Comput. Sci. (Including Subser. Lect.
Främling, K., Holmström, J., Ala-Risku, T., Kärkkäinen, M., 2003. Product agents for Notes Artif. Intell. Lect. Notes Bioinformatics). Springer, pp. 44–53, http://dx.doi.
handling information about physical objects. Rep. Lab. Inf. Process. Sci. Ser. B 153, org/10.1007/978-3-030-40907-4 5.
1–20 (Accessed 11 December 2020) https://research.aalto.fi/en/publications/ Joo, T., Shin, D., 2019. Formalizing human–Machine interactions for adaptive
product-agents-for-handling-information-about-physical-objects. automation in smart manufacturing. IEEE Trans. Human-Machine Syst. 49 (6),
Gabor, T., Belzner, L., Kiermeier, M., Beck, M.T., Neitz, A., 2016. A simulation-based 529–539, http://dx.doi.org/10.1109/THMS.2019.2903402.
architecture for smart cyber-physical systems. Proc. - 2016 IEEE Int. Conf. Auton. Josifovska, K., Yigitbas, E., Engels, G., 2019. Reference framework for digital twins
Comput. ICAC 2016, IEEE, 374–379, http://dx.doi.org/10.1109/ICAC.2016.29. within cyber-physical systems. In: Proc. - 2019 IEEE/ACM 5th Int. Work. Softw.
Gaham, M., Brahim, B., Achour, N., 2015. Human-in-the-loop cyber-physical pro- Eng. Smart Cyber-Physical Syst. SEsCPS 2019, Institute of Electrical and Elec-
duction systems control (HiLCP2sC): a multi-objective interactive framework tronics Engineers Inc, pp. 25–31, http://dx.doi.org/10.1109/SEsCPS.2019.00012.
proposal. Stud. Comput. Intell. 594, 315–325, http://dx.doi.org/10.1007/978-3- Kapteyn, M.G., Knezevic, D.J., Huynh, D.B.P., Tran, M., Willcox, K.E., 2020. Data-driven
319-15159-5 29. physics-based digital twins via a library of component-based reduced-order
Gebhardt, N., Jeken, O., Windt, K., 2011. Exploitation of manufacturing flexibilities models. Int. J. Numer. Methods Eng., 1–18, http://dx.doi.org/10.1002/nme.6423.
in decision methods for autonomous control of production processes: findings Kassner, L., Hirmer, P., Wieland, M., Steimle, F., Königsberger, J., Mitschang, B.,
from industrial practice and theoretical analysis. In: Auton. Coop. Control Logist. 2017. The social factory: connecting people, machines and data in manufac-
Springer, Berlin Heidelberg, Berlin, Heidelberg, pp. 169–187, http://dx.doi.org/ turing for context-aware exception escalation. Proc. 50th Hawaii Int. Conf. Syst.
10.1007/978-3-642-19469-6 13. Sci., Hawaii International Conference on System Sciences, http://dx.doi.org/10.
Giret, A., Botti, V., 2004. Holons and agents. J. Intell. Manuf. 15, 645–659, http://dx. 24251/hicss.2017.202.
doi.org/10.1023/B:JIMS.0000037714.56201.a3. Klein, P., van der Vegte, W.F., Hribernik, K., Klaus-Dieter, T., 2019. Towards an
Glaessgen, E., Stargel, D., 2012. The Digital Twin Paradigm for Future NASA and U.S. approach integrating various levels of data analytics to exploit product-usage
Air Force Vehicles., http://dx.doi.org/10.2514/6.2012-1818, p. 1818. information in product development. In: Proc. Int. Conf. Eng. Des. ICED, Cam-
Gockel, B., Tudor, A., Brandyberry, M., Penmetsa, R., Tuegel, E., 2012. Challenges With bridge University Press, pp. 2627–2636, http://dx.doi.org/10.1017/dsi.2019.269.
Structural Life Forecasting Using Realistic Mission Profiles., http://dx.doi.org/10. Koulamas, C., Kalogeras, A., 2018. Cyber-physical systems and digital twins in the
2514/6.2012-1813, p. 1813. industrial internet of things [cyber-physical systems]. Computer (Long. Beach.
