ArticlePDF Available

Augmented reality technology in the manufacturing industry: A review of the last decade


Abstract and Figures

The aim of this article is to analyze and review the scientific literature relating to the application of Augmented Reality (AR) technology in industry. AR technology is becoming increasingly diffuse, due to the ease of application development and the widespread use of hardware devices (mainly smartphones and tablets) able to support its adoption. Today, a growing number of applications based on AR solutions are being developed for industrial purposes. Although these applications are often little more than experimental prototypes, AR technology is proving highly flexible and is showing great potential in numerous areas (e.g., maintenance, training/learning, assembly or product design) and in industrial sectors (e.g., the automotive, aircraft or manufacturing industries). It is expected that AR systems will become even more widespread in the near future. The purpose of this review is to classify the literature on AR published from 2006 to early 2017, to identify the main areas and sectors where AR is currently deployed, describe the technological solutions adopted, as well as the main benefits achievable with this kind of technology.
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
IISE Transactions
ISSN: 2472-5854 (Print) 2472-5862 (Online) Journal homepage:
Augmented reality technology in the
manufacturing industry: A review of the last
Eleonora Bottani & Giuseppe Vignali
To cite this article: Eleonora Bottani & Giuseppe Vignali (2019) Augmented reality technology in
the manufacturing industry: A review of the last decade, IISE Transactions, 51:3, 284-310, DOI:
To link to this article:
© 2019 The Author(s). Published with
license by Taylor & Francis, LLC.
Accepted author version posted online: 03
Jul 2018.
Published online: 22 Feb 2019.
Submit your article to this journal
Article views: 276
View Crossmark data
Augmented reality technology in the manufacturing industry: A review of the
last decade
Eleonora Bottani and Giuseppe Vignali
Department of Engineering and Architecture, University of Parma, Parma, Italy
The aim of this article is to analyze and review the scientific literature relating to the application
of Augmented Reality (AR) technology in industry. AR technology is becoming increasingly diffuse,
due to the ease of application development and the widespread use of hardware devices (mainly
smartphones and tablets) able to support its adoption. Today, a growing number of applications
based on AR solutions are being developed for industrial purposes. Although these applications
are often little more than experimental prototypes, AR technology is proving highly flexible and is
showing great potential in numerous areas (e.g., maintenance, training/learning, assembly or prod-
uct design) and in industrial sectors (e.g., the automotive, aircraft or manufacturing industries). It
is expected that AR systems will become even more widespread in the near future.
The purpose of this review is to classify the literature on AR published from 2006 to early 2017,
to identify the main areas and sectors where AR is currently deployed, describe the technological
solutions adopted, as well as the main benefits achievable with this kind of technology.
Received 6 October 2016
Accepted 18 June 2018
Augmented reality (AR);
manufacturing industry;
literature review
1. Introduction and background on AR systems
Augmented Reality (AR) is a term used to identify a set of
technologies that allows the view of real world environment
to be augmentedby computer-generated elements or
objects (Van Krevelen and Poelman, 2010). More specific-
ally, AR describes a mediated reality, where the visual per-
ception of the physical real-world environment is enhanced
by means of computing devices. Compared with Virtual
Reality (VR), i.e., a set of technologies that allow the user to
interact with a computer in a simulated environment (either
a simulation of the real world or an imaginary world, Khan
et al.(2011)), AR does not aim to replace the real world
with a simulated one and is consequently often classified as
a Mixed Reality (MR) system. MR is a mix of reality and
VR, encompassing both AR and augmented virtuality, via
immersive technology (Milgram and Kishino, 1994).
The first AR prototypes, created by computer graphics
pioneer Ivan Sutherland and his students at Harvard
University and the University of Utah, appeared in the late-
1960s and exploited a see-through display fitted to a helmet
to present 3D graphics (Tamura, 2002). During the 1970s
and 1980s, mobile devices such as the Sony Walkman
(1979), digital watches and personal digital organizers were
introduced. This paved the way for wearable computing in
the 1990s as personal computers became small enough to be
worn at all times (Van Krevelen and Poelman 2010). It was
only in the early-1990s that the term augmented realitywas
coined by Caudell and Mizell (1992), both scientists at the
Boeing Corporation, who developed an experimental AR sys-
tem to help workers put together wiring harnesses. Since the
late-1990s, AR has been a specific field of research, as dem-
onstrated by the fact that several conferences on AR have
been held (e.g., the International Workshop and Symposium
on AR and the International Symposium on MR). By classify-
ing the literature published in the 1990s, Azuma (1997)
found six main classes of potential AR applications, such as
medicine, maintenance/repair, annotation, robotics, entertain-
ment, and military settings; in an update of that review, new
areas were identified, namely outdoor, mobile AR and collab-
orative AR (Azuma et al., 2001). Recently, Johnson et al.
(2011) predicted that AR technologies would have emerged
more fully within the next 2 to 3 years; learning, education,
and entertainment were identified among the most promising
areas of AR application. Georgel (2011) coined the term
Industrial Augmented Reality(IAR) to describe the use of
AR to support an industrial process and identified product
design, manufacturing, assembly, maintenance/inspection,
and training as the key areas for IAR applications.
The success of any emerging technology is typically the
result of various aspects, including social and technical
issues (Liao, 2016). From a technical point of view, since the
late-1990s it has become much easier to develop AR applica-
tions rapidly, thanks to freely available software toolkits
CONTACT Eleonora Bottani
Color versions of one or more of the figures in the article can be found online at
ß2019 The Author(s). Published with license by Taylor & Francis, LLC.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (,
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in
any way.
2019, VOL. 51, NO. 3, 284310
(e.g., the ARToolKit) and the development of camera sys-
tems able to analyze the physical environment in real time
and relate positions between objects and environment in
that period. This type of camera system now represents the
basis for the integration of virtual objects with reality in AR
systems. The smart phone adoption and the proliferation of
AR browser applications are further aspects that contribute
to the diffusion of AR solutions (Johnson et al., 2011). From
a social perspective, usually AR is not perceived as a tool to
replace human workers, but rather to help them; this may
do well also against social issues relating to its industrial
deployment (Azuma, 1997). As a result, today AR solutions
are increasingly diffuse in industrial contexts, where they are
developed for different ends (Nee et al., 2012). Personal
information or personal assistance applications have been
developed on the basis of AR systems (H
ollerer and Feiner,
2004). Other applications are reported in the tourism sector:
Vlahakis et al.(2001) developed ARCHEOGUIDE, which
reconstructs a cultural heritage site in Olympia, Greece.
Other areas interested in AR applications include the mili-
tary (Henderson and Feiner, 2009; van Krevelen and
Poelman, 2010), automotive (Regenbrecht et al., 2005), and
medical sectors (Vogt et al., 2006). Construction is the
industry with the highest number of AR applications
(Behzadan et al., 2015). In this context, AR applications are
well established and used for many purposes, including
building damage reconnaissance, georeferenced visualization
of underground utilities, collaborative design and informa-
tion delivery.
An emerging area of AR applications is the manufactur-
ing industry, which is concerned with the process of trans-
forming raw materials into finished goods with added value.
Due to their complex internal processes and the increasing
globalization of supply chains, manufacturing companies
need real-time information exchanges at the various stages
of the product lifecycle, i.e., design, prototyping, production,
assembly, maintenance/repair, etc. In this scenario, AR can
be of great help, due to its capability to simulate, assist and
improve its processes before they are carried out (Ong et al.,
2008). Indeed, the virtual objects display information that
the user cannot directly detect with his own senses; the
information conveyed by the virtual objects may help a user
perform most of the product-related tasks (Azuma, 1997).
AR applications to the manufacturing industry have been
developed for several purposes, including process monitor-
ing and control, real-time evaluation of plant layout, plant
and machinery maintenance, plant and building construc-
tion, as well as for enhancing industrial safety (Georgel,
2011). Moreover, authors argue that the future of mobile
technology promises to revolutionize AR mobile applications
in the manufacturing industry (Carmigniani et al., 2011).
Despite this, AR applications in manufacturing are perceived
to be still in an exploratory stage (Ong et al., 2008) and full-
scale deployments of IAR solutions have been carried out
only in a limited number of cases (Georgel, 2011).
On the basis of these premises, this study proposes a lit-
erature review of recent AR studies in industry, expressly
focusing on those which may help to carry out operations
typical of the manufacturing industry (e.g., assembly, main-
tenance or facility management). Besides the fact that AR
applications are increasingly used within the manufacturing
industry, we decided to target this context as it is one of
Italys most prominent industrial sectors. Italy is currently
the fifth largest manufacturing producer in the world
(OECD, 2013). In our analysis of the manufacturing indus-
try, we have not focused on the other industrial sectors, and
in particular the construction industry, despite the wide use
of AR systems in this context, as it has been analyzed in the
recent review by Behzadan et al. (2015). The overall aim of
this study is to categorize the recent literature, examine the
state-of-the-art solutions and highlight the key benefits of
AR technology within the manufacturing industry.
