ACPS: Adaptive Cyber-Physical Systems in Industry 4.0

Sebastien Ducos

and Ernesto Exposito

LIUPPA, University of Pau et des Pays de l’Adour, E2S UPPA, Anglet, France

Keywords: Industry 4.0, Cyber-Physical Systems, Decision Support Systems, Autonomic Computing, Reconfigurable

Systems.

Abstract: Nowadays, with the rapid growth of connected objects and produced data involved in industrial processes, it

is increasingly difficult to design and implement efficient cyber-physical systems (CPS) meeting business

needs. As a consequence, architectures of CPS have to be able to integrate different heterogeneous actors

(people, objects, data, services) coordinated by autonomous and self-adaptive processes capable of

implementing the different business missions of a company. Moreover, with the emergence of Industry 4.0,

interest in elastic services provided by cloud architectures is booming. Indeed, these architectures allow the

smooth and scalable interconnection of interdependent systems in order to provide efficient solutions to

facilitate the management of industrial processes. In this paper, we propose a generic architecture for

Integration Platforms as a Service (iPaaS). This architecture offers key functionalities, namely integration and

interoperability, but also self-decision support. One implementation based on open-source solutions and

illustrating the benefits of this proposal in the area of the Agriculture 4.0 domain is proposed.

1 INTRODUCTION

In their dynamic of continuous improvement and

digitalization, organizations are seeking to integrate

advanced and innovative technologies to ensure their

transition to Industry 4.0. Indeed, the emergence of

Industry 4.0 brings a technological and philosophical

revolution in companies, forcing them to question

their business models. The term "Industry 4.0"

encompasses a set of technologies and concepts

related to the re-organization of the value chain

(Hermann et al., 2015). This term is related to the

accelerated advances enabled and promoted by

information and communication technologies (ICT).

It relies on the communication of real-time

information to monitor and act on physical systems,

thus exploiting a new paradigm: the cyber-physical

systems (CPSs). Different systems communicate and

cooperate with each other, but also with humans, to

decentralize decision-making. Its deployment

requires the integration of different digital technology

know-how (Danjou et al., 2017). The fourth industrial

revolution do not only concern production processes,

but also aim to revolutionize new horizons such as

https://orcid.org/0000-0001-7990-5547

https://orcid.org/0000-0002-3543-2909

new generation smart products and services (Godreui

et al., 2016). It requires the design and

implementation of smart cyber-physical systems

following an appropriate methodology and based on

a concrete architecture that meet the challenges of

integrating IoE actors and their intelligent

coordination (agile, adaptable, reconfigurable and

flexible). They should autonomously provide

information about themselves and exchange

information with other CPSs that are part of the

industrial networks. They should be able to be

adaptive to respond to multi-domain challenges

involving different paradigms. We are talking about

cyber-physical systems of systems (CPSoS).

This article is structured as follows: the next

section presents related works proposing smart

solutions for CPS. The third section describes our

proposal aimed at creating a flexible platform to

manage cognitive processes in CPS able to integrate

compliant data science approaches for decision-

making in the area of the Agriculture 4.0, followed by

some first results. Finally, the conclusions and

perspectives of this work are presented.

704

Ducos, S. and Exposito, E.

ACPS: Adaptive Cyber-Physical Systems in Industry 4.0.

DOI: 10.5220/0012146600003538

In Proceedings of the 18th International Conference on Software Technologies (ICSOFT 2023), pages 704-711

ISBN: 978-989-758-665-1; ISSN: 2184-2833

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

REFERENCE

ARCHITECTURES FOR CPS

In order to guide organizations in their transition to

the 4th industrial revolution and to create an

environment conducive to innovation and to the

creation of CPS adapted to efficiently meet the new

needs, several initiatives have been launched around

the world and several architectures have been

proposed.

The most significant initiatives have been

accompanied by government agencies and private

organizations from countries in the most developed

economies (Wang et al., 2017) (Zhong et al., 2017).

Reference architectures provide a framework for

developing solutions for Industry 4.0 in a structured

way that allows all participants to be involved in a

uniform manner. In this sense, standards can be

identified and optimized. Among these architectures,

we will find well known ones like the Reference

Architectural Model Industrie 4.0 (RAMI4.0) and the

Industrial Internet Reference Architecture (IIRA).

Other different but interesting work has been

carried out with so-called cognitive architectures

because they allow the integration of self-

management capabilities (Kephart et al., 2003).