Gonçalves, G., Peschl, M., 2018. Adaptability in smart manufactur- Calif). 51, 95–98, http://dx.doi.org/10.1109/MC.2018.2876181.
ing systems. In: Seventh Int. Conf. Intell. Syst. Appl. INTELLI 2018, Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W., 2018. Digital Twin in manu-
IARAIA, pp. 36–42 (Accessed 18 July 2019) https://www.google.com/ facturing: a categorical literature review and classification. IFAC-Papers On Line
url?client=internal-uds-cse&cx=016414469947898432502:-hnzqf1l 51, 1016–1022, http://dx.doi.org/10.1016/j.ifacol.2018.08.474.
ic&q=http://www.thinkmind.org/download.php%3Farticleid%3Dintelli 2018 Kulvatunyou, B.S., Wallace, E., Kiritsis, D., Smith, B., Will, C., 2018. The industrial
2 30 65004&sa=U&ved=2ahUKEwjv2rP31b7jAhXC0KQKHaFZAqMQFjAAeg ontologies foundry proof-of-concept project. In: IFIP Adv. Inf. Commun. Tech-
QIBRAB&usg=AOvVaw16hSPyNsmZdybM1w93mgvd. nol. Springer New York LLC, pp. 402–409, http://dx.doi.org/10.1007/978-3-319-
Gorecky, D., Schmitt, M., Loskyll, M., Zühlke, D., 2014. Human-machine-interaction 99707-0 50.
in the industry 4.0 era. In: Proc. - 2014 12th IEEE Int. Conf. Ind. Informatics, INDIN Lee, A.N., Martinez Lastra, J.L., 2013. Enhancement of industrial monitoring systems
2014, Institute of Electrical and Electronics Engineers Inc, pp. 289–294, http:// by utilizing context awareness. In: 2013 IEEE Int. Multi-Disciplinary Conf. Cogn.
dx.doi.org/10.1109/INDIN.2014.6945523. Methods Situat. Aware. Decis. Support, IEEE, pp. 277–284, http://dx.doi.org/10.
Grieves, M., Vickers, J., 2016. Digital twin: mitigating unpredictable, undesirable 1109/CogSIMA.2013.6523858.
emergent behavior in complex systems. In: Transdiscipl. Perspect. Complex Syst. Lee, J., Lapira, E., Yang, S., Kao, A., 2013. Predictive manufacturing system - trends of
New Find. Approaches. Springer International Publishing, Cham, pp. 85–113, next-generation production systems. Ifac Proc. Vol. 46, 150–156, http://dx.doi.
http://dx.doi.org/10.1007/978-3-319-38756-7 4. org/10.3182/20130522-3-BR-4036.00107.
Guo, J., Zhao, N., Sun, L., Zhang, S., 2019. Modular based flexible digital twin for Lee, H., Yoo, S., Kim, Y.-W., 2016. An energy management framework for smart
factory design. J. Ambient Intell. Humaniz. Comput. 10, 1189–1200, http://dx. factory based on context-awareness. In: 2016 18th Int. Conf. Adv. Commun.
doi.org/10.1007/s12652-018-0953-6. Technol., IEEE, pp. 685–688, http://dx.doi.org/10.1109/ICACT.2016.7423520.
Hartmann, D., Herz, M., Wever, U., 2018. Model order reduction a key technology Leitão, P., Colombo, A.W., Restivo, F., 2003. An Approach to the Formal Specification
for digital twins. In: Reduced-Order Model. Simul. Optim. Powerful Algorithms of Holonic Control Systems. Springer, Berlin, Heidelberg, pp. 59–70 https://doi.
as Key Enablers Sci. Comput. Springer International Publishing, pp. 167–179, org/10.1007/978-3-540-45185-3 6.
http://dx.doi.org/10.1007/978-3-319-75319-5 8. Leng, J., Zhang, H., Yan, D., Liu, Q., Chen, X., Zhang, D., 2019. Digital twin-driven man-
He, Q.P., Wang, J., 2018. Statistical process monitoring as a big data analytics tool ufacturing cyber-physical system for parallel controlling of smart workshop. J.
for smart manufacturing. J. Process Control 92, 333–351, http://dx.doi.org/10. Ambient Intell. Humaniz. Comput. 10, 1155–1166, http://dx.doi.org/10.1007/
1016/j.jprocont.2017.06.012. s12652-018-0881-5.
