Methods for Model-Driven Development of IoT Applications:

Requirements from Industrial Practice

Benjamin Nast

1 a

and Kurt Sandkuhl

1,2 b

Institute of Computer Science, Rostock University, Albert-Einstein-Str. 22, 18059 Rostock, Germany

School of Engineering, J

onk

oping University, Gjuterigatan 5, 55111 J

onk

oping, Sweden

Keywords:

Internet of Things, Model-Driven Development, Modeling Methodologies, Systematic Literature Review.

Abstract:

The Internet of Things (IoT) has become a crucial topic in research and industry over recent years. Enterprises

often fail to create business value from IoT technology because they have difﬁculties deﬁning organizational

integration. Model-driven Development (MDD) is considered an effective technique for IoT application de-

velopment. We argue that methods for MDD should comprise the organizational as well as the system de-

velopment and integration. This paper aims to provide an overview of the current state of research on MDD

of IoT applications. For this purpose, we conducted a structured literature review (SLR). A research gap was

identiﬁed as no speciﬁc research could be found on MDD of IoT applications with a focus on organizational

and system aspects. We also derived requirements from an industrial use case. The main contributions of this

paper are (a) requirements from medium-sized enterprises (SMEs) to methodical and technical IoT develop-

ment support derived from a use case, (b) the results of a systematic literature analysis in this ﬁeld, and (c) an

initial structure for the methodical support and initial architecture for the accompanying tool support.

1 INTRODUCTION

From the initial technological idea in the late 1990s

to extend the reach of the Internet also to include

physical “things”, the Internet of Things (IoT) con-

cept developed into an essential element in digital

transformation and the Internet economy (Hansen and

Bøgh, 2021). Many physical devices, like vehi-

cles, household devices, or production lines in man-

ufacturing, are equipped with sensors and control

units connected to the Internet, which facilitates IoT-

enabled services, such as remote monitoring and con-

trol, predictive maintenance, or product-as-a-service.

IoT can be deﬁned as the “seamless interplay among

dynamic communities of human users, conventional

computer systems, and smart objects, which interact

with each other and the surrounding environment”

(Fortino et al., 2015). Other technological trends,

such as cyber-physical systems (CPS) and industry

4.0 (Horvath and Gerritsen, 2012), or new business

models, such as smart connected products (Porter and

Heppelmann, 2015), data-driven services, and quanti-

ﬁed products (Sandkuhl, 2022), supported the devel-

https://orcid.org/0000-0003-4659-9840

https://orcid.org/0000-0002-7431-8412

opment of infrastructures and organizational models

related to IoT.

From a research perspective, there is a substantial

body of knowledge on technological and development

aspects of how to specify, design, and implement IoT

solutions (Alberti et al., 2019; Fortino et al., 2015).

However, our observation is that many, primarily

small and medium-sized enterprises (SMEs), hesitate

to invest in IoT efforts because they consider the de-

velopment processes too complex and have difﬁcul-

ties designing the organizational integration (Sand-

kuhl and Seigerroth, 2021). The motivation for our

work is that existing model-based approaches for IoT

solutions focus on the software and systems perspec-

tive and show a need for more integration with orga-

nizational and business model aspects. We argue that

methodologies for model-driven development (MDD)

of IoT solutions need to be better prepared for use in

SMEs and should also include organizational integra-

tion. This viewpoint is conﬁrmed by the ﬁndings of

(Fortino et al., 2021), who state that the IoT develop-

ment products, such as methodologies, frameworks,

platforms, and tools, mainly focus on the software

components of IoT systems.

The long-term objective of our work is to develop

methodical and technological support tailored to the

170

Nast, B. and Sandkuhl, K.

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice.

DOI: 10.5220/0011973500003464

In Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2023), pages 170-181

ISBN: 978-989-758-647-7; ISSN: 2184-4895

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

needs of SMEs that allows for the implementation

of IoT solutions integrating business (i.e., business

model and organizational integration) and IT aspects

(i.e., software and systems engineering). The main

contributions of this paper are (a) requirements from

SMEs to methodical and technical IoT development

support derived from a use case, (b) the results of a

systematic literature analysis in this ﬁeld, and (c) an

initial structure for the methodical support and an ini-

tial architecture for the accompanying tool support.

The paper is structured as follows: Section 2

brieﬂy introduces the research method used in our

work. Section 3 summarizes the state-of-the-art

in IoT development methodologies and speciﬁcs of

SMEs to be taken into account in methodology devel-

opment. Section 4 introduces the industrial case study

motivating our work and requirements for methodolo-

gies from an SME perspective. Section 5 presents a

literature study on method support for IoT develop-

ment. Section 6 presents the initial version of method-

ical and technical development support for SMEs.

Section 7 summarizes ﬁndings and future work.

2 RESEARCH METHOD

Research work in this paper started from the follow-

ing research question, which is based on the motiva-

tion presented in section 1: RQ: When implementing

IoT solutions in SMEs, what are the enterprises’ re-

quirements for methodical and technological support,

to what extent are these requirements covered by ex-

isting methodologies, and how should the methodi-

cal/technical support look like?

The research method used for working on this re-

search question is a combination of literature study,

descriptive case study, and argumentative-deductive

work. Since our previous work (see (Nast and Sand-

kuhl, 2021)) indicated that the speciﬁcs of SMEs were

not sufﬁciently represented in existing work and not

well understood, we decided to perform a case study

to gather information pertinent to the subject area. A

qualitative case study is an approach to research that

facilitates the exploration of a phenomenon within its

context using various data sources. This ensures that

the subject under consideration is not explored from

only one perspective but rather from various perspec-

tives, allowing for multiple facets of the phenomenon

to be revealed and understood. Within the case study,

we used three different perspectives, which at the

same time represent sources of data: we observed

the activities during IoT development, examined the

enterprise’s business model, and interviewed differ-

ent roles involved. Yin differentiates various kinds

of case studies (Yin, 2009): explanatory, exploratory,

and descriptive. The case study presented in section

4 has to be considered descriptive, as it describes the

phenomenon of process outsourcing and the real-life

context in which it occurs.

