All One Needs to Know about Metaverse: A Complete Survey on Technological Singularity, Virtual Ecosystem, and Research Agenda

Abstract

Since the popularisation of the Internet in the 1990s, the cyberspace has kept evolving. We have created various computer-mediated virtual environments including social networks, video conferencing, virtual 3D worlds (e.g., VR Chat), augmented reality applications (e.g., Pokemon Go), and Non-Fungible Token Games (e.g., Upland). Such virtual environments, albeit non-perpetual and unconnected, have bought us various degrees of digital transformation. The term `metaverse' has been coined to further facilitate the digital transformation in every aspect of our physical lives. At the core of the metaverse stands the vision of an immersive Internet as a gigantic, unified, persistent, and shared realm. While the metaverse may seem futuristic, catalysed by emerging technologies such as Extended Reality, 5G, and Artificial Intelligence, the digital `big bang' of our cyberspace is not far away.

This survey paper presents the first effort to offer a comprehensive framework that examines the latest metaverse development under the dimensions of state-of-the-art technologies and metaverse ecosystems, and illustrates the possibility of the digital `big bang'. First, technologies are the enablers that drive the transition from the current Internet to the metaverse. We thus examine eight enabling technologies rigorously - Extended Reality, User Interactivity (Human-Computer Interaction), Artificial Intelligence, Blockchain, Computer Vision, IoT and Robotics, Edge and Cloud computing, and Future Mobile Networks.
In terms of applications, the metaverse ecosystem allows human users to live and play within a self-sustaining, persistent, and shared realm. Therefore, we discuss six user-centric factors -- Avatar, Content Creation, Virtual Economy, Social Acceptability, Security and Privacy, and Trust and Accountability. Finally, we propose a concrete research agenda for the development of the metaverse.

Full text

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 1
All One Needs to Know about Metaverse: A
Complete Survey on Technological Singularity,
Virtual Ecosystem, and Research Agenda
Lik-Hang Lee1, Tristan Braud2, Pengyuan Zhou3,4, Lin Wang1, Dianlei Xu6, Zijun Lin5, Abhishek Kumar6,
Carlos Bermejo2, and Pan Hui2,6,Fellow, IEEE,
Abstract—Since the popularisation of the Internet in the 1990s,
the cyberspace has kept evolving. We have created various
computer-mediated virtual environments including social net-
works, video conferencing, virtual 3D worlds (e.g., VR Chat),
augmented reality applications (e.g., Pokemon Go), and Non-
Fungible Token Games (e.g., Upland). Such virtual environments,
albeit non-perpetual and unconnected, have bought us various
degrees of digital transformation. The term ‘metaverse’ has
been coined to further facilitate the digital transformation in
every aspect of our physical lives. At the core of the metaverse
stands the vision of an immersive Internet as a gigantic, unified,
persistent, and shared realm. While the metaverse may seem
futuristic, catalysed by emerging technologies such as Extended
Reality, 5G, and Artificial Intelligence, the digital ‘big bang’ of
our cyberspace is not far away.
This survey paper presents the first effort to offer a compre-
hensive framework that examines the latest metaverse develop-
ment under the dimensions of state-of-the-art technologies and
metaverse ecosystems, and illustrates the possibility of the digital
‘big bang’. First, technologies are the enablers that drive the
transition from the current Internet to the metaverse. We thus
examine eight enabling technologies rigorously - Extended Real-
ity, User Interactivity (Human-Computer Interaction), Artificial
Intelligence, Blockchain, Computer Vision, IoT and Robotics,
Edge and Cloud computing, and Future Mobile Networks. In
terms of applications, the metaverse ecosystem allows human
users to live and play within a self-sustaining, persistent, and
shared realm. Therefore, we discuss six user-centric factors –
Avatar, Content Creation, Virtual Economy, Social Acceptability,
Security and Privacy, and Trust and Accountability. Finally, we
propose a concrete research agenda for the development of the
metaverse.