He, R., Chen, G., Dong, C., Sun, S., Shen, X., 2019. Data-driven digital twin technology Lindberg, P., 1990. Strategic manufacturing management: a proactive
for optimized control in process systems. ISA Trans. 95, 221–234, http://dx.doi. approach. Int. J. Oper. Prod. Manag. 10, 94–106, http://dx.doi.org/10.1108/
org/10.1016/j.isatra.2019.05.011. 01443579010001221.
Hochhalter, J.D., Leser, W.P., Newman, J.A., Glaessgen, E.H., Gupta, V.K., Yamakov, Louie, A.H., 2010. Robert Rosen’s anticipatory systems. Foresight 12, 18–29, http://
V., Cornell, S.R., Willard, S.A., Heber, G., 2014. Coupling Damage-Sensing dx.doi.org/10.1108/14636681011049848.
Particles to the Digitial Twin Concept., pp. 10 https://ntrs.nasa.gov/search. Lu, Y., 2017. Industry 4.0: a survey on technologies, applications and open research
jsp?R=20140006408. issues. J. Ind. Inf. Integr. 6, 1–10, http://dx.doi.org/10.1016/J.JII.2017.04.005.

10
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

Lu, J., Zheng, X., Gharaei, A., Kalaboukas, K., Kiritsis, D., 2019. Cognitive Twins for Sup- Ríos, J., Hernández, J.C., Oliva, M., Mas, F., 2015. Product avatar as digital counterpart
porting Decision-Makings of Internet of Things Systems (Accessed 19 February of a physical individual product: literature review and implications in an aircraft.
2020) http://arxiv.org/abs/1912.08547. Adv. Transdiscipl. Eng., 657–666, http://dx.doi.org/10.3233/978-1-61499-544-
Lu, Y., Liu, C., Wang, K.I.-K., Huang, H., Xu, X., 2020. Digital Twin-driven smart 9-657.
manufacturing: connotation, reference model, applications and research issues. Rosen, R., von Wichert, G., Lo, G., Bettenhausen, K.D., 2015. About the importance
Robot. Comput. Manuf. 61, 101837, http://dx.doi.org/10.1016/J.RCIM.2019. of autonomy and digital twins for the future of manufacturing. IFAC-Papers On
101837. Line 48, 567–572, http://dx.doi.org/10.1016/J.IFACOL.2015.06.141.
Lucke, D., Constantinescu, C., Westkämper, E., 2008. Smart factory - a step towards Rosenberger, P., Gerhard, D., 2018. Context-awareness in industrial applications:
the next generation of manufacturing. In: Manuf. Syst. Technol. New Front., definition, classification and use case. Procedia CIRP 72, 1172–1177, http://dx.
http://dx.doi.org/10.1007/978-1-84800-267-8 23. doi.org/10.1016/j.procir.2018.03.242.
Lützenberger, J., Klein, P., Hribernik, K., Thoben, K.-D., 2016. Improving product- Sancarlos, A., Cameron, M., Abel, A., Cueto, E., Duval, J.L., Chinesta, F., 2021. From ROM
service systems by exploiting information from the usage phase. A case study. of electrochemistry to AI-based battery digital and hybrid twin. Arch. Comput.
Procedia CIRP 47, 376–381, http://dx.doi.org/10.1016/J.PROCIR.2016.03.064. Methods Eng. 28, 979–1015, http://dx.doi.org/10.1007/s11831-020-09404-6.