The remainder of this article is organized as follows.
Section 2 details the research methodology adopted for the
literature survey and presents some preliminary information
about the studies analyzed. Section 3 details the survey
results and includes not only descriptive statistics on the sam-
ple of papers reviewed, but also their categorization and
their detailed analysis. Section 4 summarizes the key
findings from the review, discusses the related scientific
and practical implications, and indicates potential
future research.
2. Methodology
2.1. Survey methodology
The research methodology chosen for this study is system-
atic literature review (Tranfield et al., 2003). This kind of
review requires two steps (Alderson et al., 2004). First, the
inclusion criteria have to be clearly specified, with the pur-
pose of correctly selecting the studies to be reviewed. To
this end, we have only included in the review studies:
1. in English and published in peer-reviewed inter-
national journals;
2. that expressly focus on AR (as opposed to VR) solutions
in the manufacturing industry.
Because of a previous review study (Ong et al., 2008) that
included papers published up to 2005, the publication time-
span was limited to be between 2006 and 2017.
The first step of a systematic literature review is to define
the strategy of locating and selecting the studies. In this art-
icle, a computerized search was made using three different
databases: Scopus (, Web of Science
(WOS) (, and Ebsco
( to identify pertinent studies.
The search was performed following the steps shown in
Figure 1.
As can be seen from Figure 1, we began by making a
search query on Scopus, WOS, and Ebsco databases with the
general keywords or topics augmented realityþmachine
or equipment,”“manufacturing,”“maintenance,”“safety,
risk,”“emergency,”“hazard,”“assembly. The search, which
was carried out in its final version between December 2016
and February 2017, returned a total of 1285 papers, which
were reduced to 848 after merging the results from the data-
bases and excluding duplicates. In this phase, we also
excluded articles that turned out to be preliminary confer-
ence papers, whose extended version was published later on
an international journal, or articles that were erroneously
classified as journal papers in the scientific databases, but
were actually conference papers. Out of the 848 papers, 715
were published since 2006. By applying the second inclusion
criterion, we excluded the studies that did not target the
manufacturing industry, but rather were carried out in a dif-
ferent context, e.g., in the medical field or in the construc-
tion sector. We obtained 216 papers in English published in
peer-reviewed international journals since 2006 and target-
ing the manufacturing industry.
These papers were all retrieved and examined individu-
ally, by checking the title, abstract and main contents. Forty-
two of these papers were further excluded from the analysis,
because of their limited relevance to AR or because they
focused on VR instead of AR, obtaining 174 papers, which
constitute the sample of studies reviewed. The whole set of
216 papers retrieved and examined is reported in Table 1.
2.2. Classification and analysis
The 174 papers reviewed were initially classified into four
groups (see Table 1):
1. Review papers (15).
2. Technical papers (69), i.e., papers whose main focus is
on the development, calibration or refinement (and pos-
sibly testing) of a technical hardware or software feature
of the AR system.
3. Application papers (70), i.e., papers whose main focus is
on the development, deployment and possibly testing of
the AR solution in a real or laboratory environment.
4. Conceptual papers (20), i.e., papers that do not either
develop a new AR solution or apply an existing AR
system, but rather discuss some specific aspects or
issues of AR adoption in industry.
For the whole sample of papers reviewed, we provide
some descriptive statistics on the year of publication and
geographical origin of the study, to verify whether the focus
on IAR systems has increased over time and is equally dis-
tributed across the various countries where it has been ana-
lyzed (subsection 3.1). Then, the different groups of papers
are analyzed (subsection 3.2). Review papers and conceptual
papers, which are fewer in number, are examined almost
individually, by providing an overview of the main topics
treated. As application papers and technical papers are more
numerous, the framework for their analysis is grounded on
the classification of the study keywords as way to capture
the essential topics covered, according to Fadlalla and
Amani (2015). These authors suggested evaluating two main
parameters, namely the frequency of use of the keywords
and their persistence. Frequencyrefers to the number of
times a concept is used as a keyword by researchers; from a
quantitative perspective, it is measured as the number of
articles where a given keyword appears. Persistenceis a
time-based measure reflecting the continuity of a given con-
cept over time; it can be measured as the number of years
since a concept was first introduced as a keyword. Such ana-
lysis is expected to generate an overview of the main
research areas relating to AR in industry, as well as to cat-
egorize the research topics on the basis of their importance
to the scientific community. We used the Exportfunction
of Scopus/WOS to retrieve the keywords of technical and
application papers automatically; for the papers found on
the Ebsco database, keywords were retrieved manually.
As further aspects of the analysis, the main fields of
application of AR, the technological solutions deployed, and
the results achieved from the AR usage are described in
Sections 3.3, 3.4, and 3.5, respectively, for the different
group of papers and together with their evolution in time
(where appropriate).
Figure 1. Scheme of the query and related results.
Table 1. Full list of papers retrieved and analyzed.
Reference Included (Y/N)
Application paper Technical paper Review paper Conceptual paper
Adcock & Gunn (2015)YX
Ahn & Han (2012)YX
Ajanki et al. (2011)Y X
Al-Mouhamed et al. (2006)Y X
Anastassova & Burkhardt (2009)Y X
Armesto et al. (2008)N
Aromaa & V
anen (2016)Y X
Arshad et al. (2016)Y X
Aschenbrenner et al. (2016)YX
Aurich et al. (2009)N
Behzadan & Kamat (2010)Y X
Benbelkacem et al. (2013)Y X
Bhowmik (2017)N
Biocca et al. (2007)Y X
Bleser et al. (2015) N
Blum et al. (2013)N
Borsci, Lawson, Jha, Burges and Salanitri (2016)N
Borsci, Lawson, Salanitri and Jha (2016)N
Bottecchia et al. (2009)YX
cet al. (2014)N
Buker et al. (2012)N
Candela et al. (2014)Y X
Canessa et al. (2014)Y X
Carmigniani et al. (2011)Y X
Caruso et al. (2015)Y X
Cauchard et al. (2012)N
Celozzi et al. (2013)N
Chandaria et al. (2007)Y X
Chen (2014)Y X
Chen, Hong and Wang . (2014) Y X
Chen, Jin and Wang (2014)N
Chen, He, Mo, Li and Yang, (2016)YX
Chen, Chi, Kang and Hsieh . (2016) Y X
Cheok et al. (2007)N
Chimienti et al. (2010)Y X
Choi et al. (2015)N
po & Wers
enyi (2013)N
De Crescenzio et al. (2011)YX
De Lucia et al. (2011)Y X
De Marchi et al. (2013)Y X
De Marsico et al. (2014)N
Di Cecca et al. (2016)YX
Doshi et al. (2017)YX
Duarte et al. (2010)N
Eck et al. (2015)YX
El Kabtane et al. (2016)YX
Elia et al. (2016)Y X
ındola et al. (2013)YX
Ferrise et al. (2013)YX
Fiorentino et al. (2013)YX
Fiorentino et al. (2014)Y X
Fiorentino et al. (2016)YX
Fox (2010)Y X
Fox et al. (2011)N
Franceschini et al. (2016)YX
Galambos et al. (2015)N
Gattullo et al. (2015)Y X
Gavish et al. (2015)YX
Gedik & Alatan (2013)Y X
Geng et al. (2015)N
Georgel et al. (2009)Y X
Gimeno et al. (2013)Y X
nski et al. (2014) Y X
Gonzalez-Sanchez et al. (2012)YX
Green et al. (2008)Y X
Gurevich et al. (2015)YX
Hagbi et al. (2011)YX
Haist (2008) Y X
Han & Zhao (2015)Y X
Han et al. (2011)N
Harwood & Revell (2017)N
Heemskerk et al. (2011)N
Table 1. Continued.