Index Terms—Metaverse, Immersive Internet,
Augmented/Virtual Reality, Avatars, Artificial Intelligence,
Digital Twins, Networking and Edge Computing, Virtual
Economy, Privacy and Social Acceptability.
I. INT ROD UC TI ON
METAVERSE, combination of the prefix “meta” (imply-
ing transcending) with the word “universe”, describes
a hypothetical synthetic environment linked to the physical
world. The word ‘metaverse’ was first coined in a piece
of speculative fiction named Snow Crash, written by Neal
Stephenson in 1992 [1]. In this novel, Stephenson defines
the metaverse as a massive virtual environment parallel to
Corresponding Authors: Lik-Hang Lee, E-mail: (likhang.lee@kaist.ac.kr)
1KAIST, South Korea; 2HKUST, Hong Kong SAR; 3USTC China; 4
MCT Key Lab of CCCD; 5UCL, UK; 6Uni. Helsinki, Finland.
Manuscript submitted in October 2021.
Fig. 1. We propose a ‘digital twins-native continuum’, on the basis
of duality. This metaverse vision reflects three stages of development. We
consider the digital twins as a starting point, where our physical environments
are digitised and thus own the capability to periodically reflect changes to their
virtual counterparts. According to the physical world, digital twins create
digital copies of the physical environments as ‘many’ virtual worlds, and
human users with their avatars work on new creations in such virtual worlds,
as digital natives. It is important to note that such virtual worlds will initially
suffer from limited connectivity with each other and the physical world,
i.e., information silo. They will then gradually connect within a massive
landscape. Finally, the digitised physical and virtual worlds will eventually
merge, representing the final stage of the co-existence of physical-virtual
reality similar to the surreality). Such a connected physical-virtual world give
rise to the unprecedented demands of perpetual and 3D virtual cyberspace as
the metaverse.
the physical world, in which users interact through digi-
tal avatars. Since this first appearance, the metaverse as a
computer-generated universe has been defined through vastly
diversified concepts, such as lifelogging [2], collective space
in virtuality [3], embodied internet/ spatial Internet [4], a
mirror world [5], an omniverse: a venue of simulation and
collaboration [6]. In this paper, we consider the metaverse as
a virtual environment blending physical and digital, facilitated
by the convergence between the Internet and Web technolo-
gies, and Extended Reality (XR). According to the Milgram
and Kishino’s Reality-Virtuality Continuum [7], XR integrates
digital and physical to various degrees, e.g., augmented reality
(AR), mixed reality (MR), and virtual reality (VR). Similarly,
the metaverse scene in Snow Crash projects the duality of
the real world and a copy of digital environments. In the
metaverse, all individual users own their respective avatars, in
analogy to the user’s physical self, to experience an alternate
life in a virtuality that is a metaphor of the user’s real worlds.
To achieve such duality, the development of metaverse
has to go through three sequential stages, namely (I) digital
twins, (II) digital natives, and eventually (III) co-existence
of physical-virtual reality or namely the surreality. Figure 1
depicts the relationship among the three stages. Digital twins
refer to large-scale and high-fidelity digital models and entities

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 2
Fig. 2. The cyberspace landscape of real-life applications, where superseding relationships exists in the information richness theory (left-to-right) as well as
transience-permanence dimension (bottom-to-top).
duplicated in virtual environments. Digital twins reflect the
properties of their physical counterparts [8], including the
object motions, temperature, and even function. The con-
nection between the virtual and physical twins is tied by
their data [9]. The existing applications are multitudinous
such as computer-aided design (CAD) for product design
and building architectures, smart urban planning, AI-assisted
industrial systems, robot-supported risky operations [10]–[14].