Among these, we find the Adaptive Control of

Thought - Rational (ACT-R) architecture or the Soar

architecture.

2.1 Industry 4.0 Architectures

As mentioned above, the RAMI 4.0 and IIRA

reference architectures aim to facilitate the Industry

4.0 knowledge sharing paradigm, guide

organizational transitions, and specifically advise on

leveraging ICT advances. Both initiatives seek to

build a more efficient industrial world particularly

through complex, connected and intelligent systems.

A notable difference is that RAMI 4.0 extends this

vision to the entire value chain and product life cycle,

while IIRA maintains a more concrete vision of the

ICT world.

2.1.1 Reference Architecture Model

Industrie 4.0 (RAMI4.0)

The RAMI4.0 architecture is based on three

dimensions, as we can see in Figure 1 below, namely:

the layers (properties and system structures), the

hierarchy levels (from the product to the connected

world) and the life cycle and value stream (product

lifecycle).

The first vertical axis proposes 6 layers (asset,

integration, communication, information, functional

and business) allowing to break down the properties

of a machine on different levels. Thanks to this, even

the most complex systems can be divided and

managed more easily.

Regarding the second right horizontal axis, the

hierarchy levels, from IEC 62264, represents the

different functionalities of organizations. This

dimension characterizes the Industry 4.0 revolution

with the introduction of "Products" as well as the

"Connected World" with the emergence of the

connection of things and services (IoT).

Finally, the left horizontal third axis targets the

products and facilities lifecycle, based on IEC 62890.

We can identify 2 phases: types and instances. The

type phase is characterized by the design and

prototyping of a product. When this phase is

completed and the product is manufactured, the type

phase is transferred to the instance phase (ISA).

Figure 1: Reference Architectural Model Industrie 4.0

(RAMI4.0).

2.1.2 Industrial Internet Reference

Architecture (IIRA)

The Industrial Internet Reference Architecture was

introduced, in 2015, by the Industrial Internet

Consortium (IIC) and updated in 2017 to become an

open standards-based architecture for the Industrial

Internet of Things (IIoT). The architecture proposes 3

dimensions, as we can see in Figure 2, comparable to

the Reference Architectural Model Industrie 4.0

(RAMI4.0), namely: the Viewpoints (Business,

Usage, Functional and Implementation), the Process

Lifecycle (IIoT system conception, design and

implementation) and the Industrial Sectors.

A major goal of IIoT is to connect larger, complex

systems and implement hierarchies for machines.

This architecture is also based on IIoT systems for the

functional part with a decomposition in 5

interconnected domains, namely: control (control and

ACPS: Adaptive Cyber-Physical Systems in Industry 4.0

705

actuation), operations (management and

maintenance), information (data collection and

analysis), application (use-case application) and

business (business goals) (Expósito, 2019).

Figure 2: Industrial Internet Reference Architecture (IIRA).

2.2 Architectures Evaluation

Reference architectures such as RAMI4.0 and IIRA

support integration, interoperability and scalability

needs but do not explicitly consider decision part.

While both IIRA and RAMI 4.0 provide valuable

reference architectures for the design and

implementation of industrial internet systems and

smart factories, they also have some limitations:

 Complexity: both architectures are complex

and can be difficult to implement, especially

for smaller organizations with limited

resources and expertise.

 Standardization: for both architectures, there

is a lack of universal standards in some areas,

such as communication protocols and data

formats.

 Cost: Implementing the architectures can

require significant investment in hardware,

software, and personnel.

Overall, while both IIRA and RAMI 4.0 provide

valuable guidance for the design and implementation

of industrial internet systems and smart factories,

organizations must carefully evaluate the specific

needs and resources before embarking on

implementation.

Based on this analysis, our work proposes an

alternative referential architecture intended to cope

with the criteria and generic enough to be adapted to

different application domains of the Industry 4.0.

3 ADAPTIVE CYBER-PHYSICAL

SYSTEMS

This chapter presents our architecture proposal which

consists in designing a generic architecture for

building cyber-physical systems capable of deploying

autonomous processes, composed of Monitoring,

Analysis, Planning and Execution (MAPE) phases

and including knowledge bases, built using

information provided by experts, to guide automated

decision-making. The solution must consider the need

to make these knowledge bases evolve to deal with

new contexts, new objectives and constraints of

industrial processes.