Reference Included (Y/N)
Application paper Technical paper Review paper Conceptual paper
Heikkinen & Handroos (2013)N
Henderson & Feiner (2010)Y X
Henderson & Feiner (2011)YX
Hou & Wang (2013)Y X
Hovanec et al. (2014)Y X
Hovanec et al. (2015)N
Huang et al. (2012)YX
Huang et al. (2016)Y X
Huenerfauth (2014) Y X
Igarashi & Inami (2015)Y X
Irizarry et al. (2014)YX
Itoh et al. (2015)YX
Januszka & Moczulski (2011)YX
Jiang & Nee (2013)YX
Jiang et al. (2014)YX
Jimeno-Morenilla et al. (2013)Y X
Jung et al. (2010)YX
Kadavasal & Oliver (2009)N
Karbasi et al. (2016)N
Kellner et al. (2012)Y X
Kim & Lee (2016)YX
Kim & Moon (2013)YX
Klein & Murray (2010)Y X
Koch et al. (2014)YX
Krajcovic et al. (2014)YX
Lakshantha & Egerton (2016)YX
Lamberti et al. (2015)Y X
Lamberti et al. (2017)Y X
Lambrecht et al. (2013)Y X
Langley et al. (2016)Y X
Langlotz et al. (2011)Y X
Lee & Akin (2011)YX
Lee & Rhee (2008)YX
Lee et al. (2009)YX
Lee, Lee, Kim and Kim (2010)YX
Lee et al. (2010) Y X
Lee, Billinghurst and Woo (2011)Y X
Lee, Han and Yang (2011)N
Lee et al. (2016)YX
Leu et al. (2013)Y X
Li et al. (2014)N
Lijun et al. (2008)YX
Liu & Zhang (2015)Y X
Liu et al. (2013)YX
Liu et al. (2014)YX
Liu et al. (2015)YX
Liu et al. (2016)YX
Liverani et al. (2006)N
Luh et al. (2013)YX
Makris et al. (2013)Y X
ınez et al. (2013)YX
Monroy Reyes et al. (2016)YX
Morkos et al. (2012)YX
Moser et al. (2015)YX
Mossel (2015)YX
Mourtzis et al. (2013)Y X
Mourtzis et al. (2015)N
Mourtzis et al. (2017)YX
Nakai & Suzuki (2016)YX
Nakanishi & Sato (2015)Y X
Nakanishi et al. (2010)Y X
Narducci et al. (2016)N
Nathanael et al. (2016)N
Nazir et al. (2013)N
Nedel et al. (2016)N
Nee et al. (2012)Y X
Neges et al. (2017)YX
Neubert et al. (2012)Y X
Ng et al. (2013)YX
Novak-Marcincin & Novakova-Marcincinova (2013)Y X
Novak-Marcincin et al. (2013)YX
Oliveira et al. (2013)Y X
Table 1. Continued.
Reference Included (Y/N)
Application paper Technical paper Review paper Conceptual paper
Olsson et al. (2012)Y X
Ong & Wang (2011)YX
Ong & Zhu (2013)YX
Ong et al. (2007)YX
Ong et al. (2008)Y X
Orlosky et al. (2015)Y X
Pai et al. (2015)YX
Pai et al. (2016)YX
Pang et al. (2006)YX
Park & Kim (2013)YX
Park et al. (2009)YX
Piekarski (2006)YX
Pirvu et al. (2016)YX
Radkowski (2016) Y X
Radkowski et al. (2015)Y X
Rajendran et al. (2015)Y X
Rapaccini et al. (2014)Y X
Re et al. (2016)Y X
Rehman & Cao (2017)Y X
Reif & G
unthner (2009)YX
Reif et al. (2010)YX
Reinhart & Eursch (2008)YX
Rohidatun et al. (2016)Y X
Sangineto & Cupelli (2012)N
ınet al. (2007)YX
Schega et al. (2014)Y X
Shaaban et al. (2015)YX
Siewiorek & Smailagic (2016)N
Stork & Schub
o(2010)Y X
Suhaifi et al. (2015)Y X
c& Te
c (2017) Y X
Tegeltija et al. (2016)YX
Tsai et al. (2012)YX
Tuma et al. (2016)YX
Turner et al. (2016)N
Ueng & Chen (2016)Y X
Umetani et al. (2014)YX
Valentini (2009) Y X
Van West et al. (2007)Y X
Vanderroost et al.(2017)Y X
Verbelen et al. (2014)Y X
Vignais et al. (2013)YX
Vitzthum & Hussmann (2006)Y X
Vogl et al. (2006)YX
Vukobratovic (2010) N
Wang & Dunston (2006)Y X
Wang et al. (2011)N
Wang, Ong and Nee (2013)Y X
Wang, Ng, Ong and Nee (2013)Y X
Wang et al. (2016a)Y X
Wang et al. (2016b)YX
Wang et al. (2016c)YX
Webel et al. (2013)YX
Weidlich et al. (2008)Y X
Weinert et al. (2008)YX
Weng et al. (2012)YX
Westerfield et al. (2015)YX
ojcicki (2014)YX
Xiong et al. (2006)YX
Xu et al. (2017)YX
Yamauchi & Iwamoto (2010)Y X
Yang et al. (2014)YX
Yannakakis et al. (2009)N
Yew et al. (2016)YX
Yin et al. (2015)Y X
Yuan et al. (2008)YX
Zhang et al. (2010a)YX
Zhang et al. (2010b)YX
Zhang et al. (2011)YX
Zhang et al. (2014)N
Zhu et al. (2013)YX
Zhu et al. (2014)YX
3. Review results
3.1. Descriptive statistics
We begin by categorizing the groups of papers on the basis
of year of publication with the aim of demonstrating
increasing interest in AR systems; results are reported in
Figure 2.
As can be seen from Figure 2, the number of papers pub-
lished from 2006 to 2012 was between 5 and 14 papers per
year, whereas after that period research on AR has
increased, with more than 20 papers per year published
from 2013 to 2016. This shows that the interest on AR sys-
tems has grown steadily in the last decade and in recent
years in particular, with a peak of 30 and 29 papers pub-
lished in 2016 and 2013, respectively. Another interesting
aspect is that technical papers were published with good
continuity since 2006; with conceptual papers and review
papers appearing in 2008 and since then being published
with quite good continuity as well. Obviously, review studies
on a given subject can be carried out only after a wide num-
ber of research papers have been published. At the same
time, conceptual papers discussing specific aspects/issues
relating to AR usage or implementation (e.g., psychological
issues or technology acceptance issues see Section 3.2.2)
are probably motivated by the development of new technical
solutions or by new applications of AR in industry. A key
factor that is likely to contribute to the increase in AR
research in recent years is the widespread diffusion of
mobile devices (smartphones or tablet PCs), which has
meant a significant reduction in both the cost and effort
required to realize an AR system. In fact, 69% (110 out of
174 papers) of the studies reviewed were carried out in the
last 5 years.
Interesting considerations emerge from the geographical
distribution of the studies reviewed, which is shown in
Figure 3. To determine the country of the study, if conflicts
existed, the nationality of the first author was taken as refer-
ence, which is a common approach in review studies (e.g.,
Gao et al.2017).
Figure 3 shows that the majority of the studies come
from Singapore (12.64%), Germany (10.92%), Italy (10.34%),
USA (9.77%), and China (7.47%). This result can have a
twofold justification. First, many of the companies who are
manufacturing industrial devices for AR are located in these
countries (Market Reports Center, 2017). Moreover, several
well-known research groups dedicated to exploring the use
of AR for industrial applications are based in both Germany
and the USA (Friedrich, 2002; Raczynski and Gussmann,
2004; Westerfield et al., 2015). In this respect, it is interest-
ing to note that most of the studies carried out in Germany,
the USA and China are technical papers, which suggests
that these countries are particularly active in exploring new
hardware/software AR features and confirms the relation-
ships with the presence of research groups operating in two
of these countries. Finally, many countries have carried out
very few (i.e., one or two) studies on AR, demonstrating
that this technology is still in its early stage of adoption.
3.2. Topics and keywords analysis
3.2.1. Review papers
The 15 review papers provide a summary of the existing lit-
erature about different aspects of AR solutions and applica-
tions. These papers propose general state-of-the-art analyses
of AR technologies (Carmigniani et al.,2011), with a focus
in the main patents (De Lucia et al., 2011), as well as in the
manufacturing context (Ong et al., 2008). Other studies
review AR systems that can be used for specific purposes,
such as to enhance human computer interaction (HCI)
(Igarashi and Inami, 2015; Rajendran et al., 2015)or
human-robot collaboration (Green et al., 2008). Several
papers analyze AR applications devoted to a particular
Figure 2. Distribution of the group of papers as a function of the publication year (Note: partial results for 2017).
industrial task; the most investigated tasks are assembly (Stork
and Schub
o, 2010;Leuet al.2013; Wang et al., 2016a) , main-
tenance (Oliveira et al., 2013;Lambertiet al.2015), design
(Nee et al., 2012;Yinet al., 2015), safety/ergonomics (Hovanec
et al.,2014) and food logistics (Vanderroost et al., 2017).
3.2.2. Conceptual papers
Conceptual papers mainly include empirical/statistical studies
and position papers, with the only exception of the study by
Elia et al. (2016), who proposed a decision support system to
help production managers in selecting the most efficient AR
solution to be applied in specific manufacturing processes.