After establishing a digital copy of the physical reality, the
second stage focuses on native content creation. Content
creators, perhaps represented by the avatars, involve in digital
creations inside the digital worlds. Such digital creations can
be linked to their physical counterparts, or even only exist in
the digital world. Meanwhile, connected ecosystems, including
culture, economy, laws, and regulations (e.g, data ownership),
social norms, can support these digital creation [15]. Such
ecosystems are analogous to real-world society’s existing
norms and regulations, supporting the production of physical
goods and intangible contents [16]. However, research on such
applications is still in a nascent stage, focusing on the first-
contact point with users, such as input techniques and author-
ing system for content creation [17]–[20]. In the third and
final stage, the metaverse could become a self-sustaining and
persistent virtual world that co-exists and interoperates with
the physical world with a high level of independence. As such,
the avatars, representing human users in the physical world,
can experience heterogeneous activities in real-time charac-
terised by unlimited numbers of concurrent users theoretically
in multiple virtual worlds [9]. Remarkably, the metaverse
can afford interoperability between platforms representing
different virtual worlds, i.e., enabling users to create contents
and widely distribute the contents across virtual worlds. For
instance, a user can create contents in a game, e.g., Minecraft1,
and transfer such contents into another platform or game, e.g.,
Roblox2, with a continued identity and experience. To a further
extent, the platform can connect and interact with our physical
world through various channels, user’s information access
through head-mounted wearable displays or mobile headsets
(e.g. Microsoft Hololens3), contents, avatars, computer agents
in the metaverse interacting with smart devices and robots, to
name but a few.
According to the diversified concepts of computer-mediated
universe(s) mentioned above, one may argue that we are
already situated in the metaverse. Nonetheless, this is only
partially correct, and we examine several examples to jus-
tify our statement with the consideration of the three-stage
metaverse development roadmap. The Earth 3D map4offers
picture frames of the real-world but lacks physical properties
other than GPS information, while social networks allow users
to create contents but limited to texts, photos, and videos
with limited options of user engagements (e.g., liking a post).
Video games are getting more and more realistic and impres-
sive. Users can experience outstanding graphics with in-game
physics, e.g., Call of Duty: Black Ops Cold War, that deliver a
sense of realism that resembles the real world in great details.
A remarkable example of an 18-year-old virtual world, Second
Life5, is regarded as the largest user-created 3D Universe.
Users can build and shape their 3D environments and live
in such a virtual world extravagantly. However,video games
still lack interoperability between each other. The emerging
1https://www.minecraft.net/en-us
2https://www.roblox.com/
3https://www.microsoft.com/en-us/hololens
4https://earth3dmap.com/
5https://id.secondlife.com

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 3
Fig. 3. Connecting the physical world with its digital twins, and further shifting towards the metaverse: (A) the key technologies (e.g., blockchain, computer
vision, distributed network, pervasive computing, scene understanding, ubiquitous interfaces), and; (B) considerations in ecosystems, in terms of avatar, content
creation, data interoperability, social acceptability, security/privacy, as well as trust/accountability.
platforms leveraging virtual environments (e.g., VRChat6and
Microsoft Mesh7) offer enriched environments that emulate
virtual spaces for social gatherings and online meetings. How-
ever, these virtual spaces are not perpetual, and vanish after the
gatherings and meetings. Virtual objects in AR games (e.g.,
Pok´
emon Go8) have also been attached to the physical reality
without reflecting any principles of the digital twins.
Figure 2 further demonstrates the significant gap that re-
mains between the current cyberspace and the metaverse.
Both x- and y-axes demonstrate superseding relationships:
Left-to-Right (e.g., Text <Image) and Bottom-to-Top (e.g.,
Read and Write (RW) <Personalisation). The x-axis depicts
various media in order of information richness [21] from
text, image, audio, video, gaming, virtual 3D worlds, virtu-
ality (AR/MR/VR, following Milgram and Kishino’s Reality-
Virtuality Continuum [7]) and eventually, the physical world.
The y-axis indicates user experience under a spectrum be-
tween transience (Read and Write, RW) and permanence
(Experience-Duality, ED). We highlight several examples to
show this superseding relationship in the y-axis. At the Read
& Write level, the user experience does not evolve with the
user. Every time a user sends a SMS or has a call on Zoom,
their experience is similar to their previous experiences, as
well as these of all the other users. With personalisation,
users can leverage their preference to explore cyberspaces like
Spotify and Netflix. Moving upward to the next level, users can
proactively participate in content creation, e.g., Super Mario
Marker allows gamers to create their tailor-made game level(s).