3.1 5C Layered Referential

Architecture for CPS

In order to facilitate and assist in the design,

implementation and management of cyber-physical

systems for Industry 4.0, a referential architecture in

5C layers will be presented in this section. This

referential architecture is intended to build and

coordinate CPS and to allow cooperation and

collaboration of CPSoS. This architecture is well

suited for CPS involved in Industry 4.0

manufacturing processes, as well as for the

elaboration of smart products and the provision of

smart services.

This proposal promotes a generic and concrete

architectural framework, based on a 5C layered

architecture and resulting from an improvement of the

previously presented reference architectures and the

integration of the Internet of Everything (IoE)

concept.

3.1.1 5C Layers

The 5C Layered architecture follows an incremental

approach that allows the assembly of components of

a CPS and also goes as far as its composition to create

systems of systems (Sanchez, 2020).

The two lowest layers (C1..C2) are intended to

cope with the integrability (connectivity) and

interoperability (communication) challenges of the

heterogeneous actors involved in CPS (people,

things, data, services, etc.). The three highest layers

(C3..C5) are intended to incrementally integrate

monitoring, analysis, planning and management

capabilities required to allow coordination of CPS as

well as cooperation and collaboration of Cyber-

Physical Systems of Systems (CPSoS).

The Table 1 presents each layer and describe the

architectural functionalities offered to the involved

entities.

ICSOFT 2023 - 18th International Conference on Software Technologies

706

Table 1: Architecture layers and functionalities.

Layer Description

Architectural

functionality

C1:

Connection

entities share a

common medium or

channel

Network

Connectivity

C2:

Communication

two or more entities

are able to understand

each other by

exchanging messages

via a common

medium or channel

Integrability

Interoperability

Interaction

modes

C3:

Coordination

two or more entities

working together

following the orders

or the instructions of

a coordinato

Intra-system

orchestration

(CPS)

C4:

Cooperation

two or more entities

work together to

achieve individual

goals

Inter-systems

orchestration

(CPSoS)

C5:

Collaboration

two or more entities

work together to

achieve a common

global goal

Inter-systems

choreography

(CPSoS)

3.1.2 Autonomic Management Dimension

In addition to the 5 levels previously presented

representing a structural dimension for the design of

CPS and CPSoS, our architecture must also integrate

a behavioral dimension allowing the intelligent

management of the structural elements involved.

This behavioral dimension must offer a generic

and scalable approach, allowing to offer self-

adaptation capabilities to the context in order to

enable the achievement of the goals established for

the CPS.

We believe that the architecture proposed by

autonomic computing (AC) offers the framework

required to integrate this behavioral dimension for

self-management.

This architecture offers several structural and

behavioral recommendations to implement self-

management capabilities and thus build an autonomic

system. Adaptive actions are implemented by

adaptive algorithms operating within a closed-loop

control system. These algorithms can be generically

described as a process that includes monitoring,

analysis, planning and execution (MAPE) activities

that share a common knowledge base.

With regard to our 5-levels structure, the

autonomic behavior would develop progressively,

starting from the lowest levels thanks to the

implementation of the required functionalities at the

level of connection and communication to retrieve

observations and execute adaptation actions on the

CPS actors. At the coordination level, the autonomic

MAPE process will allow to self-manage the actors

involved in order to achieve the objectives set for the

CPS. At the cooperation and collaboration levels, the

CPS will function as actors that can be monitored and

who can carry out adaptation actions in order to

achieve the individual or shared objectives of the

CPSoS.

Having now the structural and behavioral

dimensions of our architecture in place, a suitable

methodology is still required to guide the process of

building CPS based on our reference architecture.

To achieve effective orchestration in an

autonomic system, it is necessary to have a high

degree of automation, real-time monitoring and

analysis, and the ability to adapt to changing

conditions.

The following section will introduce a well-suited

system engineering methodology that could be

followed to build CPS based on the Autonomic 5C

layered architecture.

3.1.3 System Engineering Methodology

In order to help innovation and development project

managers in their transition to a more connected

industry adapted to tomorrow's needs, we propose a

methodological approach to determine and define

precisely the different phases allowing designing and

integrating complex systems related to Industry 4.0.

In the area of software engineering and systems

engineering, several methodologies and modeling

frameworks have been proposed for the development

of complex systems.

A recent methodology successfully used at the

industrial level for system engineering and based on

this standard is the ARCADIA methodology

(Capella). This methodology is an example of an

MBSC methodology that also includes a language

(Roques, 2016). We cannot directly compare UML or

SySML with ARCADIA because ARCADIA is both

a language and a method.