Based on the research question deﬁned and the re-

quirements visible in the case study, we started identi-

fying research areas with relevant work for this ques-

tion and analyzed the literature in these areas. The

purpose of the analysis was to ﬁnd existing theories,

methods, or technologies that guide the implementa-

tion of IoT solutions in SMEs from both IT and busi-

ness perspectives. Due to the focus on SMEs, the

speciﬁc characteristics of SMEs had to be identiﬁed

and taken into account before starting the literature

analysis (see section 3.2). Section 5 presents and dis-

cusses the results of the literature study. Based on

the case study results and the literature study results,

we propose an initial structure of the method and

an architecture for the method’s tool support. This

argumentative-deductive part of our work is the sub-

ject of section 6. The overall research project follows

the ideas and principles of design science research

(Hevner et al., 2004) with the methodical and tech-

nical support being the core artifact. In our previous

work, we investigated the need for a domain-speciﬁc

modeling language (DSML) by analyzing business

needs in a use case and existing literature in the ﬁeld

of DSML in IoT (Nast and Sandkuhl, 2021). This

conﬁrmed the relevance of research in modeling IoT

solutions and forms the starting point for this paper.

The next step is to understand the requirements for the

application of models in developing IoT applications,

i.e., model-based development of IoT, and the possi-

bility of reusing procedural or conceptual knowledge

from existing methods. To investigate the former, as

also mentioned above, we use an industrial case study

and investigate the requirements. For the latter, a lit-

erature analysis was performed that also takes into ac-

count the results of the requirement analysis from the

industrial case. As a conclusion for both steps, we

identify future tasks in method development for MDD

of IoT solutions.

3 BACKGROUND

This section describes the required theoretical back-

ground for this research approach. Section 3.1 fo-

cuses on IoT and MDD. Speciﬁcs of non-IT SMEs

in this context are described in section 3.2.

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

171

3.1 Internet of Things and MDD

The IoT has become an important topic in research

and industry, contributing to digital transformations

in many industrial domains (Hansen and Bøgh, 2021).

IoT technologies are at the center of industry 4.0 ap-

plication scenarios and support the implementation of

CPS. Many approaches are available today to support

the speciﬁcation, design, and implementation of IoT

solutions (Fortino et al., 2021) from a technological

and development perspective. However, many enter-

prises are still struggling to derive business value from

IoT technology or are reluctant to invest in IoT efforts

because of difﬁculties in deﬁning organizational inte-

gration (Sandkuhl and Seigerroth, 2021).

There have been many approaches proposed to

cope with the complexity of developing IoT solutions

and improve their portability (Fortino et al., 2015).

MDD is based on the concepts of software engineer-

ing and diagrams and is considered an effective tech-

nique for managing the complexity of IoT applica-

tions (Sosa-Reyna et al., 2018). To model a sys-

tem and describe an architecture and functionality is

more efﬁcient for developers than to describe, doc-

ument, and code every system detail in a program-

ming language (Atkinson and Kuhne, 2003). Using

MDD in this way, developers can engage with high-

level abstractions to meet their system requirements

and then automatically generate the necessary code

(Kelly and Tolvanen, 2008). Existing model-based

approaches also focus on requirements from a tech-

nical perspective and show a lack of integration with

the day-to-day processes they should be supporting.

The results of (Fortino et al., 2021) state that the IoT

development products, such as methodologies, tools,

platforms, and frameworks, mainly focus on the soft-

ware components of IoT systems, thus conﬁrming this

view.

Deﬁning a DSML to describe system require-

ments enables the use of MDD. Compared to general

modeling languages, DSMLs are easier to specify, un-

derstand, and maintain (Frank, 2013). A deﬁned syn-

tax, semantics, or both, preserves the integrity of the

models (prevents nonsensical models). Moreover, of-

ten a concrete syntax in the form of a special graphical

notation for a DSML is given that helps to improve the

clearness and understanding of the models.

3.2 Speciﬁcs of SMEs in Organizational

Innovation

The speciﬁcs of SMEs have been well-researched

when it comes to the introduction of technological in-

novations, such as the IoT. One stream of work in this

ﬁeld concerns the “readiness” of organizations for in-

novations, which aims at identifying aspects and fac-

tors affecting the introduction of innovations into an

enterprise. Based on general observations when intro-

ducing information technology by (Snyder-Halpern,

2001), various adaptations for speciﬁc ﬁelds were

proposed, such as service innovation readiness (Yen

et al., 2012) or readiness for Artiﬁcial Intelligence

ohnk et al., 2021). In this work, the factor of dig-

ital innovation readiness seems to be most suitable

to identify the speciﬁcs of SMEs that have to be ob-

served when developing MDD support tailored for

SMEs. These aspects are (Lokuge et al., 2019):

• Resource readiness describes how ﬂexible ﬁnan-

cial, technological, and human resources that are

required for digital innovation can be used in an

enterprise. Essentially, this addresses the aspect

of whether an enterprise has sufﬁcient resources.

• IT readiness describes the quality of the IT portfo-

lio to facilitate digital innovation. Here, the exis-

tence of the required IT infrastructure and acces-

sibility of digital technologies is in focus.

• Cognitive readiness concerns the knowledge base

in an organization needed for digital innovation,

i.e., are the right competencies and the right

knowledge available?

• Partnership readiness addresses the aspect of ex-

ternal stakeholders and their view on an organiza-

tion’s digital innovation. This includes partners,

suppliers, and even customers.

• Cultural readiness concerns the core values of an

organization and how they facilitate digital inno-

vation, for example, by a culture of sharing of

ideas, decentralized decision-making culture, or

risk-taking.

• Strategic readiness addresses the set of manage-

rial activities available to facilitate digital innova-

tion, i.e., are there sufﬁcient managerial and lead-

ership competencies.

We argue that a methodology for IoT introduction

should take into account the speciﬁcs of SMEs for all

these factors. SMEs are characterized by:

• Resource: compared to larger enterprises, SMEs

have less ﬂexibility in the assignment of resources

to technology innovation projects. Thus, they aim

at projects with clearly deﬁned and stable resource

requirements.