Once a significant amount of user interaction records remain
in the cyberspace, under the contexts of personalisation and
content creation, the cyberspace evolves to a social community.
However, to the best of our knowledge, we rarely find real-
life applications reaching the top levels of experience-duality
that involves shared, open, and perpetual virtual worlds (ac-
cording to the concepts mentioned above in Figure 1). In brief,
the experience-duality emphasises the perpetual virtual worlds
6https://hello.vrchat.com/
7https://www.microsoft.com/en-us/mesh?activetab=pivot%3aprimaryr7
8https://pokemongolive.com/en/
that are paired up with the long-lasting physical environments.
For instance, a person, namely Paul, can invite his metaverse
friends to Paul’s physical home, and Paul’s friends as avatars
can appear at Paul’s home physically through technologies
such as AR/VR/MR and holograms. Meanwhile, the avatars
can stay in a virtual meeting room in the metaverse and talk to
Paul in his physical environment (his home) through a Zoom-
alike conversation window in a 3D virtual world.
To realise the metaverse, technologies other than the In-
ternet, social networks, gaming, and virtual environments,
should be taken into considerations. The advent of AR and
VR, high-speed networks and edge computing , artificial
intelligence, and hyperledgers (or blockchain), serve as the
building blocks of the metaverse. From a technical point
of view, we identify the fundamentals of the metaverse and
its technological singularity. This article reviews the existing
technologies and technological infrastructures to offer a critical
lens for building up the metaverse characterised by perpetual,
shared, concurrent, and 3D virtual spaces concatenating
into a perceived virtual universe. The contribution of the
article is threefold.
1) We propose a technological framework for the meta-
verse, which paves a way to realise the metaverse.
2) By reviewing the state-of-the-art technologies as en-
ablers to the development of the metaverse, such as edge
computing, XR, and artificial intelligence, the article
reflects the gap between the latest technology and the
requirements of reaching the metaverse.
3) We propose research challenges and opportunities based
on our review, paving a path towards the ultimate stages
of the metaverse.
This survey serves as the first effort to offer a comprehensive
view of the metaverse with both the technology and ecosystem
dimensions. Figure 3 provides an overview of the survey paper
– among the focused topics under the contexts of technology
and ecosystem, the keywords of the corresponding topics
reflect the key themes discussed in the survey paper. In the next
section, we first state our motivation by examining the existing
survey(s) as well as relevant studies, and accordingly position

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 4
our review article in Section II. Accordingly, we describe our
framework for the metaverse considering both technological
and ecosystem aspects (Section III).
II. RE LATE D WOR K AN D MOTI VATION
To understand the comprehensive landscape of existing
studies related to the metaverse, we decided to conduct a
review of the relevant literature from 2012 to 2021 (i.e., ten
years). In the first attempt of our search, we used the search
keyword “metaverse” in the title, the abstract, or the body
of the articles. We only focused on several primary sources
known for high-quality studies on virtual environments, VR,
AR, and XR: (ACM CHI) the ACM CHI Conference on
Human Factors in Computing Systems; (IEEE ISMAR) IEEE
International Symposium on Mixed and Augmented Reality;
(IEEE VR) IEEE Virtual Reality conference; (ACM VRST)
ACM Symposium on Virtual Reality Software and Technol-
ogy. We obtained only two effective results from two primary
databases of ACM Library and IEEE Xplorer, i.e., one full
article related to the design of artificial moral agents, appeared
in CHI [23]; and one poster paper related to multi-user
collaborative work for scientists in gamified environments,
appeared in VRST [24]. As the criteria applied in the first-
round literature search made only a few eligible research arti-
cles, our second attempt relaxed the search criteria to papers
with the identical search keyword of ‘metaverse’, regardless
of the publication venues. The two primary databases of ACM
Library and IEEE Xplorer resulted in 43 and 24 entities (Total
= 67), respectively. Then, we only included research article
written in English, and excluded demonstration, book chapters,
short papers, posters, and articles appeared as workshops,
courses, lectures, interviews, opinions, columns, and invited
talks – when the title, abstracts, and keywords in the articles
did not provide apparent reasons for exclusion, we read the
entire article and briefly summarise the remaining 30 papers
in the coming paragraphs.
