Insights into SoRa: A Reference Architecture for Cyber-physical

Social Systems in the Industry 4.0 Era

Teodor Ghetiu

and Bogdan-Constantin Pirvu

Undo Ltd., Cambridge, U.K.

Department of Industrial Engineering and Management, Lucian Blaga University of Sibiu, Sibiu, Romania

Keywords: Reference Architecture, Learning, Adaptation, Cyber-physical Social System, Anthropocentric.

Abstract: Reference architectures for Industry 4.0 tend to have a techno-centric orientation; their social dimension is

usually restricted to specifying that users exist, and they have concerns that impact the architecture of desired

systems. We take a step further to make the social element the core of future systems. A first step is to propose

a reference architecture for Industry 4.0 cyber-physical social systems (CPSS), that builds upon proposals

from well-known initiatives. Key differentiator in our design is the explicit consideration of the human –

cyber-physical relation and the way the two sides influence or adapt to each other. The final aim is that

architecture descriptions derived from this reference architecture, will enable the development of CPSSs

capable of harnessing the power of the Internet of Things (IoT), while respecting the importance of their

human members.

1 INTRODUCTION

The engineering of cyber-physical systems (CPSs)

has always been guided by architectures that took into

account, more or less, the concerns of their

stakeholders. Some argue that one essential

component of the systems - the actual user - has often

been just partially considered (Dressler, 2018).

As we progress in the Industry 4.0 age, more and

more complex reference architectures (RAs) are

being put forward, striving to accommodate the

development of new technologies and IoT related

capabilities. This is even more the case with the

advent of approaches such as Society 5.0 in Japan

(Hitachi-UTokyo Laboratory, 2020), a model

describing a people-centric super-smart society in

which humanity is a key trait for this ideal society.

In this particular context, cyber-physical social

systems (CPSSs), which are computing systems

adding social characteristics and interaction to CPSs

having key features such as: integrality, sociability,

locality, irreversibility, adaptivity and autonomy

(Pirvu et al., 2016), are considered to still be in their

“infancy”, as recent studies lack a systematic design

methodology or are application specific (Zeng et al.,

https://orcid.org/0000-0002-4637-4679

https://orcid.org/0000-0003-3961-4539

2020). More importantly, there is room for giving a

more socio-centric orientation to reference

architectures, such that the architectures derived from

them will serve in developing improved CPSSs in

approaches such as Society 5.0.

This research analyses a number of reference

architectures, from FITMAN, to OSMOSE, IoT-A or

BEinCPPS. The findings led us to propose

improvements in order to explicitly consider the

relation between the social side of a system and its

cyber-physical counterpart. Furthermore, we suggest

that establishing equilibrium between the two sides

can be achieved if we facilitate the adaptation of one

to the other. To this end, we propose a first effort to

elaborate a reference architecture for CPSS – the

Socio-centric RA (SoRA), as part of our ongoing

research. Please note that in this article only the

structural perspective of SoRA will be detailed, as

well as to propose an instantiation of a functional

perspective.

The remaining of this paper is structured as

follows. Section II presents the relevant work

regarding RAs. Section III describes SoRA from the

structural perspective as well as an instantiation from

a functional perspective while section IV discusses

Ghetiu, T. and Pirvu, B.

Insights into SoRa: A Reference Architecture for Cyber-physical Social Systems in the Industry 4.0 Era.

DOI: 10.5220/0009982300470052

In Proceedings of the International Conference on Innovative Intelligent Industrial Production and Logistics (IN4PL 2020), pages 47-52

ISBN: 978-989-758-476-3

some key aspects of this RA. In section V concluding

remarks are formulated, while in the final section the

outlook for SoRA’s development is presented.

2 RELATED WORKS

There are numerous RAs for CPSS, some more

detailed, others succinctly presented, some more

generic, others tailored to specific types of systems,

etc. In the following we are going to briefly look at a

number of RAs that we consider representative and

are sufficiently documented in order to be evaluated.