• IT: the IT situation is strictly deﬁned by organi-

zational needs. IT introduction, as such, is often

connected to or seen as innovation.

• Cognitive readiness: SMEs focus on the avail-

ability of competency and knowledge required for

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

172

their core business. Competencies in areas of in-

novation usually are present if the subject of inno-

vation is in the core area.

• Partnership: there is no general pattern of readi-

ness when it comes to customers, suppliers, or

partners of SMEs, as the readiness of these groups

highly depends on the domain and business model

of the SME.

• Cultural readiness is highly dependent on the

owners of an enterprise and the organizational cul-

ture they accept. Innovative owners tend to have a

higher openness to technological innovations.

• Strategic readiness usually is limited as SMEs

have management personnel that often has a mul-

titude of tasks in their responsibility, which re-

duces the capacity for new activities.

In this context, we explicitly exclude SMEs from the

IT sector as they usually have competencies in the

area of IoT, i.e., the focus is on non-IT SMEs.

4 INDUSTRIAL CASE: IoT FOR

ACT FACILITIES

In this section, the use case from industrial practice is

presented. First, the use case company and the appli-

cation context are described in section 4.1. The spe-

ciﬁc details and resulting requirements are described

in section 4.2.

4.1 Case Company and Application

Context

The method support work for integrating organiza-

tional aspects into IoT development methods is based

on an industrial use case from the air conditioning and

cleanroom technology (ACT) sector. Additional sen-

sors and control systems need to be integrated and

connected to a network in ACT facilities to enable

energy optimization and a basis for predictive mainte-

nance. The result is an IoT solution that forms the ba-

sis for new business services. Inspections of air han-

dling units in operation often reveal signiﬁcant dis-

crepancies from the assessed energy efﬁciency. As

the control technology becomes more automated, the

amount of data is increasing sharply. Operating the

direct and indirect processes of ACT facilities in an

energy-efﬁcient or -optimal manner needs intelligent

data processing. Therefore, the demand for sys-

tem solutions for self-recognition and self-organized

learning and the control of the systems is high.

The planned IoT solution is intended to provide

diagnostic support for possible optimizations in ACT

facilities and for the operational processes of the case

study company. Such a solution must be able to pro-

cess a large amount of information from different

sources and must be integrated into the operational

processes of the case study company to support new

types of services.

4.2 Use Case and Resulting

Requirements

In collaboration with the case study company, require-

ments for the intended system were derived. To do

this, employees of the case study company ﬁrst de-

scribed all the desired functions of the overall system.

A description has been developed for each require-

ment, along with guidelines on how to achieve the de-

sired goal. Those include required technical capabili-

ties and outputs, respectively, functions of the human

interface. Based on this information, it was then pos-

sible to identify and specify the required data for each

requirement (e.g., origin or required recording dura-

tion and measurement accuracy). The requirements

were then prioritized in terms of their importance and

feasibility for the project (see Table 1).

Table 1: Requirements from use case.

Priority Requirement

Interoperability

1 Maintenance Support

2 Detect Disruptions

2 Recognise Operating Errors

2 Energetic Inspection

3 Detect Rule Errors

4 Predictive Maintenance

4 Dynamically Adjust

Maintenance Cycles

5 Contracting

5 Facility Dimensioning

5 Commissioning

The following is a brief explanation of the require-

ments. Two of them are then described in more detail.

The overarching requirement is the Interoperabil-

ity of the overall system. This means that the imple-

mentation using new sensors or a new measuring kit

and the connection to existing building control sys-

tems must be possible. Furthermore, ﬂexibility re-

garding the data sources should be given (e.g., dif-

ferent sensor types).

Maintenance Support was cited as the most im-

portant requirement. Up to now, it has only been

possible to ﬁx acute errors during maintenance work.

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

173

Such errors, which only occur temporarily, therefore,

fall through the cracks. A historical view of existing

data should make these problems visible.

The automatic and permanent checking of com-

ponents of a facility to Detect Disruptions (e.g., a

fan fails) and to Recognise Operating Errors caused

by users (e.g., forget to reset exception schedules for

events) have the second highest priority. This en-

ables avoidable excess consumption to be identiﬁed

and thus supports the process of Energetic Inspec-

tion. This process is required by law (in Germany) for

industrial ACT facilities and is currently performed

manually by a technician on site. Here, the cur-

rent behavior of the system and the wearing parts are

checked. In some facilities, a building control system

is installed, which can sometimes provide a historical

sensor readout if conﬁgured accordingly. However,

as the employees of the use case company told us,

the available data (if any) is usually too incomplete to

evaluate the performance of the system.

The requirement Detect Rule Errors has priority

3. Incorrect regulations can cause considerable addi-

tional consumption that goes unnoticed. For exam-

ple, the air conditioner and heater may start simulta-

neously at a certain temperature. The user does not

notice this because the room temperature is satisfac-

tory.

Data obtained from the system can bring about

further changes in maintenance (priority 4). On the

one hand, it is possible to Dynamically Adjust Main-

tenance Cycles. It is better to determine maintenance

intervals by the intensity of the use of a facility rather

than ﬁxed times. On the other hand, it is possible to

detect components with deviation behavior to main-

tain or replace them before they break (Predictive

Maintenance).

Furthermore, a few requirements will play a role

in the future (priority 5). However, the technical foun-

dations for this are already to be created, at least to

some degree, as part of the current project.

Understanding avoidable excess consumption by

the system can open up new business models for the

company in the future. For example, if an oversized

fan is identiﬁed, a more suitable one can be installed.

This swap would be done free of charge, and the com-

pany would share in the electricity savings. Contract-

ing in this way is only possible without signiﬁcant risk

if a good data basis is created.

Once sufﬁcient data is available from many facil-

ities, comparisons can be made between them. This

is intended to detect, for example, the oversizing of

a facility (Facility Dimensioning). Often a facility is

designed for more people in a room or building than

are actually present.