First, we spot a number of system solutions and architec-
tures for resolving scalability issues in the metaverse, such as
balancing the workload for reduced response time in Modern
massively multiplayer online games (MMOGs) [25], unsuper-
vised conversion of 3D models between the metaverse and
real-world environments [26], High performance computing
clusters for large-scale virtual environments [27], analyzing
underground forums for criminal acts (e.g., trading stolen
items and datasets) in virtual worlds [28], exploration of new
composition and spatialization methods in virtual 3D spaces
under the context of multiplayer situations [29], governing
user-generated contents in gaming [30], strengthening the
integration and interoperability of highly disparate virtual
environments inside the metaverse [31], and redistributing net-
work throughput in virtual worlds to improve user experiences
through avatars in virtual environments [32].
Second, we spot three articles proposing user interaction
techniques for user interaction across the physical and virtual
environments. Young et al. proposed an interaction technique
9https://futuristspeaker.com/future-trends/the-history- of-the- metaverse/
10https://coinmarketcap.com/currencies/alien-worlds/
for users to make high-fiving gestures being synchronised in
both physical and virtual environments [33]. Vernaza et al.
proposed an interactive system solution for connecting the
metaverse and real-world environments through tablets and
smart wearables [34]. Next, Wei et al. made user interfaces for
the customisation of virtual characters in virtual worlds [35].
Third, the analysis of user activities in the metaverse also
gains some attention from the research community. The well-
recognised clustering approaches could serve to understand
the avatar behaviours in virtual environments [36], and the
text content created in numerous virtual worlds [37]. As
the metaverse may bridge the users with other non-human
animated objects, an interesting study by Barin et al. [38]
focuses on the crash incident of high-performance drone racing
through the first-person view on VR headsets. The concluding
remark of their study advocates that the physical constraints
such as acceleration and air resistance will no longer be
the concerns of the user-drone interaction through virtual
environments. Instead, the design of user interfaces could limit
the users’ reaction times and lead to the critical reasons for
crash incidents.
Next, we report the vastly diversified scenes of virtual
environments, such as virtual museums [39], ancient Chinese
cities [40], and virtual laboratories or classrooms [41]–[44].
We see that the existing virtual environments are commonly
regarded as a collaborative learning space, in which human
users can finish some virtual tasks together under various
themes such as learning environmental IoT [41], teaching
calculus [44], avatar designs and typographic arts in virtual
environments [45], [46], fostering Awareness of the Environ-
mental Impact of Agriculture [47], and presenting the Chinese
cultures [40].
Finally, we present the survey articles found in the collection
of research articles. Only one full survey article, two mini-
surveys, and three position papers [48], [49] exist. The long
survey written by Dionisio et al. [50] focuses on the develop-
ment of the metaverse, and accordingly discusses four aspects
of realism, ubiquity, interoperability, scalability. The two mini-
surveys focus on the existing applications and headsets for user
interaction in virtual environments, as well as various artistic
approaches to build artwork in VR [51], [52]. Regarding the
position papers, Ylipulli et al. [49] advocates design frame-
works for future hybrid cities and the intertwined relationship
between 3D virtual cities and the tangible counterparts, while
another theoretical framework classifies instance types in the
metaverse, by leveraging the classical Vitruvian principles
of Utilitas, Firmitas, and Venustas [53]. Additionally, as the
metaverse can serve as a collective and shared public space in
virtual environments, user privacy concerns in such emerging
spaces have been discussed in [48].