Position papers typically describe the potentials of AR
usage in different fields, encompassing finite elements ana-
lysis studies (Weidlich et al., 2008), virtual manufacturing
(Novak-Marcincin and Novakova-Marcincinova, 2013),
mobile learning (Chen, 2014) and lean manufacturing, this
latter in terms of waste reduction and improvement in pro-
cess efficiency (Huenerfauth, 2014). Anastassova and
Burkhardt (2009) discussed the benefits of AR as a tool to
solve some typical training problems for automotive service
technicians. The study by Langley et al.(2016) also focused
on the effectiveness of AR systems for training of users.
Other position papers provided guidelines for AR imple-
mentation, either in general terms (Chimienti et al., 2010)
or for the deployment of AR in the specific areas, e.g.,
assembly and remote maintenance (Haist, 2008). Fox (2010),
instead, described the requirements to properly design ICT
tools for the implementation of AR.
Looking at the statistical studies, Radkowski et al. (2015)
evaluated different visual features for the development of
AR-based assembly instructions with an increasing level of
difficulty. Their hypothesis with this study was that the
complexity of the visual feature should comply with the dif-
ficulty of the assembly task; therefore, the ultimate aim was
to associate different types of visual features of AR systems
to different levels of assembly task complexity. Again in the
field of assembly, Hou and Wang (2013) evaluated the
impact of genderon the post-training performance of nov-
ice assemblers, by analyzing the learning curves of test users
with two assembly treatments (i.e., AR versus 3D manuals).
Similar analyses were carried out by Fiorentino et al.(2014)
and Re et al.(2016), who evaluated empirically the effective-
ness of maintenance operations assisted with interactive AR
instructions compared with paper instructions, in terms of
execution time and error rate of the test users. A technical
comparison of two solutions for AR, i.e., a wearable head-
mounted device and a hand-held device (e.g., smartphone),
versus paper solutions was also carried out by Rehman and
Cao (2017). Aromaa and V
anen (2016) compared an AR
prototype and a VR one in terms of their suitability to sup-
port human factors/ergonomics evaluation during the
design phase.
Some statistical studies have focused on the acceptance of
AR technology. Nakanishi and Sato (2015) evaluated the
psychological and physiological effect of digital manuals pre-
sented by a retinal imaging display on workers in the manu-
facturing industry. Rapaccini et al.(2014) provided the
results of a field study of user acceptance of AR to support
the delivery of field services (e.g., maintenance activities) on
installed products, and Olsson et al.(2012) carried out a
similar study to evaluate the user acceptance, potentials and
risks associated to five different mobile AR scenarios.
Nakanishi et al. (2010) evaluated the situations where the
use of Head-Mounted Displays (HMDs) is really effective
and can be applied instead of (or in addition to) the trad-
itional auditory/visual instructions.
3.2.3. Technical papers
The analysis of the authorskeywords for the technical
papers generated an original list of 304 different terms. For
six papers, neither the authors keywords or index keywords
were available these studies have obviously been excluded
from this analysis. It should be mentioned that the general
keyword augmented reality(or AR) was also excluded
from the analysis, because it was originally used to make the
Figure 3. Geographic distribution of the studies reviewed.
search query on the scientific databases and to identify the
studies to review. It would thus turn out to be the most fre-
quent keyword, which could bias the results of the ana-
lysis itself.
A preliminary analysis revealed that authors often use
slightly different keywords to express a similar (or the same)
concept, which was expected. This is the case for 3Dand
three-dimensional,or assembly designand assembly.
We consequently screened manually the keywords to iden-
tify similarities and group them into the same thematic field.
The keywords resulting from the screening and grouping are
shown in Table 2. To be more effective, the table focuses on
keywords with a frequency of 2 and therefore it excludes
some keywords that are difficult to group with others, as
they relate to very specific topics; these are referred to as
non-groupedkeywords and clearly score a frequency of
one each.
Table 2 shows that the most relevant research issues of
technical papers focus on 3D applications (4.62% of the
grouped keywords), tracking (3.46%), interaction problems
(3.08%), VR (2.60%), calibration (1.92%), and user interface
(1.92%); as far as the application areas are concerned, the
technical solutions developed targeted primarily assembly
(2.69%), maintenance (1.15%), ergonomics (1.54%), and
manufacturing (1.12%).
The persistence of the grouped keywords was therefore
assessed to distinguish between research topics that are well-
established, emerging or have been debated in the past, but
have now disappeared from the scientific literature. By cor-
relating the frequency of the grouped keywords with their
persistence, using the data in Table 2, we obtained the
results in Figure 4. The graph in Figure 4 was divided into
quarters using a horizontal and a vertical line as separators.
For the persistence, the horizontal line is set at persistence
¼5.5, corresponding to half of the time-span covered by the
studies reviewed. For the frequency, the vertical line is set at
frequency ¼3.23, which is the mean of the frequencies for
the grouped keywords.
Table 2. Frequency of keywords of the technical papers and year of appearance.
Keyword Frequency Persistence Percentage (%) Year of first appearance
3D 12 11 4.62 2006
Tracking 9 9 3.46 2008
Interaction 8 7 3.08 2010
Assembly 7 10 2.69 2007
VR 6 10 2.31 2007
Calibration 5 5 1.92 2012
User interface 5 11 1.92 2006
Ergonomics 4 4 1.54 2013
Computer aided design (CAD) 4 11 1.54 2006
Mobile 4 6 1.54 2011
HMD 3 7 1.15 2010
Optical see-through (OST) 3 5 1.15 2012
Maintenance 3 10 1.15 2007
Camera 3 10 1.15 2007
Training 3 2 1.15 2015
Control 3 9 1.15 2008
Human-machine interaction (HMI) 3 10 1.15 2007
Display 3 11 1.15 2006
Manufacturing 3 4 1.15 2013
Eye-tracking 3 3 1.15 2014
Robotics 3 4 1.15 2013
Gesture 3 4 1.15 2013
Haptic 3 10 1.15 2007
User test 3 3 1.15 2014
Semantics 2 9 0.77 2008
Optical see-through HMD (OST-HMD) 2 2 0.77 2015
Telerobotics 2 11 0.77 2006
Indoor 2 5 0.77 2012
Human-computer interaction (HCI) 2 3 0.77 2014
Information systems 2 10 0.77 2007
Smart factory 2 1 0.77 2016
Ubiquitous computing 2 9 0.77 2008
Footwear 2 4 0.77 2013
Kinetics 2 4 0.77 2013
Pattern recognition 2 6 0.77 2011
Artificial, augmented and virtual reality (AAVR) 2 7 0.77 2010
Scene analysis 2 11 0.77 2006
Handheld AR 2 5 0.77 2012
Sensors 2 9 0.77 2008
Marker 2 7 0.77 2010
Scene structure and integration modeling language (SSIML) 2 11 0.77 2006
Mobile augmented reality (MAR) 2 2 0.77 2015
Finger detection 2 3 0.77 2014
Ontology 2 10 0.77 2007
Algorithm 2 3 0.77 2014
Kinematics 2 2 0.77 2015
Visualization 2 7 0.77 2010
Non-grouped keywords 108 41.54
Going back to the aim of analyzing the persistence and
frequency concepts, the top-right quarter of Figure 4
includes the grouped keywords which first appeared more
than 5.5 years ago and have been used more than 3.23 times
overall. This quarter is therefore expected to include well-
establishedresearch topics. The top-left quarter includes
the keywords which first appeared more than 5.5 years ago,
but have not been used frequently since then. These key-
words identify the intermittent concepts,i.e., concepts that
are discussed in the literature in an on-off manner (Fadlalla
and Amani, 2015). These are research topics that either the
researchers have not agreed on yet, or are relatively fre-
quently tackling changing topics. The bottom-left quarter
lists the keywords which have appeared more recently (i.e.,
less than 5.5 year ago) and have occurred only a few times
since then; consequently, these keywords are likely to
describe emergingresearch topics. These topics could
either disappear early or alternatively become trendy
topics, which reflect the keywords in the bottom-right quar-
ter; in this case, the research topics seem to be very promis-
ing, as the related keywords have appeared recently and
have already been used a considerable number of times.