The OSMOSE project delivered a RA (Felic, et

al., 2014) (Felic, et al., 2016) intended to enable the

development of sensing-liquid enterprises. These two

attributes were inspired from the FInES Research

Roadmap 2025 (FInES Research Roadmap Task

Force, 2012), which identified nine qualities of being

(QB) of Future Internet-based Enterprises. These nine

QBs are: 1) Humanistic Enterprise, 2) Inventive

Enterprise, 3) Agile Enterprise, 4) Cognisant

Enterprise, 5) Sensing Enterprise, 6) Community-

oriented Enterprise, 7) Liquid Enterprise, 8) Global

Enterprise and 9) Sustainable Enterprise. Key to the

OSMOSE reference architecture is the identification

of three worlds (real, digital and virtual) to which an

enterprise’s assets belong; communication between

worlds is mediated by a “membrane”, which allows

osmotic processes to take place (information entering

the membrane is processed and routed to the other

worlds according to complex event processing and

knowledge links mechanisms). From a socio-centric

perspective, the OSMOSE reference architecture

considers the human users only in terms of the data

and multimedia information the system stores for the

users, or in terms of avatars that may be used in

“what-if” simulations pertaining to the virtual world.

The FITMAN project delivered three RAs

(Rotondi et al., 2013) that define what a smart, digital

or virtual enterprise should consist of. From a socio-

centric perspective, humans are identified as end-

users, that only control the system from a logically

remote location. While the RAs benefit from the

identification of a collection of reusable components,

mainly dedicated to information processing or to

abstracting out details of lower-levels of abstraction,

the information relevant to the socio-centric

perspective is not abundant here as this is not the

focus of the FITMAN RAs.

The IoT-A project produced a comprehensive

piece of work (The Internet of Things – Architecture

project, 2013) that not only defines a RA, but also

provides the underlying architectural reference model

(ARM), as well as guidance for using the RA in order

to generate specific architectures. While the work

itself is not complete (the information model, for

example, is only partially defined), the ARM contains

eloquent descriptions of the domain, functional and

communication model, while the RA itself complies

with the framework of already established

architectural views and perspectives, such as those

denoted in (Rozanski and Woods, 2005). That said,

the overall approach is techno-centric: the RA is

composed of functional groups and functional

components, with the user interacting with the system

via the top-level applications functional group.

More recently, the Industrial Internet Consortium

(IIC) delivered the industrial Internet Reference

Architecture (Industrial Internet Consortium, 2017),

which is strongly based on the ISO/IEC 42010

(ISO/IEC/IEEE, 2011). This RA, in fact, instantiates

ISO/IEC 42010 for the Industrial Internet domain,

selecting four relevant viewpoints (i.e. business,

usage, functional and implementation) and detailing

the elements that need to be defined in order to

generate views for each viewpoint; in addition, the

RA specifies a number of cross-cutting concerns,

while also providing architectural patterns (i.e.

topologies for interconnecting physical devices or

logical layers within an enterprise) that may be

applied when constructing specific system

architectures.

In a similar timeframe, the BEinCPPS project

finalized its RA, oriented on cyber-physical

production systems (CPPS). While initially the

BEinCPPS RA (Fischer, et al., 2016) was a cross-

breed between the OSMOSE and RAMI 4.0 RAs, the

final version (Isaja et al., 2017) adopts a simplified

approach, that makes use of four perspectives in order

to define a semi-reference semi-concrete architecture.

Its simple structural perspective divides the elements

of a system into design-time and runtime, while

runtime systems are considered at different

hierarchical levels. Again, the approach is techno-

centric, the human element being either implicit

(hence undefined) or explicit only at the top-level

cloud level (similar to the FITMAN approach of

representing humans only as end-users).

Table 1 below provides a synthetic view over the

mentioned architectures, describing the types of

systems they were intended for, as well as the scope

that they can be associated with. A detailed review of

recent reference architectures for cyber-physical

systems in industry 4.0 is presented in (Ghetiu, 2018).

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics

Table 1: Some existing reference architectures and their

scope and applicability.

Reference

Architecture

Intended

system

Scope

OSMOSE Sensing-Liquid

Enterprise

Individual

enterprise

FITMAN

Smart

Factory

Smart Factory Shop floor

FITMAN

Digital

Factory

Digital Factory Factory (data

analytics)

FITMAN

Virtual

Factory

Virtual Factory Supply chains

IIRA IoT CPSoS General

applicability

IoT-A CPSoS General

applicability

BEinCPPS Cyber-physical

production

systems

Field devices,

Factory,

Cloud

3 PROPOSED WORK

SoRA adopts the BEinCPPS structural perspective as

a starting point. One of the main aspects that lead to

this decision is the structural simplicity and clarity of

this RA, in its final format (Isaja et al., 2017), which

is highly suitable for being adapted in order to present

functions and technologies that could be used in

concrete architectures; in comparison, other RAs can

be considered either too complicated (e.g. FITMAN

or OSMOSE) or too abstract (e.g. the 5C architecture

(Lee et al., 2015)). Additionally, the BEinCPPS RA

makes a clear distinction between the design and

operation concerns of an architecture, so it represents

a good foundation for creating an improved, yet not

too complicated RA.