As we find a limited number of existing studies emphasising
the metaverse, we view that the metaverse research is still
in its infancy. Therefore, additional research efforts should
be extended in designing and building the metaverse. Instead
of selecting topics in randomised manners, we focus on two
critical aspects – technology and ecosystem, with the following
justifications. First, the technological aspect serves as the
critical factor to shape the metaverse. Figure 4 describes

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 5
Fig. 4. A timeline of the Metaverse Development from 1974 to 2020 (information source partially from T. Frey11and [22]), demonstrating the evolving
understanding of the metaverse once new technological infrastructures are introduced into the metaverse. With the evolving status of the metaverse, the
metaverse has gained more enriched communication media – text, graphics, 3D virtual worlds. Recently, AR applications demonstrate highly immersive digital
overlays in the world, such as Pok´
emon GO and Super Mario AR, while VR applications (e.g., VR Chat) allow users to be fully immersed in virtual worlds
for social gatherings. The landscape of the metaverse is dynamic. For instance, cryptoassets (e.g., CryptoKitties) have appeared as in-game trading, while
Alien Worlds encourages the users to earn non-fungible tokens (NFT) that can be converted into currencies in the real world12.
the timeline of the metaverse development. The metaverse
has experienced four transitions from text-based interactive
games, virtual open worlds, Massively Multiplayer Online
Game (MMOG), immersive virtual environments on smart
mobiles and wearables, to the current status of the metaverse.
Each transition is driven by the appearance of new technology
such as the birth of the Internet, 3D graphics, internet usage
at-scale, as well as hyperledger. It is obvious that technologies
serve as the catalysts to drive such transitions of cyberspaces.
In fact, the research community is still on the way to
exploring the metaverse development. Ideally, new technology
could potentially unlock additional features of the metaverse
and drive the virtual environments towards a perceived virtual
universe. Thus, we attempt to bridge various emerging tech-
nologies that could be conducive to the further progress of the
metaverse. After discussing the potential of various emerging
technologies, the game-based metaverse can open numerous
opportunities, and eventually may reach virtual environments
that is a society parallel to the existing one in the real
world, according to the three-stage metaverse as discussed in
Section I. Our survey paper, therefore, projects the design of
metaverse ecosystems based on the society in our real world.
The existing literature only focuses on fragmented issues such
as user privacy [48]. It is necessary to offer a holistic view of
the metaverse ecosystem, and our article serves this purpose.
Before we begin the discussion of the technologies and
the issues of ecosystems in Section III, here we pinpoint the
interdisciplinary nature of the metaverse. Thus, the survey
covers fourteen diversified topics linked to the metaverse.
Technical experts, research engineers, and computer scien-
tists can understand the latest technologies, challenges, and
research opportunities for shaping the future of the metaverse.
This article connects the relationship between the eight tech-
nological topics, and we did our utmost to demonstrate their
relationship. On the other hand, social scientist, economists,
avatar and content creators, digital policy makers, and gov-
ernors can understand the indispensable six building blocks
to construct the ecosystems of the metaverse, and how the
emerging technologies can bring impacts to both physical and
virtual worlds. In addition, other stakeholders who have al-
ready engaged in the metaverse, perhaps focusing on the game-
Fig. 5. The fourteen focused areas, under two key aspects of technology
and ecosystem for the metaverse. The key technologies fuel the ‘Digital Big
Bang’ from the Internet and XR to the metaverse, which support the metaverse
ecosystem.
oriented developments, can view our article as a reflection of
when technological catalysts further drive the evolution of the
metaverse, and perhaps the ‘Digital Big Bang’.
III. FRA ME WORK
Due to the interdisciplinary nature of the metaverse, this
section aims to explain the relationship between the fourteen
focused areas under two key categories of technologies and
ecosystems, before we move on to the discussion on each
focused area(s). Figure 5 depicts the focused areas under the
two categories, where the technology supports the metaverse
and its ecosystem as a gigantic application.