Madni, A.M., Samet, M.G., Freedy, A., 1982. A trainable on-line model of the human Sanderson, D., Antzoulatos, N., Chaplin, J.C., Busquets, D., Pitt, J., German, C., Norbury,
operator in information acquisition tasks. IEEE Trans. Syst. Man Cybern. 12, A., Kelly, E., Ratchev, S., 2015. Advanced manufacturing: an industrial application
504–511, http://dx.doi.org/10.1109/TSMC.1982.4308855. for collective adaptive systems. Proc. - 2015 IEEE 9th Int. Conf. Self-Adaptive Self-
Madni, A., Madni, C., Lucero, S., 2019. Leveraging digital twin technology in Organizing Syst. Work. SASOW 2015, http://dx.doi.org/10.1109/SASOW.2015.
model-based systems engineering. Systems 7, 7, http://dx.doi.org/10.3390/ 15.
systems7010007. Saracco, R., 2019. Digital Twins: bridging physical space and cyberspace. Computer
Maes, P., 1993. Modeling adaptive autonomous agents. Artif. Life 1, 135–162, http:// (Long. Beach. Calif). 52, 58–64, http://dx.doi.org/10.1109/MC.2019.2942803.
dx.doi.org/10.1162/artl.1993.1.1 2.135. Schaumeier, J., Pitt, J., Cabri, G., 2012. A tripartite analytic framework for characteris-
Mandelbaum, M., 1978. Flexibility in Decision Making: An Exploration and Unifica- ing awareness and self-awareness in autonomic systems research. In: 2012 IEEE
tion. University of Toronto, Toronto, Canada. Sixth Int. Conf. Self-Adaptive Self-Organizing Syst. Work., IEEE, pp. 157–162,
Mourtzis, D., Doukas, M., Vlachou, A., Xanthopoulos, N., 2014. Machine availability http://dx.doi.org/10.1109/SASOW.2012.35.
monitoring for adaptive holistic scheduling: a conceptual framework for mass Schirner, G., Erdogmus, D., Chowdhury, K., Padir, T., 2013. The future of human-
customization. Procedia CIRP 25, 406–413, http://dx.doi.org/10.1016/j.procir. in-the-loop cyber-physical systems. Computer (Long. Beach. Calif). 46, 36–45,
2014.10.056. http://dx.doi.org/10.1109/MC.2013.31.
Mullen, R.J., Monekosso, D., Barman, S., Remagnino, P., 2009. A review of ant algo- Schleich, B., Anwer, N., Mathieu, L., Wartzack, S., 2017. Shaping the digital twin for
rithms. Expert Syst. Appl. 36, 9608–9617, http://dx.doi.org/10.1016/J.ESWA. design and production engineering. CIRP Ann. - Manuf. Technol. 66, 141–144,
2009.01.020. http://dx.doi.org/10.1016/j.cirp.2017.04.040.
Müller, V.C., 2012. Autonomous cognitive systems in real-world environments: less Scholz-Reiter, B., 2004. Autonomous logistic processes: New demands and first
control, more flexibility and better interaction. Cognit. Comput. 4, 212–215, approaches. Proc. 37th CIRP Int. Semin. Manuf. Syst., 357–362 http://www.
http://dx.doi.org/10.1007/s12559-012-9129-4. sfb637.uni-bremen.de/fileadmin/SFB Files/PDF Download/SFB637 CIRP 2004.
Nadin, M., 2017. Predictive and anticipatory computing. In: Encycl. Comput. Sci. pdf.
Technol. Routledge Handbooks Online, pp. 643–659, http://dx.doi.org/10.1081/ Scholz-Reiter, B., Rekersbrink, H., Wenning, B.-L., Makuschewitz, T., 2008. A survey of
E-ECST2-120054027. autonomous control algorithms by means of adapted vehicle routing problems.