Before a system can be used sensibly and produc-

tively, many details must be set and implemented in

the control system. Therefore, another requirement

is to simplify the process of Commissioning and thus

save appointments with the customer.

As described above, Maintenance Support is the

essential requirement of the planned system. Tem-

porary faults, which are caused, for example, by the

facility’s design, should be made visible with the help

of historical data. This results in the data being stored

historically in a database. To support this, automati-

cally generated reports and diagrams are to be output.

A typical example is that a facility heats and cools

simultaneously in strong sunlight. The last mainte-

nance appointment took place in the summer, and the

error did not occur then. With the collection of his-

torical data, this event can be detected. Automatically

generated reports and diagrams can make this under-

standable for every user. Among other things, data

about humidity, temperature, volume ﬂow, or CO

content is needed. The resolution should be one hour,

and the recording needs to include at least one spring

or autumn, better a year. If possible, permanent stor-

age is preferred. The measurement accuracy depends

on the measurement variable. Temperature (+/- 1° C),

electrical power (500W), and humidity (5%), for ex-

ample, need high accuracy. Steps of 10 Pascal, in con-

trast, are sufﬁcient for pressure.

To Detect Disruptions in individual components

of an ACT facility, these must always be checked for

threshold or experience values to trigger an alarm.

Data to be recorded for this purpose and the re-

quired measurement accuracy vary for each compo-

nent. These and similar data can also be used to iden-

tify operating errors. By comparing the data with nor-

mal values at a certain time, one can inform about this

irregular condition. The details of the data required

are very similar to those for Maintenance Support.

5 LITERATURE REVIEW ON

METHOD SUPPORT FOR IoT

DEVELOPMENT

This section addresses the question of what work has

already been done in MDD of IoT applications, focus-

ing on real-world enterprise use cases. The research

approach used, the data collection and the results are

described.

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

174

5.1 Search Strategy and Process

To systematically identify the existing research about

MDD of IoT applications, we performed a Systematic

Literature Review (SLR) according to the guidelines

of (Kitchenham, 2004). This approach was chosen

because it was developed to assess what has already

been published on a particular research topic, com-

pare existing research ﬁndings, and analyze potential

research gaps. Following (Kitchenham, 2004), our re-

search approach was carried out in six steps. We ﬁrst

formulated the general research question:

• RQ. What is the state of research on model-driven

development of IoT applications?

Next, we conducted the literature review and identi-

ﬁed relevant papers. In this way, we developed search

strings and deﬁned criteria for inclusion and exclu-

sion in the third step. After extracting and collecting

the relevant data from the selected papers in the fourth

step, we analyzed these data and summarized the re-

sults in section 5.2. Finally, we interpret and discuss

the results.

5.1.1 Development of Search Strings

Once the research question was deﬁned, we started

the literature review with an initial population and

identiﬁed ”Model-driven”, ”Enterprise” and ”IoT” as

the main keywords. Subsequently, we collected syn-

onyms and associated terms for these:

Table 2: Search terms for the SLR.

Model-driven Enterprise IoT

MDD Company

MDE Organization

MDSD Business

(DSL AND Model*)

The selection of synonyms is based on previ-

ous experience in this ﬁeld. The initial keywords

were used to develop several search strings. Syn-

onyms were then added to compare how the results

changed. In order to create the search terms, we used

the ”Scopus” database. We tested about twenty differ-

ent search strings, but we only give the most impor-

tant ones because of the page limitations. For each

string, only the titles were scanned initially. Next, we

reviewed the abstract and then the introduction and

summary. We ﬁnally read all selected papers and ap-

plied the selection criteria.

In order to complete the literature search, the

search strings (see Table 3) were also applied to the

databases ”IEEE-Xplore”, ”ACM Digital Library”,

”SpringerLink,” and ”AISeL”. We adapted to the syn-

tax of the respective database. Within the ”Springer-

Link” database, we also used the corresponding Ger-

man translations. Since no other suitable work was

found, we decided to retest a version of the search

term with additional keywords to see if this had a dif-

ferent effect on the search results than in ”Scopus”.

Even so, we did not ﬁnd any other works with the

search term.

The literature review was performed between Au-

gust and October 2022.

Table 3: Search strings.

Search Number of Results/

String Identiﬁed Papers

TITLE-ABS- Number of Results: 70

KEY Identiﬁed Papers: 6

((”model driven”

OR ”mdd” OR

”mde” OR ”mdsd” (Corradini et al., 2022)

OR (”dsl” AND (Moin et al., 2022)

”model*”)) AND (Nast and Sandkuhl, 2021)

(”enterprise” OR (Michael et al., 2019)

”company” OR (Brambilla et al., 2017)

”business” OR (Khaleel et al., 2017)

”organization”)

AND (”iot”))

((((”Full Text & Number of Results: 180

Metadata”:”model Identiﬁed Papers: 2

driven” OR ”mdd”

OR ”mde” OR

”mdsd”) OR (”dsl” (Ferreira et al., 2018)

AND ”model*”)) (Khaleel et al., 2017)

AND ”iot”

AND (”Full

Text & Metadata”:

”enterprise” OR

”company” OR

”business” OR

”organization”)))

5.1.2 Selection of the Papers

For the selection of relevant papers to answer the re-

search question, all abstracts had to be read and the

inclusion and exclusion criteria (see Table 4) were ap-

plied. We consider a paper relevant if the authors de-

scribe a model-based development approach of an IoT

application. Moreover, the included approach must be

the center of the paper. Papers in which the authors

only mention that those approaches can be helpful for

certain questions are not relevant to us. Work that is

not based on a real company application or has not

been evaluated in a company is not considered. We

also excluded papers in the context of healthcare, as

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

175

the requirements are different from ordinary compa-

nies.

Table 4: Inclusion and exclusion criteria.

Inclusion Criteria

A model-based development

approach of an IoT application is described.

The included approach must be the center of the

work (not only mentioned).

Exclusion Criteria

Work that is not based on or has not been

evaluated in a real company is not considered.