Arcadia has been influenced by systems

engineering and in particular the distinction between

requirements and solutions (Roques, 2016). This

method also promotes a viewpoint approach. The

central viewpoint in Arcadia is a functional

viewpoint. Functions are used to describe what the

system needs to do, and then functions to describe

what the logical or physical components do and how,

what they do, will contribute to the system. In

addition, other points of view such as performance or

security must be satisfied and conform to the context

of the specific project. This allows the same system

ACPS: Adaptive Cyber-Physical Systems in Industry 4.0

707

to be seen from many different points of view

depending on the system to be designed.

This methodology proposes 5 incremental phases

to identify the functional and non-functional

requirements of the system (operational and

functional analysis phases) and to design the system

architecture (logical and physical architectures and

EPBS). Moreover, the method has its own language

mainly due to the lack of the concept of functions with

languages like UML or SysML.

Our methodology is based on an extension of the

Arcadia methodology, in order to integrate additional

viewpoints and views, capable of incorporating the

structural and behavioral levels of our referential

architecture for Industry 4.0 CPS.

3.2 Agriculture 4.0 Domain

As this work was carried out in collaboration with the

Maïsadour agricultural cooperative, it was logical to

deploy and evaluate this approach on agricultural

processes, mainly on the cereal drying process.

As presented above, the ARCADIA method was

therefore chosen and followed in order to model an

integration Platform as a Service (iPaaS) type

approach because it allows to design its architecture

while defining, evaluating and exploiting the

collaboration of the systems (Capella). With this

method, our architecture could be divided into 4 parts:

Operational Architecture, System Architecture,

Logical Architecture and Physical Architecture.

The logical architecture, presented in Figure 3,

highlight the different functionalities of the system

and show the collaboration and communication of the

latter by detailing the different sub-functions.

The iPaaS platform is composed of 3 modules or

features: the integration module, the process manager

and the prediction module. The logical actors and

entities, on the left, represent the data sources and

collectors, which can be also interpreted as the

workspace or environment. The integration module

allows the exchange of information between all the

systems and actors involved in the process. In

addition, it will play the role of translator because it

will transform and standardize the data in order to

make all the actors collaborate. Next, we find the

process manager who ensures that the process runs

smoothly step by step. It provides an overview of the

various business processes and their interactions.

Finally, the prediction module, composed of various

decision models, allows the processing, analysis and

prediction of data thanks to knowledge bases

designed from heterogeneous sources (humans, IS,

sensors, PLCs, ...). For this last module, it is essential

to build decision models capable of integrating expert

knowledge while ensuring a suitable accuracy of the

decisions taken.

For the decisional part, i.e. recommendations or

automated decision making, 2 models were initially

integrated and tested in order to evaluate the global

approach. After that, we thought of developing a

more concrete and complete decisioning module,

namely, a generic Data-Driven Decision Support

System (DDDSS) that could meet a wide range of

needs in an adapted and precise manner.

Figure 3: Logical Architecture.

ICSOFT 2023 - 18th International Conference on Software Technologies

708

Figure 4: Data-Driven Decision Support System Architecture.

The architecture platform, Figure 4, is conformed

of six internal elements: local repository, data

evaluation, machine learning, model definition,

predictions of parameters, model monitoring and user

interphase.

Each of the elements is defined based on its

specific functionalities that receive and process data.

In addition, the DDDSS, as an open system, is

interconnected to external components databases,

stakeholders, quantum cloud services, machines to

exchange resources and information.

Our solution therefore meets the various key

elements defined above concerning the

functionalities or services of integration,

interoperability, decision model inclusion,

adaptability, auditability and finally scalability.

4 DEVELOPMENT AND

EXPERIMENTATIONS

In order to develop and evaluate this approach, an

iPaaS prototype has been implemented on several

agricultural processes, including drying, with the aim

of optimizing machine parameters and reducing

energy consumption (gas + electricity) and therefore

also CO2 emissions.

This prototype uses only open-source solutions to

prove its efficiency at low cost. For the integration

module, the Apache Camel framework has been

chosen as the main integrator. For the Process

Manager part, the adopted solution is Camunda BPM.

Moreover, we find other solutions like Apache Kafka,

which is a complementary integrator (Cestari et al.,

2020).

4.1 Implementation

The different solutions used were adopted following

a thorough state of the art on the subject. We

identified the characteristics and functionalities

necessary for the proper functioning of the system in

order to list and compare the solutions that best

corresponded to our needs.