It can be seen from Figure 4 that for technical papers
there are eight well-established research topics, which refer
to 3D, tracking, interaction, VR, AR implementation for
assembly/CAD, the development of mobile solutions, or user
interfaces. From this list it can be seen that some keywords
(e.g., trackingor interaction) falling under the category
of well-established research topics actually describe very gen-
eral concepts or common technical problem when develop-
ing AR solutions; therefore, they are likely to be mentioned
with a higher frequency by researchers. The use of AR for
assembly is primarily motivated by the fact that, on the one
hand, assembly processes constitute a significant portion of
the cost of a product (Wang, Ong and Nee, 2013; Wang,
Ng, Ong and Nee, 2013) and that, on the other one, this
cost can be dramatically reduced if a product is assembled
according to a well-planned assembly sequence. AR is there-
fore used to this end, in the attempt to automate the process
and enhance its efficiency. Technical papers address, among
others, topics relating to the generation of the assembly
sequence (Ong et al.2007; Makris et al., 2013), including
object recognition issues caused by the particular shape of
products (Wang, Ong and Nee, 2013; Radkowski, 2016), or
the automated positioning of 3D objects in the assembly
guiding system (Chen, Hong and Wang, 2014). Product
design/CAD is another well-established field of adoption of
AR; in this context, AR is typically used for product custom-
ization (Jimeno-Morenilla et al., 2013; Mourtzis et al., 2015),
to modify/correct the model (Georgel et al., 2009)orto
interact with its (virtual) 3D components (Caruso et al.
2015). The proper development of the user interface for AR
solutions, including speech recognition features, has been
addressed by Ajanki et al.(2011), Benbelkacem et al.(2013),
Figure 4. Persistence versus frequency of the keywords for technical papers.
and Caruso et al.(2015). Finally, the development of MAR
solutions has been discussed by Biocca et al.(2007),
Mourtzis et al.(2013), Verbelen et al.(2014), Han and Zhao
(2015), and Kim and Lee (2016).
Intermittent research topics of technical papers include 19
themes, covering AR applications to maintenance and tele-
robotics, as well as technical issues such as the development
of visualization systems or sensors. Among the devices used,
HMDs, cameras, and haptic devices fall into this category.
Maintenance is a crucial process for facilities and machines
to prevent failures and is frequently carried out manually,
involving significant time and cost (Koch et al.2014).
Maintenance represents an interesting problem domain for
the application of AR: indeed, most repair activities are con-
ducted by trained personnel applying established procedures
that can be effectively organized into sequences of tasks tar-
geting a particular item, machine, or location (Henderson
and Feiner, 2009; Vignali et al., 2018).
In line with these argumentations, in this area technical
AR solutions mainly aim at supporting inspectors during
the on-field inspection/diagnosis of a machine or when car-
rying out maintenance tasks (De Marchi et al., 2013), cover-
ing also facility maintenance (Koch et al., 2014). The use of
AR is expected to avoid delays and possible mistakes during
maintenance activities, thus decreasing the related costs
(Benbelkacem et al., 2013). Telerobotics is the area of
robotics concerned with the control of semi-autonomous
robots from a distance (Sheridan, 1989; Goldberg and
Siegwart, 2001). In this field, only two technical solutions
for the usage of AR were developed, i.e., an AR user inter-
face for nanoscale interaction (Vogl et al., 2006) and a real-
time client-server system that can be integrated with 3D AR
services (Al-Mouhamed et al., 2006). Visualization issues
have been dealt with by Klein and Murray (2010), who pro-
posed a method to model the artifacts produced by a small
low-cost camera and add these effects to an ideal pinhole
image produced by conventional rendering methods.
Sensors (i.e., typically inertial sensorsof mobile devices)
are used in AR environment to estimate the position, inclin-
ation, or movement of an object; they have been used to
this end by Chandaria et al.(2007) and Han and Zhao
(2015). An HMD is a display device, worn on the head or
as part of a helmet, with one or two small displays; an
exhaustive examination of display systems (including
HMDs) suitable for adoption in AR environments has been
made by Weng et al.(2012). Kellner et al.(2012) have
instead addressed the issue of calibrating these devices for
their optimal usage in AR or VR environments. Hapticis
a term derived from the Greek word hapticos,i.e., pertain-
ing to the sense of touch; accordingly, haptic technology is a
way to recreate the sense of touch by applying forces, vibra-
tions, or motions to the user (El Saddik et al., 2011). Haptic
AR systems (also called visuo-haptic augmented reality)
enable users to see and touch digital information that is
embedded in the real world (Eck et al., 2015). Van West
et al.(2007) have developed the haptic tweezer,i.e. a com-
bination of haptic technology and an electrostatic levitation
system that allows manipulating objects without direct
contact; this technological solution can be useful when
manipulating fragile or contaminated components, as it
avoids touching them. Another technical solution was devel-
oped by Henderson and Feiner (2010), to integrate haptic
technology and opportunistic controls, i.e., a class of user
interaction techniques for AR applications that support ges-
turing on and receiving feedback from affordances already
present in the domain environment (Henderson and
Feiner, 2008).
Emerging research topics (18) include the analysis of
technologies for body tracking (e.g., eye-tracking, finger, or
gesture detection), optical technologies (primarily OST-
HMD), or hand-held technologies. Manufacturing and train-
ing are among the emerging application areas of technical
papers. An OST-HMD is a wearable device that has the cap-
ability of reflecting projected images, as well as allowing the
user to see through it using AR. Eye-tracking and OST-
HMD have been examined jointly by Moser et al.(2015),
who carried out a user study to evaluate the registration
accuracy produced by three OST-HMD calibration methods,
from both an objective (quantitative) and subjective (qualita-
tive) perspective. Similarly, Orlosky et al.(2015) have pro-
posed ModulAR, a hardware and software framework
designed to improve flexibility and hands-free control of
video see-through AR displays; the framework integrates
eye-tracking for on-demand control of vision augmentations,
such as optical zoom or field of view expansion. Looking at
gesture tracking, Kim and Lee (2016) have proposed a
method for naturally and directly manipulating 3D AR
objects through touch and hand gesture-based interactions
in hand-held devices; the touch gesture is used for the AR
object selection, whereas the natural hand gesture enables
the direct and interactive manipulation of the selected
objects. Lambrecht et al.(2013) have applied a similar
approach, i.e., a combination of marker-less gesture recogni-
tion and MAR, to the programing of industrial robots.
Caruso et al.(2015) have instead proposed an interactive
AR system that enables the user to freely interact with vir-
tual objects integrated in a real environment, avoiding the
use of cumbersome equipment. The adoption of AR in an
Industry 4.0 vision (Drath and Horch, 2014) has been pro-
posed by Yew et al.(2016), with the aim to enhance the
information perception of the different types of workers
interactions in the environment. Liu and Zhang (2015) and
Rohidatun et al.(2016) have instead proposed AR solutions
respectively for welder training and assembly/disassem-
bly training.
Finally, two trendy research topics emerged for technical
papers, i.e., calibration and ergonomics. Calibration of AR
devices has been dealt with by Canessa et al. (2014) for the
case of a color camera, Liu et al. (2016) for an AR guiding
system, Kellner et al.(2012) for an HMD, and Eck et al.
(2015), Itoh et al.(2015) and Moser et al. (2015) for OST-
HDMs. Ergonomic issues were instead examined by Schega
et al.(2014), who evaluated the effect of different HMDs on
visual performance of the users, and Tuma et al. (2016) who
used AR to evaluate the ergonomic state of a workplace.
3.2.4. Application papers
The analysis of the authorskeywords for the application
papers lead to an original list of 304 different terms. As per
the previous examination, three papers lacked the authors
keywords or index keywords and where therefore excluded
from this analysis. The general keyword augmented reality
was also excluded from the analysis.
The original list of keywords was analyzed to identify
possible similarities in the concept expressed and in the case
group them; this screening reduced the keywords to 142.
Results are proposed in Table 3, which, to be more effective,
focuses again on keywords with a frequency of 2; key-
words difficult to group were referred to as non-grouped
keywords and clearly score a frequency of one each.
Table 3 shows that AR applications in industry have tar-
geted the areas of assembly (6.05% of the grouped key-
words), maintenance (4.84%), design (3.63%), and
manufacturing (2.82%). Mobile applications (2.82%), 3D
models (2.02%), and VR (2.02%) are the most relevant kinds
of applications developed.
By correlating these findings with the analysis of the per-
sistency of each keyword, which is again shown in Table 3,
the research topics were classified as proposed in Figure 5.
To build the graph in Figure 5, the horizontal line was set
again at persistence ¼5.5, while the vertical line was set at
frequency ¼3.6, which is the mean of the frequencies for
the grouped keywords. From Figure 5 it can be seen that for
application papers there are 10 well-established research
topics relating, among others, to AR applications in the field
of assembly, maintenance, design/prototyping, manufactur-
ing, and order picking. Assembly turned out to be a well-
established research topic for both application papers and
technical papers; this means that numerous works have
focused on either the development of hardware/software AR
solutions for assembly or the deployment of AR to guide
operators in assembly tasks. Application papers have
focused, in particular, on this latter aspect (Pang et al., 2006;
Yuan et al., 2008; Valentini, 2009; Ong & Wang 2011;
Gonzalez-Sanchez et al., 2012; Liu et al., 2015; Wang et al.,
2016b); the same approach can be easily extended to the dis-
assembly process, as done by Tegeltija et al.(2016). Zhang
et al.(2011) have also integrated AR with Radio-Frequency
IDentification (RFID) technology to guide operators in the
assembly process. Product design issues are sometimes inte-
grated with assembly issues, as both activities are critical to
the product development process (Ng et al., 2013). AR
Table 3. Frequency of keywords of the application papers and year of appearance.