Under the technology aspect, i.e., the eight pillars for the
metaverse, human users can access the metaverse through
extended reality (XR) and techniques for user interactivity
(e.g., manipulating virtual objects). Computer vision (CV),
artificial intelligence (AI), blockchain, and robotics/ Internet-
of-Things (IoT) can work with the user to handle various

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 6
activities inside the metaverse through user interactivity and
XR. Edge computing aims to improve the performance of
applications that are delay-sensitive and bandwidth-hungry,
through managing the local data source as pre-processing data
available in edge devices, while cloud computing is well-
recognised for its highly scalable computational power and
storage capacity. Leveraging both cloud-based and edge-based
services can achieve a synergy, such as maximising the appli-
cation performance and hence user experiences. Accordingly,
edge devices and cloud services with advanced mobile network
can support the CV, AI, robots, and IoT, on top of appropriate
hardware infrastructure.
The ecosystem describes an independent and meta-sized
virtual world, mirroring the real world. Human users situated
in the physical world can control their avatars through XR
and user interaction technique for various collective activities
such as content creation. Therefore, virtual economy is a
spontaneous derivative of such activities in the metaverse. We
consider three focused areas of Social acceptability, security
and privacy, as well as trust and accountability. Analogue to
the society in the physical world, content creation and virtual
economy should align with the social norms and regulations.
For instance, the production in the virtual economy should
be protected by ownership, while such production outcomes
should be accepted by other avatars (i.e.,human users) in the
metaverse. Also, human users would expect that their activities
are not exposed to privacy risks and security threats.
The structure of the paper is as follows. Based on the
proposed framework, we review fourteen key aspects that
critically contribute to the metaverse. We first discuss the
technological aspect – XR (Section IV), user interaction in
XR and ubiquitous interfaces (Section V), robotics and IoT
(Section VI), artificial intelligence (Section VII), computer
vision (Section IX), hyperledger supporting various user ac-
tivities and the new economy in the metaverse market (Sec-
tion VIII), edge computing (Section X), the future network
fulfilling the enormous needs of the metaverse (Section XI).
Regarding the ecosystem on the basis of the aforementioned
technologies, we first discuss the key actors of the metaverse
– avatars representing the human users in Section XII. Next,
we discuss the content creation (Section XIII) and virtual
economy (Section XIV), and the corresponding social norms
and regulations – Social Acceptability (Section XV), Privacy
and Security (Section XVI), as well as Trust and Account-
ability (Section XVII). Finally, Section XVIII identifies the
grand challenges of building the metaverse, and discusses the
key research agenda of driving the ‘Digital Big Bang’ and
contributing to a unified, shared and collective space virtually.
IV. EXT EN DE D REA LI TY (XR)
Originated from the Milgram and Kishino’s Reality-
Virtuality Continuum [7], the most updated continuum has
further included new branches of alternated realities, leaning
towards the side of physical realities [54], namely MR [55]
and the futuristic holograms like the digital objects shown in
the Star Trek franchise [56]. The varied categories inside the
continuum allow human users to experience the metaverse
through various alternated realities across both the physical
and digital worlds [57]. However, we limited our discussion to
four primary types of realities that gain a lot of attention from
the academia and industry sectors [58]–[60]. This section be-
gins with the well-recognised domain of VR, and progressively
discusses the emerging fields of AR and its advanced variants,
MR and holographic technologies. This section also serves as
an introduction to how XR bridging the virtual entities with
the physical environments.
A. Virtual Reality (VR)
VR owns the prominent features of totally synthetic views.
The commercial VR headsets provide usual way of user
interaction techniques, including head tracking or tangible
controllers [60]. As such, users are situated in fully virtual
environments, and interacts with virtual objects through user
interaction techniques. In addition, VR is known as ‘the far-
thest end from the reality in Reality-Virtuality Continuum’ [7].