Apache Camel ensures interoperability between

the various systems thanks to a multitude of

connectors (easily developed in case of absence) and

the simple integration of new services or tools.

Apache Kafka, on the other hand, will be useful for

communications with a need for real-time data, as it

is much better adapted than Camel for this part.

Camunda BPM, as process manager, can be

considered as the brain of the system. It will ensure

that the process runs smoothly via BPMN diagrams,

compatible with other editors, which are executed by

integrated engines.

To meet the scalability and elasticity needs of the

iPaaS platform, Docker and Kubernetes were

selected, among several technologies, as the main

components to best manage them.

ACPS: Adaptive Cyber-Physical Systems in Industry 4.0

709

The chosen use case, designed with Camunda,

aims at reducing energy consumption during the grain

drying process by optimizing the dryer parameters to

obtain the best settings (number of burners, burner

temperatures, extraction interval, …) and set points

(humidity, objectives) for the proper functioning of

the drying process with lower consumption.

The learning models aim to define the optimal

parameters with the lowest energy consumption

based on 4 inputs: input humidity, desired output

humidity, outside temperature and extraction weight.

The predicted parameters then become the input data

for estimating the energy consumption required for

the process.

4.2 Results

This section will detail the results obtained from the

models employed according to the methods

mentioned throughout the work. First, we will see the

results obtained for a complex regression problem,

namely the parameterization of the dryers (T°

burners, extraction intervals). Then, we will see the

results obtained for a non-complex classification

problem (number of burners).

4.2.1 Results for Regression Problems

After deploying different machine learning models to

forecast energy consumption and optimize

production parameters, we conclude that the deep

learning branch and ANN artificial neural networks

model provides the best performance overall. We

divided the data into training and test set. Seventy per

cent of the data was used as a training set and the

remainder as the test set. The ANN was trained over

700 epochs with a batch size equal to 20 and a

learning rate equal to 0.01. The metrics are in the

following Table 2.

Table 2: Model performance for parameter optimization.

Approach MAE R^2

ANN 0.206 0.86

As a result, we obtained an acceptable and reliable

MAE loss metric to predict the parameters. The

prediction error can be improved by implementing

other more robust outlier elimination techniques since

the most significant errors are obtained when the

fundamental variables are extreme points.

When the DDDSS Time Efficiency and Energy

Efficiency parameters are set to the ultra-mode, the

theoretical result is an average saving in the

production plants of 17.5%.

4.2.2 Results for Classification Problems

Three evaluation tests were carried evaluate the

feasibility of applying the quantum support vector

machines QSVM model to solve binary classification

problems to predict the number of burners. This same

problem was addressed using Support Vector

Machine in its classical approach, and the

performance of these two methods was compared.

The results obtained are shown in Table 3.

Table 3: QML and CML test definition.

Tests

Feature

Reduction

(PCA)

Number

features

Number of

Instances

Training

Testing

dataset

Test 1 Yes 3 100 0.7 0.3

Test 2 Yes 3 3000 0.7 0.3

Test 3 Yes 7 3000 0.7 0.3

After deploying the QSVM model, the result

obtained is not only promising for deploying quantum

infrastructure solutions, but it is already a reality, as

we can see in Table 4.

Table 4: QML and CML accuracy and running time results.

Test 1 Test 2 Test 3

Approach Accuracy

Time

(s)

Accuracy

Time

(s)

Accuracy

Time

(s)

QSVM 0.98 48.00 - - - -

CSVM 0.73 0.001 0.77 0.08 0.99 0.1

We obtained extraordinary results, as in the first

test, training the model only with a tiny part of the

dataset; we were able to obtain an accuracy of one

hundred per cent after being evaluated while its

counterpart provides less efficient performance. The

classical model must be trained with the complete

dataset to provide similar results as the quantum

model. It can be concluded that the quantum

properties speed up pattern recognition on little data

and are highly efficient compared to their traditional

counterpart.

However, when it came to testing two and three,

with more extensive input variables, the quantum

computing provided by IBM did not process it, due to

the resource limit offered to users. It is known that

today, leading companies continue to develop

quantum infrastructure with larger processing units.

The statement above positions the classical

method as the primary solution to address binary and

non-binary problems within an Industry 4.0

framework. However, the latter will be a prosperous

ICSOFT 2023 - 18th International Conference on Software Technologies

710

approach when quantum computers reach "quantum

supremacy" in the coming years.

5 CONCLUSIONS AND

PERSPECTIVES

In this paper, we have proposed a generic iPaaS

architecture fully composed of open-source solutions.