Nunna, S., Kousaridas, A., Ibrahim, M., Dillinger, M., Thuemmler, C., Feussner, H., In: Proc. 9th Bienn. ASME Conf. Eng. Syst. Des. Anal., ASME, Haifa, Isreal, pp. 1–6,
Schneider, A., 2015. Enabling real-time context-aware collaboration through 5G http://dx.doi.org/10.1115/ESDA2008-59077.
and mobile edge computing. In: Proc. - 12th Int. Conf. Inf. Technol. New Gener., B. Scholz-Reiter, M. Freitag, C. de Beer, T. Jagalski, Modelling and analysis of
ITNG 2015, Institute of Electrical and Electronics Engineers Inc, pp. 601–605, autonomous shop floor control, Sfb637.Uni-Bremen.De. (n.d.). http://www.
http://dx.doi.org/10.1109/ITNG.2015.155. sfb637.uni-bremen.de/fileadmin/SFB Files/PDF Download/SFB637-A5a-Ib-05-
Otto, B., ten Hompel, M., Wrobel, S., 2018. Industrial data space. In: Digitalisierung. 01.pdf (Accessed 10 July 2019).
Springer, Berlin Heidelberg, pp. 113–133, http://dx.doi.org/10.1007/978-3-662- Schuldt, A., Berndt, J.O., Herzog, O., 2011. The interaction effort in autonomous logis-
55890-4 8. tics processes: potential and limitations for cooperation. In: Auton. Coop. Control
Pachter, M., Chandler, P.R., 1998. Challenges of autonomous control. IEEE Control Logist. Springer, Berlin Heidelberg, Berlin, Heidelberg, pp. 77–90, http://dx.doi.
Syst. 18, 92–97, http://dx.doi.org/10.1109/37.710883. org/10.1007/978-3-642-19469-6 7.
Park, H.-S., Tran, N.-H., 2011. An autonomous manufacturing system for adapting to Shani, U., Franke, M., Hribernik, K.A., Thoben, K.-D., 2017. Ontology mediation to rule
disturbances. Int. J. Adv. Manuf. Technol. 56, 1159–1165, http://dx.doi.org/10. them all: managing the plurality in product service systems. 11th Annu. IEEE
1007/s00170-011-3229-2. Int. Syst. Conf. SysCon 2017 - Proc., http://dx.doi.org/10.1109/SYSCON.2017.
Park, K.T., Lee, J., Kim, H.J., Do Noh, S., 2020. Digital twin-based cyber physical 7934810.
production system architectural framework for personalized production. Int. J. Sha to, M., Conroy, M., Doyle, R., Glaessgen, E., Kemp, C., LeMoigne, J., Wang, L.,
Adv. Manuf. Technol. 106, 1787–1810, http://dx.doi.org/10.1007/s00170-019- 2012. A modeling, simulation, information technology & processing roadmap.
04653-7. Natl. Aeronaut. Sp. Adm., 1–38.
Passino, K.M., 1995. Intelligent control for autonomous systems. IEEE Spectr. 32, Stark, R., Damerau, T., 2019. Digital Twin. Int. Acad. Prod. Eng. CIRP Encycl. Prod.
55–62, http://dx.doi.org/10.1109/6.387144. Eng., http://dx.doi.org/10.1007/978-3-642-35950-7 16870-1.
Peruzzini, M., Pellicciari, M., 2017. A framework to design a human-centred adaptive Steiner, N.J., 2008. Autonomous Computing Systems, ProQuest
manufacturing system for aging workers. Adv. Eng. Inf. 43 (C), 330–349, http:// Diss. Theses. DP19191., pp. 176 (Accessed 13 June 2019)
dx.doi.org/10.1016/j.aei.2017.02.003. http://ezproxy.net.ucf.edu/login?url=http://search.proquest.
Petersen, S.A., van der Kooij, R., Puhar, P., 2016. Connecting Business Processes com/docview/1021197367?accountid=10003%5Cnhttp://sfx.fcla.
and Sensor Data in Proactive Manufacturing Enterprises. Springer, Cham, pp. edu/ucf?url ver=Z39.88-2004&rft val fmt=info:ofi/fmt:kev:mtx:
101–109, http://dx.doi.org/10.1007/978-3-319-45390-3 9. dissertation&genre=dissertations+&+theses&sid=ProQ:ProQuest+
Preuveneers, D., Joosen, W., Ilie-Zudor, E., 2018. Robust digital twin compositions for Dissertations+&+.
industry 4.0 smart manufacturing systems. In: Proc. - IEEE Int. Enterp. Distrib. Talukdar, S., Baerentzen, L., Gove, A., De Souza, P., 1998. Asynchronous teams: coop-
Object Comput. Work. EDOCW, Institute of Electrical and Electronics Engineers eration schemes for autonomous agents. J. Heuristics. 4, 295–321, http://dx.doi.