If the BEinCPPS RA is composed of two domains

and three layers, SoRA has a more elaborate

structure, as shown in figure 1 where we present the

structural perspective of SoRA; we highlight in green

the new elements added on top of the BEinCPPS

structural perspective. The design-time and runtime

domains remain but, in order to adequately transition

between the two, a third buffer domain is needed; this

approach reflects to an extent the OSMOSE

philosophy, where the distinct worlds are separated

by a “membrane”.

The new buffer domain is dedicated to

representing the relation between the social and

cyber-physical elements within the intended system;

its role is to explicitly define how the cyber and the

social actors will work well within the CPSS. The

domain consists of two layers, one dedicated to

training, the other to adaptation. The training layer

defines how the human users will be prepared for

using the CSP efficiently, whereas the adaptation

layer refers to the adaptation of the CSP to the humans

inside the CPSS.

Furthermore, the design-time domain needs to be split

into an upper layer dedicated to capturing information

pertaining to the social realm, whereas the lower layer

remains restricted to the standard design of a CPS.

The motivation for this separation stems from the

need to explicitly consider the social factors that will

influence the CPSS design.

Figure 1: Structural view of the SoRA for CPSS.

Insights into SoRa: A Reference Architecture for Cyber-physical Social Systems in the Industry 4.0 Era

In addition, the runtime domain (named

“Execution” in figure 1) contains three layers, as that

of the BEinCPPS architecture; the two RAs different

in the sense that in the SoRA, the layers have generic

titles, whereas the BEinCPPS is focused on CPPS.

Consequently, at the bottom of the hierarchy we have

CPSs (analogue for the shop floor level of the

BEinCPPS runtime domain), which are the building

block for CPSoSs and ultimately the Cloud (analogue

for the enterprise level of the BEinCPPS runtime

domain). Note that the Edge is not explicitly depicted

in this version of the architecture, but is expected to

be contained within each runtime layer.

While it is still under development, figure 2 offers

a glimpse of what may be a functional view

implementing the SoRA. Here we identify

components, technologies or processes that may be

used in each layer of the architecture.

The Training layer, for example, can be defined

by training processes and training systems; this

implies that a CPSS, designed with the social-factor

in mind, should come equipped with documentation

supporting the training of personnel, but also (if the

complexity of the systems deems it necessary) with

explicit training systems that may be used in the

training process (Gellert et al., 2020). If we were to

take the case of an Industry 4.0 production facility, in

order to support the social factor, apart from training

processes, a VR or AR enabled training system could

be developed or purchased.

The Adaptation layer refers to the systems

capability to model their user and adapt to them, so

that quality metrics can be improved. Machine

learning (ML) techniques can be employed, but also

newer approaches such as generative design or

design-space exploration, especially in co-simulated

environments, can be used.

The runtime domain can be instantiated with

agent-oriented platforms that readily implement

logic, not only for executing production functions, but

also for interfacing with users in “smart” ways. As

(Ocker et. al 2019) suggest, multi-agent systems

(MAS) are a solution to modern challenges faced by

production systems or other types of complex,

human-design systems. Software agents can be

mapped to individual devices, users or to aggregates

of such entities; they can activate solely within the

software domain or become embodied. Furthermore,

any MAS comes with a solution for inter-agent

communication; if the design of a CPSS tunes this

communication so that it is done efficiently, across

CPSs and levels of hierarchy, then MASs represent

indeed a powerful proposition for the composition of

CPSS.

What is more important is that MAS can be

employed in defining the functions of the socio-

cyber-physical relation domain. Specialized agents

can aid or fully execute the training of human users,

whereas the adaptation of the CPS to its users can be

enacted by other types of specialized agents,

embodied or not.

The bottom layer of the runtime domain (CPS)

brings together devices and users, that are interfaced

through a specific class of MAS. The middle layer

interconnects CPSs and users, through a potentially

different class of MAS. Finally, at the top-layer we

find Cloud functions: data processing (e.g. big data),

together with specific applications which support the

Figure 2: Functional view of the SoRA for CPSS.

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics

collaboration between users or the execution of other

system functions.