Keyword Frequency Persistency Percentage (%) Year of first appearance
Assembly 15 11 6.05 2006
Maintenance 12 6 4.84 2011
Design 9 11 3.63 2006
Manufacturing 7 9 2.82 2008
Mobile 7 5 2.82 2012
3D 5 6 2.02 2011
VR 5 4 2.02 2013
Picking 4 8 1.61 2009
Collaboration 4 8 1.61 2009
Prototyping 4 8 1.61 2009
Bare-hand 4 6 1.61 2011
Building information modeling (BIM) 4 6 1.61 2011
Safety 3 5 1.21 2012
MAR 3 3 1.21 2014
Interaction 3 8 1.21 2009
HMI 3 7 1.21 2010
Quality control 3 7 1.21 2010
Training 3 4 1.21 2013
MR 3 7 1.21 2010
Camera 3 9 1.21 2008
Intelligent algorithm 3 3 1.21 2014
Learning 3 5 1.21 2012
Telematics 2 1 0.81 2016
Registration 2 7 0.81 2010
Object tracking 2 5 0.81 2012
Context-awareness 2 4 0.81 2013
Acceptance sampling 2 1 0.81 2016
Ergonomics 2 4 0.81 2013
User interface 2 6 0.81 2011
Haptics 2 7 0.81 2010
Real time 2 9 0.81 2008
Authoring 2 4 0.81 2013
Remote assistance 2 2 0.81 2015
Layout planning 2 4 0.81 2013
Spatial augmented reality (SAR) 2 2 0.81 2015
CAD 2 6 0.81 2011
Tracking 2 11 0.81 2006
Visualization 2 4 0.81 2013
User test 2 8 0.81 2009
Natural interface 2 4 0.81 2013
Marker 2 1 0.81 2016
Non-grouped keywords 100 40.32
applications targeting expressively the product design issues
have been carried out by Januszka and Moczulski (2011),
Huang et al.(2012), Luh et al.(2013), Yang et al.(2014),
Lee et al.(2009), and Lee, Lee, Kim and Kim, 2010).
Maintenance is another well-established application area of
AR technology; the purpose of AR deployment in this field
could be either to train employees about maintenance tasks
(De Crescenzio et al., 2011; Webel et al., 2013; Westerfield
et al., 2015), or to enhance the effectiveness and accuracy of
the process, by supporting and guiding employees and
avoiding errors or safety issues (Henderson and Feiner,
2011; Lee and Akin, 2011; Esp
ındola et al., 2013; Ong and
Zhu, 2013; Zhu et al., 2013,2014; Fiorentino et al., 2016).
AR is particularly useful in those situations where mainten-
ance activities should be carried out in hazardous environ-
ments (De Crescenzio et al., 2011; Mart
ınez et al., 2013;
Nakai and Suzuki, 2016) or when the machine/device is
complex (Gola
nski et al., 2014;W
ojcicki, 2014). Three
papers (Reif and G
unthner, 2009; Reif et al., 2010; Krajcovic
et al., 2014) have finally applied AR to improve the accuracy
and efficiency of the order picking process. Among the tech-
nical aspects, bare-hand solutions have been implemented
quite frequently in industry, especially in the field of assem-
bly (Ong and Wang 2011;Nget al., 2013; Wang et al.,
2016c). Bare-handmeans that no device has to be worn to
interact with a computer; rather, the position of the hand
and the fingers is used to control applications directly.
These solutions can be interesting, as they do not need add-
itional equipment or physical sensors, which can be
inconvenient for employees and relatively inaccurate (Choi
et al., 2011).
Intermittent research topics of application papers
include 12 themes, embracing AR applications for CAD,
quality control, tracking or HMI. In the field of CAD,
Januszka and Moczulski (2011) have adopted AR for aiding
product designers in the development of machinery sys-
tems; Fiorentino et al.(2013) have instead proposed a
design review workspace which acquires user motion using
a combination of video and depth cameras and visualizes
the CAD models using monitor-based AR. The use of AR
for quality control has been proposed solely by
Franceschini et al.(2016). Studies dealing with HMI have
been developed by Lee et al., (2010)andTegeltijaet al.
(2016). The technical solutions that fall among the inter-
mittent concepts embrace the use of haptic devices (Lee,
Lee, Kim and Kim, 2010; Ferrise et al., 2013; Webel et al.
2013) and the development of user interfaces (Henderson
and Feiner, 2011;Gola
nski et al., 2014; Chen, Chi, Kang
and Hsieh, 2016).
Emerging research topics (17) encompass AR usage for
safety, learning, training, ergonomics, remote applications
(remote maintenance in particular), and layout planning.
The use of AR for occupational safety is a very recent topic.
The rationale for adopting AR for safety purposes is that
this technology can be useful in reducing some risk factors
for work injuries, namely insufficient training or work
experience and monotonicity of the tasks performed. More
precisely, by guiding the employees step-by-step in their
Figure 5. Persistence versus frequency of the keywords for application papers.
work, safety procedures can also be implemented success-
fully (Tati
c and Te
c, 2017). As far as learning is concerned,
AR has been adopted in e-learning (El Kabtane et al., 2016)
and mobile learning (Gonzalez-Sanchez et al., 2012) envi-
ronments. Vignais et al.(2013) have instead proposed an
AR-based tool for the evaluation of ergonomic conditions of
workers, with a particular attention on the potential risks
for musculoskeletal disorders. Ng et al.(2013) have
addressed ergonomic issues related to an AR-based assembly
process. Remote applications with the use of AR have been
proposed by Adcock and Gunn (2015), Gurevich et al.
(2015), Mourtzis et al.(2017), and Neges et al.(2017). SAR
and MAR are among the emerging technical solutions. SAR
augments real-world objects and scenes without the use of
displays such as monitors, HMDs or hand-held devices;
instead, it makes use of digital projectors to display graph-
ical information onto physical objects. MAR systems make
use of one or more of the following tracking technologies:
digital cameras and/or other optical sensors, accelerometers,
GPS, gyroscopes, solid state compasses, RFID and wireless
sensors (Chatzopoulos et al., 2013). SAR has been applied to
enhance the accuracy of automotive manufacturing proc-
esses (Doshi et al., 2017). Examples of MAR usage in indus-
try have been proposed by Shaaban et al.(2015) for the case
of laptop maintenance and Monroy Reyes et al.(2016) for
the training of students in the use of milling and lathe
machines in laboratories.
Finally, the two trendy research topics of application
papers embrace the development of mobile applications and
VR applications. Kim and Moon (2013) have implemented a
mobile application using AR to train employees in car main-
tenance. Tsai et al.(2012) have developed a mobile solution
integrating AR and geographical information to help the
evacuation of people in the case of a nuclear accident. VR
applications have been proposed by Park and Kim (2013),
Gavish et al.(2015), and El Kabtane et al.(2016) in the
training environment and Tati
c and Te
c(2017) in the
occupational safety environment.
3.3. Field of application of AR
Some initial insights about the key application fields of AR
systems have already been gained from the analysis of the
grouped keywords proposed previously. To confirm these
preliminary indications, the application field was analyzed
across the different groups of papers and as a function of
the publication year. The distribution of the application field
across the four groups of papers is proposed in Table 4.As
many papers typically mention a primary field of usage of
AR and a secondary one, the total number of applications
listed in Table 4 is higher than the number of stud-
ies reviewed.
From Table 4 it can be seen that 37 papers do not specify
an application field for AR; these are mostly technical papers
that developed either general AR solutions or conceptual/
review papers that discussed the usage of AR under a gen-
eral perspective without referring to a particular context.
Apart from these studies, results in Table 4 confirm that
assembly and maintenance are the most popular application
fields of AR, as they are mentioned (either as the primary or
secondary fields of usage of AR) in 15.09% and 14.65% of
the papers, respectively. Moreover, all paper types have tar-
geted these fields, although application papers and technical
papers are obviously prevalent. The next most widespread
fields of adoption of AR are training/learning (12.5% of the
studies reviewed), product design (7.33%), safety (6.03%),
remote assistance (5.17%), and telerobotics/robotics (5.60%).