Inc, pp. 69–78, http://dx.doi.org/10.1109/EDOCW.2018.00021. org/10.1023/A:1009669824615.
Qi, Q., Tao, F., 2018. Digital twin and big data towards smart manufacturing and Tantik, E., Anderl, R., 2017. Potentials of the asset administration shell of industrie
industry 4.0: 360 degree comparison. IEEE Access 6, 3585–3593, http://dx.doi. 4.0 for service-oriented business models. Procedia CIRP, 363–368, http://dx.doi.
org/10.1109/ACCESS.2018.2793265. org/10.1016/j.procir.2017.03.009, Elsevier B.V.
Qu, T., Lei, S.P., Wang, Z.Z., Nie, D.X., Chen, X., Huang, G.Q., 2016. IoT-based real-time Tao, Fei, Zuo, Ying, Da Xu, Li, Zhang, Lin, 2014. IoT-based intelligent perception and
production logistics synchronization system under smart cloud manufacturing. access of manufacturing resource toward cloud manufacturing. IEEE Trans. Ind.
Int. J. Adv. Manuf. Technol. 84, 147–164, http://dx.doi.org/10.1007/s00170-015- Informatics 10, 1547–1557, http://dx.doi.org/10.1109/tii.2014.2306397.
7220-1. Tao, F., Zhang, H., Liu, A., Nee, A.Y.C.Y.C., 2019a. Digital twin in industry: state-of-
Radziwon, A., Bilberg, A., Bogers, M., Madsen, E.S., 2014. The smart factory: exploring the-art. IEEE Trans. Ind. Inform. 15, 2405–2415, http://dx.doi.org/10.1109/TII.
adaptive and flexible manufacturing solutions. Procedia Eng. 69, 1184–1190, 2018.2873186.
http://dx.doi.org/10.1016/j.proeng.2014.03.108. Tao, F., Qi, Q., Wang, L., Nee, A.Y.C., 2019b. Digital Twins and cyber–physical systems
Rasheed, A., San, O., Kvamsdal, T., 2019. Digital Twin: Values, Challenges and toward smart manufacturing and industry 4.0: correlation and comparison.
Enablers., pp. 1–31, Eprint ArXiv:1910.01719 http://arxiv.org/abs/1910.01719 Engineering 5, 653–661, http://dx.doi.org/10.1016/j.eng.2019.01.014.
(Accessed 29 October 2019). Tennenhouse, D., 2000. Proactive computing. Commun. ACM 43, 43–50, http://dx.
Rassolkin, A., Vaimann, T., Kallaste, A., Kuts, V., 2019. Digital twin for propulsion doi.org/10.1145/332833.332837.
drive of autonomous electric vehicle. In: 2019 IEEE 60th Annu. Int. Sci. Conf. Tuegel, E., 2012. The Airframe Digital Twin: Some Challenges to Realization., http://
Power Electr. Eng. Riga Tech. Univ. RTUCON 2019 - Proc., Institute of Electrical dx.doi.org/10.2514/6.2012-1812, p. 1812.
and Electronics Engineers Inc., http://dx.doi.org/10.1109/RTUCON48111.2019. E. Tuegel, A. Ingraffea, . . . T.E.-I.J. of, undefined 2011, Reengineering aircraft struc-
8982326. tural life prediction using a digital twin, Hindawi.Com. (n.d.). https://www.
Reifsnider, K., Majumdar, P., 2013. Multiphysics Stimulated Simulation Digital hindawi.com/journals/ijae/2011/154798/abs/ (Accessed 29 October 2019).