Work in the context of healthcare is excluded.

Before reading the abstracts, we excluded any

items we found that were no papers. In ”Scopus”,

14 of the 70 items had to be ﬁltered out because they

are tables of contents, summaries of books, and so

on. Two of them were included twice in the results.

We then read the abstracts of 54 remaining papers and

identiﬁed only six relevant papers according to our

criteria. If the situation was unclear, we read the en-

tire text to decide whether the text was relevant or not.

The second string from Table 3 was applied in ”IEEE

XPLORE” (180 results) and added one new relevant

paper, resulting in a total of seven relevant texts iden-

tiﬁed.

5.1.3 Summary of the Search Process

We started this SLR with the population. As

a ﬁrst step, we applied an initial search string

to the previously deﬁned literature sources (”IEEE

Xplore”, ”ACM Digital Library”, ”SpringerLink” and

”AISeL”).

By adding synonyms to reﬁne the search string,

a broader range of research papers could be covered,

building the basis for identifying relevant papers. We

established criteria to determine which papers were

relevant to answering our research questions. Before

we started identifying relevant papers, we excluded

articles that are, for example, tables of contents, sum-

maries of books, or that appeared more than once in

the results. After reading the abstracts, seven relevant

papers remained, which are analyzed in section 5.2.

5.2 Search Results

In this section, we answer the research question from

section 5.1 using the collected data from the identiﬁed

relevant papers. First, general information is provided

on the research topics, research approaches used, ac-

tive researchers in the ﬁeld studied, and the overall

activity. Then, the approaches presented in each case

are summarized to provide an overview of the current

state of research. It continues to examine the extent to

which the approaches considered organizational as-

pects in the development of IoT solutions and what

is written about research gaps or problems and criti-

cism. A table was ﬁrst created in Excel to collect the

relevant data from each paper.

5.2.1 Research Activity and Topics

Since our own experience leading up to the SLR has

already shown that general MDD approaches tend to

focus on technical aspects of IoT solution develop-

ment, so we did not expect many contributions. Thus,

it is not surprising that we found only seven contri-

butions in which the organizational requirements of a

real enterprise are considered during development.

We can see in Table 3 that three of the seven rele-

vant papers were written in the last two years (2021-

2022). Two were written in 2017, and one each in

2018 and 2019. The seven papers are written by 36

authors, and none appeared more than once. They

described approaches with different emphases and at

different stages of progress. Four papers are written

by authors of the same university (Corradini et al.,

2022), (Moin et al., 2022), (Nast and Sandkuhl, 2021)

and (Ferreira et al., 2018). The other papers were

written in collaboration between two or more univer-

sities and different types of organizations or compa-

nies.

In summary, it seems that there is no research

group which working on this topic for years. This

could be because this topic is very practice-oriented

and therefore requires applied research rather than ba-

sic research projects.

In none of the selected papers is a holistic platform

proposed. The use of tools in combination with their

own meta-models or modeling languages or a combi-

nation of different tools is always described.

5.2.2 Approaches of MDD

(Corradini et al., 2022) propose an approach called

FloWare that organizes the modeling and develop-

ment of IoT applications into distinct steps, man-

ages complexity in representing IoT variability, and

enables reusability of design decisions and artifacts.

This provides modeling support through functional

models to fully represent and handle the potential

variability of devices in a given IoT application do-

main. Once a particular conﬁguration is selected, it

is augmented with speciﬁc information to automati-

cally derive fragments of IoT applications that the de-

veloper within a low-code development environment

successively combines. FloWare is fully supported by

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

176

a toolchain that has been released for public use. A

tool was designed to create functional models and a

platform including templates that automatically gen-

erate the basic application logic for interacting with

the selected devices, helping the developer to develop

more complex applications. Requirements from a real

Smart Campus scenario were taken into account dur-

ing development.

An open-source research prototype called ML-

Square is presented by (Moin et al., 2022). It is based

on the Eclipse Modeling Framework and the state-

of-the-art in model-driven software engineering liter-

ature for intelligent CPS and the IoT. The intended

users are mainly software developers who may not

have deep knowledge and skills related to heteroge-

neous IoT platforms and various artiﬁcial intelligence

technologies, especially machine learning. It pro-

vides a domain-speciﬁc modeling methodology for its

users. Therefore, it enables the automatic generation

of source code from the models for the entire soft-

ware solution of IoT or CPS. The users gain access to

the APIs of the libraries and frameworks of machine

learning at the modeling level. Using the enhanced

software models in ML-Quadrat thus enables the pro-

vided model-to-code transformations not only to pro-

duce the source code but also the ML models for the

target IoT solution and possibly re-train them by ob-

serving new data samples later. The approach focuses

on Smart IoT services and was initially validated

through requirements from four computer science ex-

perts (two from industry and two from academia) with

different levels of skills in, for example, IoT or Ma-

chine Learning. The extended version of this ap-

proach, called DriotData, is currently under develop-

ment and is intended to be offered as a subscription-

based service for SMEs. This will include model ed-

itors for domain experts without knowledge or skills

in the IT area.

The approach described by (Nast and Sandkuhl,

2021), which is integrated with enterprise modeling

techniques, considers organizational and technical as-

pects of IoT application development. A method

component for IoT modeling is added to the enter-

prise modeling language 4EM. The overall objective

is methodical and technical support for model-based

development of IoT solutions, focusing on organiza-

tional integration into business processes, organiza-

tion structures, and business models. The creation of

a model is intended to conﬁgure the IoT application

and thus provide a basis for data processing without

requiring IT skills from the user. The stage of this ap-

proach described in the paper is limited to the model-

ing phase. It contains a meta-model and a DSML for

ACT facilities, which were derived based on require-

ments from an actual use case and interviews with do-

main experts. Upcoming work needs to extend this to

multiple cases and explore the use of the developed

models to conﬁgure IoT platforms to be used during

the operation of the enterprise IoT solution.