Going forward, the ARM for the SoRA

architecture should be supplied. We take as starting

point IoT-A’s ARM and identify aspects that should

be modified or added so that it becomes relevant. This

work will be presented in a follow-up article.

4 DISCUSSION

The information provided in the previous section is

just a brief introduction to SoRA - a developing

reference architecture for CPSS. In this paper, we

only define its structural perspective and sketch what

could be its functional perspective.

The structural perspective builds on that provided

by the BEinCPPS RA; a key differentiator is the

addition of an explicit buffer domain, that bridges the

design and runtime domains. Its purpose is to expose

the importance of the socio - cyber-physical relation,

for the development of successful CPSSs. Our initial

thoughts go towards the need for defining the way in

which the human factor is going to be accustomed to

the cyber-physical one (i.e. training), as well as on

explicitly allowing (or even requesting) the CPS to

adapt itself to the human factor, so that it can better

support it.

When looking at the functional perspective, we

consider it is necessary to put more focus on the tools

and means for designing the social factors into the

architectures of CPSSs, in contrast with focusing on

the technical aspects of CPS architectures. While

existing RAs vary in terms of scope, qualities built

into the targeted system or even the level of detail

with which they are made public, we aim to obtain a

RA that is socio-oriented, builds on existing best

practices and conforms to the Future Internet

prospects for Industry 4.0.

Another defining trait of the SoRA is the centering

on multi-agency as a philosophy for addressing the

needs of a modern CPSS. MAS can be employed in

order to implement the training and adaptation

functions of our buffer defined, as well as in bringing

together CPSs and human actors, at all hierarchical

levels within a CPSS.

5 CONCLUSIONS

In this work we have introduced SoRA, a new RA

aimed at facilitating the development of architecture

descriptions for CPSSs, that reflect the importance of

the human factor as a key element in the control loop.

To achieve this, SoRA builds on previous work that

spans decades of research and implementation. Key

in making a difference is considering that the relation

between the social and the cyber-physical elements

needs to be added to the core of all CPSSs

architectures.

Previous RAs have been of a predominantly

techno-centric nature: the human was the external

factor, the user that interacts with the system via a

more or less advanced interface; important was

achieving the functions of the system. In the best of

cases, ISO/IEC 42010 (ISO/IEC/IEEE, 2011) was

brought into attention, reminding that architectures

have to consider the concerns of stakeholders or,

specific to the IoT domain, architectures would need

to define an entire business view (Industrial Internet

Consortium, 2017). Other architectures were too

succinctly defined (such as the SmartFactory or

RAMI 4.0 RAs) to evaluate their orientation, but we

can go by the rule that if something is not explicitly

defined, it does not exist; as such, we cannot attribute

a socio-centric nature to such architectures.

There are cases when focus is explicitly laid on

other attributes (or qualities of being, as (FInES

Research Roadmap Task Force, 2012) considers); the

OSMOSE RA (Felic, et al., 2016) aims at identifying

architectural constructs that enable the development

of architectural descriptions for sensitive and liquid

enterprises. RAs need to map out the spectrum of QBs

and OSMOSE’s approach is valid from this

perspective.

In this paper, we have detailed only the structural

and functional perspectives of SoRA. Its structural

perspective builds on that provided by the BEinCPPS

RA, to which it adds a new domain: that of the socio

- cyber-physical relationship. The intention is to

explicitly consider this relationship from the onset, so

that the resulting architecture will lead to the

implementation of a CPSS where human users are

well trained in efficiently using the CPSS, while the

cyber-physical components will continuously adapt

to the social factors that interact with them. SoRA’s

functional perspective is still in a developing stage,

but one aspect can be considered defined: its reliance

on MAS to achieve functions related to training,

adaptation, and human-computer interaction.

6 FUTURE WORK

As part of the research conducted for SoRA, the next

step is to further elaborate on the means through

which social factors can be (best) taken into account

Insights into SoRa: A Reference Architecture for Cyber-physical Social Systems in the Industry 4.0 Era

at design time. The current proposal is to look at

quality perspectives, using notions derived from

standards such as ISO/IEC 9126-4 or ISO/IEC 20510.

We need to evaluate if proposals such as IIRA’s

business viewpoint are satisfactory from a socio-

centric perspective, or a more in-depth view is

needed, including user modelling, profiling.