Twin Methods for Fleet Management., http://dx.doi.org/10.2514/6.2013-1578, Tzanis, N., Andriopoulos, N., Magklaras, A., Mylonas, E., Birbas, M., Birbas, A., 2020.
p. 1578. A hybrid cyber physical digital twin approach for smart grid fault prediction.

11
K. Hribernik, G. Cabri, F. Mandreoli et al. Computers in Industry 133 (2021) 103508

In: Proc. - 2020 IEEE Conf. Ind. Cyberphysical Syst. ICPS 2020, Institute of Elec- proceedings, and he has received 7 best paper awards. He is (and has been) involved
trical and Electronics Engineers Inc, pp. 393–397, http://dx.doi.org/10.1109/ in international and national projects, of which 2 FP7 projects and 1 H2020 project.
ICPS48405.2020.9274723. He has been general chair of IEEE WETICE 2004, of ACM PPPJ 2008 and of IEEE WETICE
Van Brussel, H., 2014. Holonic manufacturing systems. In: CIRP Encycl. Prod. Eng. 2014.
Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 654–659, http://dx.doi.org/
10.1007/978-3-642-20617-7 6556.
C. Wagner, J. Grothoff, U. Epple, . . . R.D.-2017 22nd I., undefined 2017, The role of Federica Mandreoli is an Associate Professor in the Ital-
the Industry 4.0 asset administration shell and the digital twin during the life ian SSD ING-INF/05 Ïnformation Management Systemsät
cycle of a plant, Ieeexplore.Ieee.Org. (n.d.), https://ieeexplore.ieee.org/abstract/ the University of Modena and Reggio Emilia. She got
document/8247583/ (Accessed 16 December 2019). the Italian National Habilitation (Abilitazione Scientifica
Wang, L., Haghighi, A., 2016. Combined strength of holons, agents and function Nazionale) as full professor in the Scientific Sector 09/H1
blocks in cyber-physical systems. Int. J. Ind. Manuf. Syst. Eng. 40, 25–34, http:// (Information Processing Systems) and 01/B1 (computer
dx.doi.org/10.1016/J.JMSY.2016.05.002. science). Since several years, she coordinates the research
Wang, C., Matthies, H.G., Xu, M., Li, Y., 2018. Epistemic uncertainty-based model activity of the Ïnformation Systemsg̈roup (www.isgroup.
validation via interval propagation and parameter calibration. Comput. Methods unimo.it). Her research concerns the management of
Appl. Mech. Eng. 342, 161–176, http://dx.doi.org/10.1016/j.cma.2018.08.001. massive information and knowledge and data science in
Wang, J., Ye, L., Gao, R.X., Li, C., Zhang, L., 2019. Digital Twin for rotating machinery different application contexts. In the last years, her inter-
fault diagnosis in smart manufacturing. Int. J. Prod. Res. 57, 3920–3934, http:// ests are geared towards clinical medicine and Industry 4.0.
dx.doi.org/10.1080/00207543.2018.1552032. On these topics, she has several national and international
Wieland, M., Leymann, F., Schafer, M., Lucke, D., Constantinescu, C., Westkamper, collaborations and published more than 100 papers in refereed journals, book, and
E., 2010. Using context-aware workflows for failure management in a smart in proceedings of international and national conferences. Moreover, she is associate
factory. UBICOMM 2010 - 4th Int. Conf. Mob. Ubiquitous Comput. Syst. Serv. editor of the two international journals Data Science and Engineering (Elsevier) and
Technol. International Journal of Knowledge-Based and Intelligent Engineering Systems (IOS
Windt, K., Becker, T., Jeken, O., Gelessus, A., 2010. A classification pattern for Press). She participated in many national and international projects and she is cur-
autonomous control methods in logistics. Logist. Res. 2, 109–120, http://dx.doi. rently one of the PI of the Marie-Curie RISE Project FIRST focused on Digital and Smart
org/10.1007/s12159-010-0030-9. factories. She served as PC member of many international conferences and sympo-
Wuest, T., Hribernik, K., Thoben, K.-D., 2013. Digital Representations of Intelligent siums and she is one of the organizers of the CAiSE workshop K̈ET4DF: Key Enabling
Products: Product Avatar 2.0. Springer, Berlin Heidelberg, pp. 675–684, http:// Technologies for Digital Factories¨. She was referee for European Research Coun-
dx.doi.org/10.1007/978-3-642-30817-8 66. cil (ERC) Starting Grant funding calls, Italian Ministry of University and Research
Zaccaria, V., Stenfelt, M., Aslanidou, I., Kyprianidis, K.G., 2018. Fleet monitoring and (MIUR) FIRB and SIR funding calls,France-South America STIC AmSud Program, and
diagnostics framework based on digital twin of aero-engines. Proc. ASME Turbo Austria FWF Program.