(Michael et al., 2019) discuss a way to create IoT

systems using a model-based approach to support pri-

vacy and data transparency. The relevance and appli-

cation are shown on a use case from industrial pro-

duction processes. Tools and frameworks in connec-

tion with a set of domain-speciﬁc languages (DSLs)

are used. A possible DSL model structure, includ-

ing domain models, a privacy model, and possible in-

stantiations, as well as relevant aspects that must be

considered for the system design, are shown. The ap-

proach was validated in a real-world production line

use case. The realization makes use of two different

MDE tools. First, a modeling language development

workbench is used that supports agile and composi-

tional development of DSLs. The second tool is a

generator for enterprise management, which is based

on the ﬁrst tool and uses a set of models. These are

parsed using a templated engine and transformed to-

wards the target, namely output ﬁles in the target lan-

guage. As a result, an information system is created

of class diagrams and graphical interface models.

The work of (Ferreira et al., 2018) deals with

the complexity of IoT application development for

integrating CPS modules into SMEs manufacturing

processes. For this purpose, an architecture-based

process modeling and simulation of the different

phases of existing SME factory production is pro-

posed. This enables real-time information through

IoT data collection to feed different production im-

provement mechanisms (e.g., planning, scheduling,

and monitoring). This will be done by providing

data on the components in the current environment of

SMEs to improve the performance of their production

processes. The solution is divided into design time

and runtime. At design time, user knowledge leads to

a representation of the processes in the factory. There

is a tool for creating a model that does not require

the permanent support of a specialist. The resulting

BPMN diagram is then executed by another tool (af-

ter being reﬁned there). An IoT device deployed on

a machine enables the coordination of hardware and

software components. Information such as the time

of completion activities, the number of products man-

ufactured, wasted products, and production or setup

times for the machines can be given in real-time. To

avoid any complication or waste of time in the pro-

duction process, an additional tool was developed to

be used by the factory workers to check the produc-

tion schedule or detect component errors. The exper-

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

177

iment prototype was validated in a polishing process

as part of a real company’s production process of cut-

lery.

(Brambilla et al., 2017) focus on requirements and

usage scenarios that cover the front-end aspects of IoT

systems. The proposed model-driven approach to de-

sign interfaces includes deﬁning speciﬁc components

and design patterns using a visual modeling language

for IoT applications and describing an implementa-

tion of the solution that includes automatic code gen-

eration from the models. It contains the deﬁnition of

the main domain-speciﬁc concepts for IoT and typ-

ical use cases, a visual modeling language focusing

on user interaction aspects, and a set of design prac-

tices that increase productivity and simplify the de-

sign of IoT solutions. Moreover, tools for the design,

deployment, and execution phases of IoT applications

are implemented. The validation of the approach in

three different industrial use cases in cooperation with

a company is described.

The Enabling Business-based Internet of Things

and Services (ebbits) platform with a focus on the

industrial domain is presented from (Khaleel et al.,

2017). Heterogeneous applications were deployed

and tested, such as a wireless sensor and actuator

network for monitoring industrial machinery and an

RFID-based system for operator management, locat-

ing, and authorization. The latter includes an inter-

active user interface for wearable devices to visual-

ize real-time information from devices in the physical

world. To simplify the process of creating IoT appli-

cations, tools for MDD are used. These developments

are based on an IoT middleware designed to enable

the seamless integration of heterogeneous technolo-

gies and processes into traditional enterprise systems.

This paper presents the deployment of the prototype

in the automotive industry.

5.2.3 Interpretation of the Results

This section completes the SLR and addresses the in-

terpretation of the results. To provide an overview of

the current state of the research on MDD of IoT ap-

plications, the identiﬁed papers are examined in the

context of the overarching research question.

The results showed that there is not much research

on MDD of IoT applications based on requirements

from real-world use cases from enterprises. A small

number of papers were identiﬁed. However, that ad-

dressed the topic in a variety of ways based on the

inclusion criteria selected. Only (Corradini et al.,

2022), (Nast and Sandkuhl, 2021), and (Ferreira et al.,

2018) combined technological and organizational as-

pects in the development process. The rest of the pa-

pers validated their approaches after development in

real-world use cases in cooperation with companies.

In fact, the results showed that there are already

many different approaches to handling the complex-

ity of IoT application development. Tools containing

meta-models and a graphical modeling language are

often used in MDD.

Not much is written in the selected papers about

further research or problems in this area. (Corradini

et al., 2022), (Moin et al., 2022), (Nast and Sandkuhl,

2021), (Michael et al., 2019), and (Brambilla et al.,

2017) intend to apply their described approaches in

different domains to validate them. A need to improve

the application logic and functions and to achieve

user-friendliness and thus higher acceptance is identi-

ﬁed by (Corradini et al., 2022), (Ferreira et al., 2018),

and (Brambilla et al., 2017). (Khaleel et al., 2017)

merely summarizes the results and does not provide

an outlook on further action.

The work reviewed in this chapter provides

promising approaches that focus on the technological

aspects of model-driven IoT application development

(at different maturity levels). Usability plays an im-

portant role in all approaches, to a greater or lesser

extent. (Moin et al., 2022) and (Nast and Sandkuhl,

2021) even explicitly target users without special IT

skills and thus non-IT SMEs. In summary, there is no

validated method that combines technological and or-

ganizational aspects in all phases of the development

process, and there is no speciﬁc support for SMEs.

6 TOWARDS

METHODICAL/TECHNICAL

SUPPORT

After describing the requirements for the overall ap-

plication resulting from the use case in section 4.2,

section 6.1 now presents the requirements that re-

late directly to the methodological/technical support.

Section 6.2 presents the initial version of methodi-

cal/technical development support for SMEs.

6.1 Requirements

ACT facilities for the industry are not mass-produced

but designed individually for each customer. In these

systems, many components have to work together,

and we need a variety of sensors to understand what

is going on inside them. However, the signals from

these sensors are not readily interpretable. Often, they

only measure an analog signal of 0 to 10 volts, which

must be converted into a physical unit of measure-

ment. To make sense of these measurement points

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

178

and understand the performance of the system, they

must then be processed and put into context. As the

conﬁgurations of these air conditioning systems vary,

so does the arrangement of the measurements. Only

some systems have a cooling unit, and some use it

only for cooling, while others also use it for dehu-

midifying. These heterogeneous structures also have

an impact on the IT architecture. On the one hand,

the systems are heterogeneous in their conﬁgurations,

and the data must be preprocessed at great expense

to be usable. On the other hand, there is great poten-

tial in comparing systems with similar conﬁgurations.