Another line of research is that of implementing

prototype systems, that reflect the new functions

described in the socio – cyber-physical relation

domain of SoRA. The first prototype is an adaptive

system to correctly learn how to manually assemble

products without a human instructor. The adaptation

aims at adjusting the instructions for the user

according to the previous and current performance in

the execution of the task, the chosen components, the

physical and emotional state of the operator as well as

the detected user profile. The second prototype is a

modular production system prototype having a

distributed low-level control architecture together

with a MAS for the high-level control. Fully

automatic the system can produce standard orders

(i.e. modular tablets), orders which have limited or

predefined customization; in case of highly

customized orders, the automated system collaborates

with human operators to manufacture the special

orders.

ACKNOWLEDGEMENTS

This work is supported through the DiFiCIL project

(contract no. 69/08.09.2016, ID P_37_771, web:

http://dificil.grants.ulbsibiu.ro), co-funded by ERDF

through the Competitiveness Operational Programme

2014-2020.

REFERENCES

Dressler, F., 2018. Cyber Physical Social Systems:

Towards Deeply Integrated Hybridized Systems,

International Conference on Computing, Networking

and Communications. IEEE.

Felic, A., Rubattino, C., Di Vito, G., Agostinho, C., Lucena,

C., Fischer, K., Liudmila, D., Sesana, M. 2014. D 3.11

- First OSMOSE Models and Architecture.

Felic, A., Sesana, M., Dobriakova, L., Rubattino, C.,

Ferreira, J., Marques, M., Fisher K., Collado, L.,

Gonzales, L., Sanguini, R., 2016. D 3.12 Final

OSMOSE Models and Architecture.

FInES Research Roadmap Task Force., 2012. Research

Roadmap 2025. EU COMMISSION.

Fischer, K., Jara, A., Geven, A., Rotondi, D., Coscia, E.,

Aştefănoaei, L., Rooker, M., Voss, S., 2016. D2.1 -D

2.1a BEinCPPS Architecture and Business Processes.

Gellert, A., Precup, SA., Pirvu, BC., Zamfirescu, CB.,

2020. Prediction-Based Assembly Assistance System,

In ETFA 2020, 25th International Conference on

Emerging Technologies and Factory Automation.

IEEE.

Ghetiu, T., 2018. A review of recent reference architectures

for cyber-physical systems, in the industry 4.0 era. In

Acta Universitatis Cibiniensis – Technical Series, Vol.

LXX. SCIENDO.

Hitachi-UTokyo Laboratory., 2020. Society 5.0 - A People-

centric Super-smart Society, Springer, Singapore, 1

edition.

Industrial Internet Consortium., 2017. Volume G1:

Reference Architecture. INDUSTRIAL INTERNET

CONSORTIUM.

Isaja, M., Fischer, K., Rotondi, D., Coscia, E., Rooker, M.,

2017. D2.2 - BEinCPPS Architecture & Business

Processes.

ISO/IEC/IEEE., 2011. Systems and software engineering —

Architecture description. ISO/IEC.

Lee, J., Bagheri., B, Kao, H.-A., 2015. A Cyber-Physical

Systems architecture for Industry 4.0-based

manufacturing. Manufacturing Letters, Vol. 3.

SCIENCEDIRECT.

Ocker, F., Kovalenko, I., Barton, K., Tilbury, D., Vogel-

Heuser, B., 2019. A Framework for Automatic

Initialization of Multi-Agent Production Systems Using

Semantic Web Technologies, IEEE Robotics and

Automation Letters, Volume 4, Issue 4. IEEE.

Pirvu, B.-C., Zamfirescu, C.-B., Gorecky, D. 2016.

Engineering insights from an anthropocentric cyber-

physical system: A case study for an assembly station,

Mechatronics, Vol. 34. ELSEVIER.

Rotondi, D., Benedicto, J., Lázaro, O., Sesana, M.,

Gusmeroli, S., 2013. Deliverable D1.5: FITMAN

Reference Architecture.

Rozanski, N., Woods, E., 2005. Software System

Architecture. ADDISON-WESLEY

PROFESSIONAL.

The Internet of Things – Architecture project., 2013.

Deliverable D1.5 - Final architectural reference model

for the IoT v3.0.

Zeng, J., Yang, L. T., Lin, M., Ning, H., Ma, J., 2020. A

survey: Cyber-physical-social systems and their

system-level design methodology. Future Generation

Computer Systems, Vol. 105. ELSEVIER.

IN4PL 2020 - International Conference on Innovative Intelligent Industrial Production and Logistics