It is worth mentioning that some application fields are fre-
quently considered together: this is, for instance, the case for
assembly and training (eight papers), maintenance and
remote assistance (seven papers), maintenance and training
(seven papers), safety and ergonomics (three papers), assem-
bly and product design (three papers).
Looking at the evolution of the application fields in time,
Table 5 shows that the interest towards AR adoption in the
field of assembly and maintenance has grown in time, with
most of the papers (22 out of 35 for assembly and 23 out of
34 for maintenance) published between 2013 and 2016; this
Table 4. Field of application versus paper type.
Application field Application papers Conceptual papers Review papers Technical papers Total
Assembly 15 6 3 11 35
Maintenance 23 2 2 7 34
Product design 11 2 4 17
Safety 8 1 5 14
Remote assistance 6 2 1 3 12
Telerobotics/robotics 1 3 9 13
Ergonomics 1 2 1 3 7
Training/learning 13 6 3 7 29
Quality control 3 1 4
Facility inspection or management 4 4
Outdoor environment 1 3 4
Picking 3 1 4
Diagnostic 3 3
Prototyping 1 1 1 3
Information 1 1 1 3
Navigation 1 1 2
2D/3D CAD 1 1 2
Layout planning 1 1
Welding 1 1
Machining simulation 1 1
Other 1 1 2
not specified 4 4 29 37
suggests that these fields of application of AR, although
quite explored, are still attracting new research. Similar con-
siderations can be made for AR adoption for training/
learning purposes. Safety, telerobotics/robotics, and remote
assistance, instead, are application fields of AR which have
emerged in recent years and are now investigated with
good continuity.
Further indications can be derived from an analysis
of the industrial sector of interest for the implementation of
AR. In this case, we have not considered the main area of
application of the AR solution, as done previously, but
instead focus on the industrial sector where the application
was developed and implemented (e.g., automotive sector, oil
& gas sector, etc.). The results of this analysis are shown in
Table 6.
Table 6 shows that in most cases (81 out of 174 papers,
46.55%) the systems implementation was carried out in
laboratory settings, which prevents the possibility of identi-
fying a particular industrial sector. Most of the times this is
the case for technical papers (49), as it was reasonable to
expect. Any AR solution, however, is ultimately expected to
leave the lab and be used in the real industrial context
(Georgel, 2011); in this respect, laboratory studies typically
represent a proof-of-concept of the AR applicability and
should be seen as the first step for the development of a
scalable solution.
Among studies developed in a particular sector, the man-
ufacturing (10.34%), machine tool industry (6.90%), AECO
(6.32%), and automotive (6.32%) industries are the most
popular contexts where AR has been implemented. Other
sectors, which are less frequent among the studies reviewed,
are the aircraft, chemical, food, footwear, and electronic
industries, the warehouse processes and nuclear/power plant
sector. Probably, the cost-effectiveness, benefits, and scalabil-
ity of AR solutions in these contexts has still to be demon-
strated (Weidenhausen et al., 2003).
By examining jointly the industry sector and primary
field of usage of AR (where specified), it is easy to see that
most implementations of AR for assembly were carried out
in laboratory settings or targeted the machine tool industry
(Table 7). AR studies in the manufacturing industry, instead,
have primarily focused on maintenance or product design
issues. It is interesting to note that AR studies targeting
ergonomics issues have been entirely carried out in labora-
tory settings, with real case applications still lacking in
the literature.
Table 5. Field of application versus year of publication (Note: partial results for 2017).
Application field 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Total
2D/3D CAD 1 1 2
Assembly 1 1 1 2 3 3 1 8 2 4 8 1 35
Product design 1 1 1 3 1 1 2 4 1 1 1 17
Diagnostic 111 3
Ergonomics 1 1 2 1 2 7
Facility inspection or management 1 1 2 4
Information 1 1 1 3
Layout planning 1 1
Machining simulation 1 1
Maintenance 1 2 1 3 1 10 5 4 4 3 34
Navigation 2 2
Training/learning 1 1 5 2 1 6 2 4 7 19
Telerobotics/robotics 3 1 2 3 4 13
Outdoor environment 1 1 1 1 4
Picking 1 1 1 1 4
Prototyping 1 1 1 3
Quality control 2 1 1 4
Remote assistance 2 1 2 1 5 1 12
Safety 1 2 1 2 2 1 1 3 1 14
Welding 1 1
Other 112
not specified 1 1 3 3 3 4 3 4 9 5 1 37
Table 6. Industrial sector versus paper type.
Industrial sector Application paper Conceptual paper Review paper Technical paper Total
Aircraft 3 3
Architecture, engineering, construction and operations (AECO) 5 1 5 11
Automotive 6 2 3 11
Chemical plants 1 1
Electronics 4 2 6
Food industry 1 1
Footwear 2 2
Laboratory 23 6 3 49 81
Machine tool 7 3 2 12
Manufacturing 9 2 3 4 18
Warehousing 4 4
Nuclear/power plants 4 3
Other 1 1 1 3
not specified 1 3 8 5 17
3.4. Technology evolution
We now examine the technological solutions adopted in the
implementation of AR systems. According to Reif et al.
(2010) and Jeon et al.(2010), a typical AR system includes
some essential components, such as the visualization/captur-
ing device, the interaction device, and the tracking system.
Capturing technologies are technological solutions adopted
to capture a scene, which can then be viewed by a user with
superimposed information. These technologies are therefore
responsible for collecting information about the environ-
ment. Visualization devices are used to display the results of
image processing or the image of the real-world enriched
with additional useful information (Milgram and Kishino,
1994). Interaction devices are used for commands that affect
information processing and displaying. Finally, tracking
technologies typically refers to the technological solutions
used to enable the AR system to recognize the key compo-
nents in the captured scene and thus provide the correct
information when augmenting the scene itself. Tracking
technologies are also essential for identifying the users pos-
ition within the industrial environment.
Most of the papers reviewed describe all the components
of the AR solutions; however, some papers describe more
than one AR application (and thus the hardware compo-
nents are more numerous) whereas other papers (in particu-
lar, technical papers, review papers or conceptual papers) do
not always focus on the description of the full AR architec-
ture. Consequently, the number of solutions described below
does not reflect the number of published studies.
Table 8 provides an overview of the technological solu-
tions as a function of the paper type, whereas Table 9
describes the evolution of these technologies over time.
From these tables it is easy to see that camera,or its vari-
ant camera connected to a monitor,are the most com-
monly used solutions for capturing the scene of the external
environment; in that case, the monitor displays the scene to
the users with the relevant additional information. Another
technological solution frequently adopted for visualization
purposes is the HMD. Compared with the first solution, the
HMD is easier to transport and does not require cameras or
monitors to be installed in the production area. The use of
HMD solutions has slightly increased in time: since 2011,
more than six papers per year make use of this solution.
However, some authors have recently criticized the use of
HMD, especially for remote assistance purposes (Gurevich
et al., 2015); as a matter of fact, the use of HMD could force
the worker to limit his/her head movements in an attempt
not to make the view of the remote assistant unstable. Some
studies also reported that users can suffer from decreased
visual acuity while looking at a physical target through
HMDs (Livingston et al., 2005).
In studies published since 2010, most of the technological
solutions for visualization make use of tablets, smartphones
or other mobile devices, e.g., ultra-mobile PCs. In this case,
the scene is captured by the device camera and is immedi-
ately visualized with superimposed information on the
device display. Compared with HMD solutions, the use of
tablets or smartphones is more socially accepted and has the
advantage of being even easier to transport. However,
mobile devices are hand-held when used, which can hinder
the operator when he/she has to carry out manual tasks
(e.g., assembly or maintenance tasks). For this reason, bare-
hand solutions have been proposed, starting from 2011.
These solutions aim at developing a natural and intuitive
hand-based interaction with virtual objects and have been
applied in seven papers; moreover, as this is a relatively new
interaction technique, five papers have investigated bare-
hand interaction from a technical point of view.
In some of the papers reviewed, non-conventional AR
systems were developed. Jimeno-Morenilla et al.(2013) pro-
vided an example of use of a non-conventional system, i.e.,
an infrared emitter coupled with a pair of active glasses.
Table 7. Industrial sector versus primary application field of AR.
Aircraft AECO Automotive
plants Electronics Food Footwear Laboratory
tool Manufacturing
plant Warehousing
Assembly 1 3 2 17 4
Diagnostic 1
Ergonomics 4
Facility inspection or
1 1
Information 1
Layout planning 1
Training/learning 1 9 1 1
Machining simulation 1
Maintenance 2 2 3 1 8 3 4 1
Navigation 2
Outdoor environment 3
Picking 13
Product design 1 1 2 4 1 3
Quality control 1 1 1
Remote assistance 1 2 2 2
Safety 3 1 2 3
Telerobotics/robotics 8 2
Welding 1
Other 1
Similarly, Liu et al.(2013) described a technical issue of AR
systems equipped with holographic displays.