At the same time, the implemented system must be

usable by climate experts. Emphasis is placed on a

system that does not require a lot of IoT and analysis-

speciﬁc knowledge but is usable by the end user.

While the requirements described in section 4.2

relate primarily to the data and, thus, to the over-

all system, it was possible to derive further require-

ments in the various development phases. The use of a

method for the development of IoT applications must

be usable without requiring knowledge of IT or data

processing. For this purpose, appropriate interfaces

must be provided for conﬁguration and later also for

operation and maintenance. Therefore, for conﬁgu-

ration, options are created that require only domain-

speciﬁc knowledge. In addition, a kind of vocabulary

for the designation of sensors or their IDs and clear

deﬁnitions of the conﬁguration data (unit of measure-

ment, data format, or range of values) must be de-

ﬁned. Another requirement is that ACT facilities can

be visualized. On the one hand, this helps ACT tech-

nicians in designing the solution. On the other hand,

these visualizations can be used for presentations to

customers to explain repair measures or the replace-

ment of components in an understandable way. The

monitoring of ACT systems must be made possible,

for example, through alarms or comprehensible dia-

grams and data evaluations.

6.2 Initial Method Structure and Tool

Architecture

We developed a tool that includes a meta-model and a

DSML for ACT facilities to support the ACT techni-

cians in designing the IoT solution. The tool was de-

veloped using the ADOxx meta-modeling platform.

This is a development and conﬁguration platform for

the implementation of modeling methods. Using this

platform, it is possible to realize a full-ﬂedged mod-

eling software that contains procedures and function-

alities in the form of mechanisms and algorithms in

addition to the modeling language.

The modeling language for ACT facilities is based

on the requirements of the case study. An example

model of a ﬁctitious facility containing some of the

modeling objects is shown in Figure 2.

The classes of the meta-model represent the possi-

ble components in such facilities. The attributes con-

tain, e.g., information about the type of the compo-

nents or the unique identiﬁers of the sensors. The

symbols are based on the European standard DIN EN

12792 and the knowledge of the domain experts in-

volved in the project. The graphical representation, as

well as the designation of the components, depends on

the entered attributes. Arrows show the logical rela-

tionship between the components and the airﬂow di-

rection. Sensors are represented with a circle and a

line to the component or location in the model. The

letters in the circle indicate the sensor type. An exam-

ple model of a real facility (see Figure 1) shows some

of the objects in the modeling language and contains

eight sensors: two each for controlling the supply and

exhaust air (temperature in °C and CO

content in

ppm). Furthermore, the two air ﬁlters are equipped

with a sensor for the pressure difference (unit of mea-

surement: Pascal), and the fans with a sensor for the

volume ﬂow (unit of measurement: m³/h).

Figure 1: Model of the example facility.

The creation of a new facility in the tool is started

with the input of conﬁguration data. Required infor-

mation about the location, the operator and owner,

and the components of a facility, are requested via

dialog boxes. The input of these values automati-

cally generates the corresponding modeling objects,

including the attribute values (e.g., fan size). Unique

IDs are assigned to the sensors, depending on their

type and position.

The conﬁguration data and the sensors’ IDs are

converted into a JSON ﬁle and sent to a server appli-

cation using an HTTP request (see Figure 3). This

information can also be exported as documentation of

the facility and used for presentation or internal pur-

poses of the case study company. For instance, this

documentation can help the customer understand his

facility, the results of energetic inspections, or even

the need for renewal and repairs.

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

179

Figure 2: Objects of the DSML.

The developed tool allows the technician to con-

ﬁgure the installed sensors. Parts of the physical con-

ﬁguration are covered by collecting hardware-related

sensor information and mappings to measured as-

pects. The conﬁguration enables the data processing

of the logical part by indicating what type of ACT fa-

cility is present. The user of the tool does not need

speciﬁc knowledge or skills in IoT and data analysis.

Figure 3: Architecture of the overall system.

7 SUMMARY AND FUTURE

WORK

Starting from requirements to methodical/technical

support for IoT implementation in an industrial use

case, this paper investigated how SMEs can be sup-

ported in developing and introducing IoT solutions

that integrate business (i.e., business model and orga-

nizational integration) and IT aspects (i.e., software

and systems engineering). The initial structure for

method support and a ﬁrst architecture of tool support

for this method was proposed. The work was also an-

chored on a systematic literature analysis in the ﬁeld.

The successful application of the method and tool

support in the case company shows that it meets

the requirements of at least this speciﬁc SME. The

method and tool support also address the situation

in SMEs when it comes to resource, cognitive, and

IT readiness: clearly deﬁned method steps and free-

of-charge tool support contribute to low resource re-

quirements. The modeling tool and the possibility

to generate part of the conﬁguration reduce the need

for special employee training and lower the cognitive

readiness threshold. The use of standardized compo-

nents minimizes IT readiness expectations as much as

reasonable. The other readiness aspects require more

investigation in future work.

The main limitation of our work is that we base

the requirements for the method support on only one

use case. This poses the threat of having tailored the

method proposal to the demand of the use case en-

terprise only. However, the results of the literature

analysis and similarities to established methods give

reasons to believe that the method might be transfer-

able.

The future work will have conceptual and empiri-

cal nature. From a conceptual perspective, a more de-

tailed description of the method and architecture and

the extension of the tool support is required. From

an empirical perspective, more case studies are neces-

sary to ensure validity. Furthermore, evaluation with

stakeholders in SMEs is needed.

ACKNOWLEDGEMENTS

Part of the research presented in this paper was sup-

ported by grant no. TBI-V-1-426-VBW-145 of the

Ministry of Economics, Infrastructure, Tourism and

Labour of the State of Mecklenburg-Vorpommern us-

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

180

ing funds from the European Regional Development

Fund.