This shows that this solution can work at very low

cost even if some tasks will be a bit heavier to

manage. All technologies used could, of course, be

replaced by proprietary solutions. We could see that

the architecture allows to satisfy the requirements of

integrability, interoperability and extensibility.

To optimize the complex regression case results,

it’s essential to increase the data preprocessing

methods to achieve formidable performance for

diverse problems. Therefore, some robust techniques

will be introduced to the system for this purpose, e.g.,

data imputation using linear regression. Second, it

will be fundamental to optimize the hyperparameters

of the algorithms to obtain desired results, this last

will be possible by implementing the Grid-Search

technique.

Moreover, this work presents an alternative to the

existing options reviewed throughout state of the art,

including machine learning methods in its quantum

version to address binary classification tasks. The

latter approach was possible to deploy by using IBM

quantum resources. Moreover, the properties of

entanglement and superposition provided a speedup

to determine the number of burners needed to dry a

production batch, with exceptional accuracy and

minimal training.

This architecture allows for a simple integration

of the DDDSS which makes it adaptive and that will

clean and standardize the data and define the most

suitable decision models. The models will be able to

be evaluated, adjusted and used simultaneously to

support the decision-making process or to make it

directly while providing auditable results. The

objective is to acquire as much knowledge as possible

to compensate for the retirement of experts who are

not necessarily replaced and are becoming

increasingly rare, particularly in certain fields such as

grain drying and agriculture in general. The

collaboration of these models will bring a strong

adaptability and robustness to future CPS. The

integration of quantum decision models is also not to

be excluded in the coming years. Finally, in the

future, an evaluation of the scalability and elasticity

of the solution will be performed in a multi-tenant

scenarios context.

ACKNOWLEDGEMENT

We thank Pierre Lestage, Olivier Sourbets and the

silo managers for sharing their expertise in the field

of agriculture and more specifically in the drying

process. More generally, we also thank

MAÏSADOUR and UPPA for their financial and

institutional support in this research project.

REFERENCES

M. Hermann, T. Pentek et Otto B Design principles for

industrie 4.0 scenarios [Online]. Technische Universität

Dortmund, 2015.

DANJOU Christophe, RIVEST Louis et PELLERIN

Robert Industrie 4.0 : des pistes pour aborder l’ère du

numérique et de la connectivité [Online]. CEFRIO,

2017,

GODREUI Benjamin et SAUDEAU Emmanuelle Les

technologies de l’usine du futur au service de la

maintenance industrielle [Online]. F.F.E., 2016.

Wang, Yübo, Thilo Towara, and Reiner Anderl.

"Topological Approach for mapping technologies in

reference architectural model Industrie 4.0 (RAMI

4.0)." Proceedings of the World Congress on

Engineering and Computer Science. Vol. 2. 2017.

Zhong, R. Y., Xu, X., Klotz, E., & Newman, S. T. (2017).

Intelligent manufacturing in the context of industry 4.0:

a review. Engineering, 3(5), 616-630.

J. O. Kephart and D. M. Chess, “The vision of autonomic

computing,” Computer, vol. 36, no. 1, pp. 41–50, Jan.

2003, doi: 10.1109/MC.2003.1160055.

“RAMI 4.0 - ISA,” isa.org.

Ernesto Expósito. Semantic-Driven Architecture for

Autonomic Management of Cyber-Physical Systems

(CPS) for Industry 4.0. MEDI 2019 International

Workshops, DETECT, DSSGA, TRIDENT, Toulouse,

France, October 28–31, 2019, Proceedings, Oct 2019,

Toulouse, France. pp.5-17, ff10.1007/978-3-030-

32213-7_1ff. ffhal-02432944f.

M. Sanchez, “Autonomic process management for

Integration in Industry 4.0,” These de doctorat, Pau,

2020. [Online]. Available: https://www.theses.

fr/2020PAUU3006.

“Capella MBSE Tool - Arcadia.” https://www.eclipse.

org/capella/arcadia.html.

Roques, Pascal. "MBSE with the ARCADIA Method and

the Capella Tool." 2016. "Arcadia (engineering)".

R. H. Cestari, S. Ducos, and E. Exposito, “iPaaS in

Agriculture 4.0: An Industrial Case,” Sep. 2020, pp.

48–53, doi: 10.1109/WETICE49692.2020.00018.

ACPS: Adaptive Cyber-Physical Systems in Industry 4.0

711