Expo, American Society of Mechanical Engineers (ASME), http://dx.doi.org/10.
1115/GT2018-76414. Gregoris Mentzas is full Professor of Management Infor-
Zhang, H., Liu, Q., Chen, X., Zhang, D., Leng, J., 2017. A Digital Twin-based approach mation Systems, School of Electrical and Computer
for designing and multi-objective optimization of hollow glass production line. Engineering, National Technical University of Athens
IEEE Access 5, 26901–26911, http://dx.doi.org/10.1109/ACCESS.2017.2766453. (NTUA) and Founder and Director of the Information Man-
Zhang, J., Zhao, C., Ni, B., Xu, M., Yang, X., 2019. Variational Few-Shot Learning. agement Unit (IMU) at the Institute of Communication and
Zheng, P., Sivabalan, A.S., 2020. A generic tri-model-based approach for product- Computer Systems (ICCS). His main research focus is on AI-
level digital twin development in a smart manufacturing environment. Robot. enabled decision making: augmenting human intelligence
Comput. Manuf. 64, 101958, http://dx.doi.org/10.1016/j.rcim.2020.101958. with data-driven approaches and using Artificial Intelli-
gence to elevate human decision-making competence. He
Karl A. Hribernik studied Computer Science at the Uni- has published 4 books and more than 300 papers in inter-
versity of Bremen. He worked as a software developer national peer-reviewed journals and conferences, has 5
at Productec Ingenieurgesellschaft mbH on both research best papers awards, sits on the editorial board of five inter-
and commercial e-logistics and e-commerce projects from national journals and has served as (co-)Chair or Program
1997 to 2002. He joined BIBA – Bremer Institut für Committee Member in more than 55 international conferences. He has led or con-
Produktion und Logistik GmbH as a research scientist tributed in more than 50 European research and development projects. Research
in 2002. He has been the head of the department of carried out by his group has led to the establishment of three internet technology
Intelligent ICT for Co-operative Production at BIBA since companies. His experience includes twelve years of management consulting in cor-
2013 and manages BIBA’s iotfablab. His research interests porate strategy and information systems strategy. He holds a Diploma Degree in
are focussed on product-centric information manage- Engineering (1984) and a Ph.D. in Operations Research and Information Systems
ment and closed-loop product lifecycle management, and (1988) both from NTUA.
include the investigation of the Internet of Things and
Industrie 4.0 in the product lifecycle; semantic models,
interoperability and search; item-level digital product representations and digital
twins; data analysis, AI and machine learning and knowledge-based engineering. He
has contributed to more than 100 papers in national and international proceedings
and has received multiple best paper awards.

Giacomo Cabri is full professor in Computer Engineering

at the Department of Physics, Informatics and Mathemat-
ics of Università degli studi di Modena e Reggio Emilia,
Italy. He is IEEE senior member and ACM senior mem-
ber. He is Rector Delegate for teaching at the Università
degli studi di Modena e Reggio Emilia, Deputy Head of
the Department of Physics, Informatics and Mathematics
and member of the board of the department. He teaches
different courses in the field of Computer Science and
Information Technology, both base ones and advanced
ones. He is member of the council of the doctoral course
in ICT.His research activities are mainly related to the fol-
lowing areas: software agents, distributed applications,
autonomic computing, self-org systems. In these areas, he has published around
180 papers in international and national journals, and in international and national

12