REFERENCES

Alberti, A. M., Santos, M. A. S., Souza, R., Da Silva, H.

D. L., Carneiro, J. R., Figueiredo, V. A. C., and Ro-

drigues, J. J. P. C. (2019). Platforms for smart envi-

ronments and future internet design: A survey. IEEE

Access, 7:165748–165778.

Atkinson, C. and Kuhne, T. (2003). Model-driven develop-

ment: A metamodeling foundation. IEEE Software,

20(5):36–41.

Brambilla, M., Umuhoza, E., and Acerbis, R. (2017).

Model-driven development of user interfaces for iot

systems via domain-speciﬁc components and patterns.

Journal of Internet Services and Applications, 8(1).

Corradini, F., Fedeli, A., Fornari, F., Polini, A., and Re, B.

(2022). Floware: A model-driven approach fostering

reuse and customisation in iot applications modelling

and development. Software and Systems Modeling,

pages 1–28.

Ferreira, J., Lopes, F., Ghimire, S., Doumeingts, G.,

Agostinho, C., and Jardim-Goncalves, R. (2018).

Cyber-physical production systems to monitor the

polishing process of cutlery production. In 2018 Inter-

national Conference on Intelligent Systems (IS), pages

926–933. IEEE.

Fortino, G., Guerrieri, A., Russo, W., and Savaglio, C.

(2015). Towards a development methodology for

smart object-oriented iot systems: A metamodel ap-

proach. In 2015 IEEE International Conference on

Systems, Man, and Cybernetics, pages 1297–1302.

IEEE.

Fortino, G., Savaglio, C., Spezzano, G., and Zhou, M.

(2021). Internet of things as system of systems: A

review of methodologies, frameworks, platforms, and

tools. IEEE Transactions on Systems, Man, and Cy-

bernetics: Systems, 51(1):223–236.

Frank, U. (2013). Domain-speciﬁc modeling languages:

Requirements analysis and design guidelines. In Do-

main Engineering, pages 133–157. Springer.

Hansen, E. B. and Bøgh, S. (2021). Artiﬁcial intelligence

and internet of things in small and medium-sized en-

terprises: A survey. Journal of Manufacturing Sys-

tems, 58:362–372.

Hevner, A. R., March, S. T., Park, J., and Ram, S. (2004).

Design science in information systems research. MIS

Quarterly, pages 75–105.

Horvath, I. and Gerritsen, B. H. (2012). Cyber-physical

systems: Concepts, technologies and implementation

principles. In Proceedings of TMCE, volume 1 (2),

pages 7–11.

ohnk, J., Weißert, M., and Wyrtki, K. (2021). Ready or

not, ai comes— an interview study of organizational

ai readiness factors. Business & Information Systems

Engineering, 63(1):5–20.

Kelly, S. and Tolvanen, J.-P. (2008). Domain-speciﬁc mod-

eling: Enabling full code generation. John Wiley &

Sons.

Khaleel, H., Conzon, D., Kasinathan, P., Brizzi, P., Pas-

trone, C., Pramudianto, F., Eisenhauer, M., Cultrona,

P., Rusina, F., Lukac, G., and Paralic, M. (2017).

Heterogeneous applications, tools, and methodologies

in the car manufacturing industry through an iot ap-

proach. IEEE Systems Journal, 11(3):1412–1423.

Kitchenham, B. (2004). Procedures for performing sys-

tematic reviews. Keele, UK, Keele University,

33(2004):1–26.

Lokuge, S., Sedera, D., Grover, V., and Dongming, X.

(2019). Organizational readiness for digital innova-

tion: Development and empirical calibration of a con-

struct. Information & Management, 56(3):445–461.

Michael, J., Netz, L., Rumpe, B., and Varga, S. (2019).

Towards privacy-preserving iot systems using model

driven engineering. In MDE4IoT/ModComp@ MoD-

ELS, pages 15–22.

Moin, A., Mituca, A., Challenger, M., Badii, A., and

unnemann, S. (2022). Ml-quadrat & driotdata: A

model-driven engineering tool and a low-code plat-

form for smart iot services. In Proceedings of the

ACM/IEEE 44th International Conference on Soft-

ware Engineering: Companion Proceedings, pages

144–148.

Nast, B. and Sandkuhl, K. (2021). Meta-model and tool

support for the organizational aspects of internet-of-

things development methods: Organizational aspects

of iot development methods. In 2021 3rd Interna-

tional Conference on Advanced Information Science

and System (AISS 2021), pages 1–6.

Porter, M. E. and Heppelmann, J. E. (2015). How smart,

connected products are transforming companies. Har-

vard Business Review, 93(10):96–114.

Sandkuhl, K. (2022). Features of quantiﬁed products and

their design implications. In International Baltic Con-

ference on Digital Business and Intelligent Systems,

pages 152–163. Springer.

Sandkuhl, K. and Seigerroth, U. (2021). Digital transfor-

mation of enterprises: Case studies and transforma-

tion paths. In Twenty-ﬁfth Paciﬁc Asia Conference on

Information Systems, Dubai, UAE, 2021.

Snyder-Halpern, R. (2001). Indicators of organizational

readiness for clinical information technology/systems

innovation: A delphi study. International Journal of

Medical Informatics, 63(3):179–204.

Sosa-Reyna, C. M., Tello-Leal, E., and Lara-Alabazares, D.

(2018). Methodology for the model-driven develop-

ment of service oriented iot applications. Journal of

Systems Architecture, 90:15–22.

Yen, H. R., Wang, W., Wei, C.-P., Hsu, S. H.-Y., and Chiu,

H.-C. (2012). Service innovation readiness: Dimen-

sions and performance outcome. Decision Support

Systems, 53(4):813–824.

Yin, R. K. (2009). Case study research: Design and meth-

ods, volume 5. SAGE.

Methods for Model-Driven Development of IoT Applications: Requirements from Industrial Practice

181