By coupling the technological solution with the primary
application field of usage of AR (where specified), we got
the results shown in Table 10. This table shows that cam-
eras, HMDs and tablets/smartphones have been adopted in
almost all application fields of AR. The use of HMDs and
mobile devices is slightly prevalent in the areas where
instructions should be given to the user, i.e., assembly and
maintenance. This is consonant with the fact that HMD
instructions are more effective than paper instructions (Tang
et al., 2003) and that the use of mobile devices decreases the
risk of errors in assembly operations. Bare-hand solutions
instead have been primarily adopted in the (augmented)
Table 8. Technological devices versus paper type.
Visualization, capturing and interaction devices Application paper Conceptual paper Review paper Technical paper Total
3D scanner 1 1 2
Bare-hand 7 5 12
Camera 31 1 1 37 70
Camera connected to a monitor 10 3 3 16
Display 4 2 1 7
Haptic devices 5 1 1 8 15
Head-worn camera 2 3 5
HMD 28 8 6 20 62
Holographic display 1 1
Inertial measurement unit 1 2 3
Infrared machine 1 1
Laptop 5 2 5 12
Projector 5 2 5 12
Tablets, smartphone or other mobile devices 25 3 7 16 51
Table 9. Technological devices versus year of publication (Note: partial results for 2017).
Visualization, capturing and interaction devices 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Total
3D scanner 1 1 2
Bare-hand 114 6 12
Camera 4 2 4276198815470
Camera connected to a monitor 2 1 1 5 2 1 4 16
Display 1 1 1 3 1 7
Haptic devices 1 1 1 2 1313 2 15
Head-worn camera 1 1 1 1 1 5
HMD 3253585778 6362
Holographic display 1 1
Inertial measurement unit 1 2 3
Infrared machine 1 1
Laptop 112231 1112
Projector 1135212
Tablets, smartphone or other mobile devices 1 2 3232387610451
Table 10. Technological device versus primary application field of AR.
Application field
scanner Bare-hand Camera
to a
monitor Display
camera HMD
machine Laptop Projector
or other
mobile devices
2D/3D CAD 1
Assembly 4 9 4 5 1 5
Product design 1 2 5 3 2 2
Diagnostic 1 1 1
Ergonomics 1 4 1
Facility inspection or management 1 1 1
Information 1
Layout planning 1
Training/learning 4 3 2 4 6 2 2
Machining simulation 1
Maintenance 7 1 1 7 12
Navigation 1 1
Outdoor environment 1 1 2
Picking 13
Telerobotics / robotics 1 6 2 2
Prototyping 11
Quality control 2 1 1
Remote assistance 4 2 1 3
Safety 2 1 1 6
Welding 1
assembly process, probably because assembly is a manual
process where ergonomic issues cannot be ignored (Azuma,
1997; Wang, Ng, Ong and Nee, 2013).
Tracking systems for AR can be classified into two main
categories, namely marker-based trackingand marker-less
tracking (Ababsa et al., 2010). The analysis of the selected
papers (Table 11 and Table 12) reveals that marker-based
systems are the most frequently adopted solutions, as they
are mentioned in 89 out of 174 papers reviewed (51.1%)
and of 125 papers (71.2%) that expressively describe the
tracking solutions adopted. Marker-based solutions have
been implemented in both older and more recent studies,
with an increase in usage in the last years. In these systems,
a kind of tag is placed on the elements of interest, to allow
the AR system to recognize it and provide the related infor-
mation when required. The studies analyzed made use of
barcodes/QR codes (two papers), fiducial markers (14),
optical markers (25), physical markers (11) or RFID tags
(three) to this end.
Marker-less systems are mentioned in 36 out of 174 stud-
ies reviewed (20.7%). When marker-less systems are
adopted, several different solutions can be used to track the
elements of interest. For instance, Zhang et al.(2010b)
developed a hybrid tracking method combining area-based
and feature-based tracking, whereas De Crescenzio et al.
(2011) adopted feature tracking: the application recognizes a
particular feature, which acts as a marker. Koch et al.
(2014), Fiorentino et al.(2016), and Neges et al.(2017)
made use of natural markers, i.e., markers that are already
available on-site and that because of their particular shape
or color have great potential to be used for optical tracking.
3.5. Results of the AR implementation
As discussed in the previous sections, most of the papers
reviewed have developed an AR application that could help
operators carry out tasks typical of industrial procedures.
According to Georgel (2011), the fact that the system was
tested is one of the key evaluation criteria for assessing AR
solutions. For the purpose of this study, we distinguish
between AR systems that have been not tested,”“technically
tested(i.e., tested only in terms of technical functioning) or
user tested,that is tested with a group of users, to evaluate
user acceptance and performance. Out of the 174 studies
reviewed, 139 papers have carried out either a technical test
or a user study. Looking at the application papers, most of
them (32 out of 70, 45.7%) carried out a user study to evalu-
ate the benefits generated when AR systems (instead of trad-
itional methods) are used to assist the user in the execution
of a task; in addition, 28 papers (40.0%) propose a technical
test of the developed application, aimed at illustrating how
it functions (Table 13). The majority of technical papers (45
out of 69, 65.2%) carried out a technical test aimed at
assessing the functioning of the technical solutions devel-
oped; conversely, user studies are more limited in number
(18 out of 69, 26.1%). User tests were conducted also in 12
conceptual papers and one review paper.
As far as the results observed are concerned, obviously
many studies measured more than one outcome generated
by the AR implementation. The distribution of the results
observed as a function of the type of test carried out (either
user studies or technical tests) and paper classification
(Table 14) shows that technical tests are mainly intended to
evaluate the effectiveness of the technical solutions, which is
mentioned among the test results in 54 out of 66 cases
(81.8%). Technical papers contribute mainly to this out-
come, with 34 studies that have tested the effectiveness of
the solution developed. Looking at the papers that carried
out user studies, their focus is primarily on evaluating the
savings in the time required to carry out a given task: this is
assessed in 26 cases out of 102 (25.5%). Further outcomes
frequently measured in user studies are again the perform-
ance of the technical solution, in terms of its effectiveness
and ease of usage (reported in 23 and 20 cases, respectively)
and the possibility of reducing errors associated with the
task executed (16 cases).
By combining the results observed through user studies
or technical tests with the primary application field of AR
(where specified), we found that assembly and maintenance
can benefit from faster execution of activities to the greatest
extent (Table 15). Better training is the typical result for AR
application targeting learning/training issues. Testing the
effectiveness of the solution developed is an important
aspect of many application areas of AR, including assembly,
maintenance, product design, and telerobotics/robotics.
Table 11. Tracking technology versus paper type.
Classification Marker-based tracking system Marker-less tracking system
Application paper 42 19
Conceptual paper 7 3
Review paper 8 1
Technical paper 32 13
Total 89 36
Table 12. Tracking technology versus year of publication (Note: partial results
for 2017).
Year of publication 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Total
Marker-based tracking system 3 1 4 4 8 7 3 15 8 14 17 5 89
Marker-less tracking system 2 1 3 2 1 5 4 6 3 7 2 36
Table 13. Test of the AR system versus paper type.
Classification Test type Total
Application papers user test 32
no test 10
technical test 28
Conceptual papers user test 12
no test 6
technical test 2
Review papers user test 1
no test 14
Technical papers user test 18
no test 6
technical test 45
Among the studies that carried out a thorough assess-
ment of the benefits generated by AR applications, De
Crescenzio et al.(2011) presented a prototype based on an
AR system to help operators in executing maintenance tasks
on aircraft. Their proposed system was tested with 10 opera-
tors, who showed improved performance and satisfaction
when using the prototype. Operators were asked to evaluate
the system; workload, performance and usage satisfaction
were all judged positively. Webel et al.(2013) presented an
AR system that can be used to train and guide technicians
through complex assembly and maintenance tasks. The
application has been tested at Sidel. The results obtained
with this application showed that AR had a positive impact
on the execution of maintenance tasks: technicians equipped
with the application required 14% less time to complete the
assigned job. The average number of unsolved errors scored
0.30 for equipped operators, whereas for unequipped ones it
scored 1.30. The only drawback was that employee training
time increased by 20% when the AR application was
adopted. The subjects involved in the test were the same age
and had equivalent work experience. AR also shows poten-
tial for operations requiring remote assistance. For instance,
Gurevich et al.(2015) developed a remote video
collaboration system called tele-advisorthat allows a
remote expert to natura