That is, the users with VR headsets have to pay full attention to
the virtual environments, and hence separate from the physical
reality [55]. As mentioned, the users in the metaverse will
create contents in the digital twins. Nowadays, commercial
virtual environments enable users to create contents, e.g.,
VR painting11. The exploration of user affordance can be
achieved by user interaction with virtual entities in a virtual
environment, for instance, modifying the shape of a virtual
object, and creating new artistic objects. Multiple Users in
such virtual environments can collaborate with each other in
real-time. This aligns with the well-defined requirements of
virtual environments: a shared sense of space, a shared sense
of presence, a shared sense of time (real-time interaction),
a way to communicate (by gesture, text, voice, etc.), and a
way to share information and manipulate objects [61]. It is
important to note that multiple users in a virtual world, i.e., a
subset of the metaverse, should receive identical information
as seen by other users. Users also can interact with each
other in consistent and real-time manners. In other words, how
the users should precept the virtual objects and the multi-
user collaboration in a virtual shared space would become
the critical factors. Considering the ultimate stage of the
metaverse, users situated in a virtual shared space should
work simultaneously with any additions or interactions from
the physical counterpart, such as AR and MR. The core of
building the metaverse, through composing numerous virtual
shared space, has to meld the simultaneous actions, among
all the objects, avatars representing their users, and their
interactions, e.g., object-avatars, object-object, and avatar-
avatar. All the participating processes in virtual environments
should synchronise and reflect the dynamic states/events of the
virtual spaces [62]. However, managing and synchronising the
dynamic states/events at scale is a huge challenge, especially
when we consider unlimited concurrent users collectively
act on virtual objects and interact with each other without
sensible latency, where latency could negatively impact the
user experiences.
11Six artists collaborate to do a VR painting of Star
Wars with Tilt Brush:https://www.digitalbodies.net/virtual-reality/
six-artists- vr-painting-star-wars/

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 7
B. Augmented Reality (AR)
Going beyond the sole virtual environments, AR delivers
alternated experiences to human users in their physical sur-
roundings, which focuses on the enhancement of our physical
world. In theory, computer-generated virtual contents can be
presented through diversified perceptual information channels,
such as audio, visuals, smell, and haptics [63]–[65]. The first-
generation of AR system frameworks only consider visual
enhancements, which aim to organise and display digital
overlays superimposing on top of our physical surroundings.
As shown in very early work in the early 1990s [66], a bulky
see-through display did not consider user mobility, which
requires users to interact with texts and 2D interfaces with
tangible controllers in a sedentary posture.
Since the very first work, significant research efforts have
been made to improve the user interaction with digital en-
tities in AR. It is important to note that the digital entities,
perhaps from the metaverse, overlaid in front of the user’s
physical surroundings, should allow human users to meld the
simultaneous actions (analogue to VR). As such, guaranteeing
seamless and lightweight user interaction with such digital
entities in AR is one of the key challenges, bridging human
users in the world physical with the metaverse [65]. Freehand
interaction techniques, as depicted in most science fiction
films like minority report12, illustrate intuitive and ready-to-
use interfaces for AR user interactions [58]. A well-known
freehand interaction technique named Voodoo Dolls [67] is
a system solution, in which users can employ two hands to
choose and work on the virtual contents with pinch gestures.
HOMER [68] is another type of user interaction solution that
provides a ray-casting trajectory from a user’s virtual hand,
indicating the AR objects being selected and subsequently
manipulated.
Moreover, AR will situate everywhere in our living envi-
ronments, for instance, annotating directions in an unfamiliar
place, and pinpointing objects driven by the user contexts [69].
As such, we can consider that the metaverse, via AR, will
integrate with our urban environment, and digital entities
will appear in plain and palpable ways on top of numerous
physical objects in urban areas. In other words, users with
AR work in the physical environments, and simultaneously
communicate with their virtual counterparts in the metaverse.
This requires significant efforts in the technologies of detection
and tracking to map the virtual contents displayed with the
corresponding position in the real environment [70]–[73]. A
more detailed discussion will be available in Section IX.
Touring Machine is considered as the first research prototype
that allows users to experience AR outdoors. The prototype
consists of computational hardware and a GPS unit loaded
on a backpack, plus a head-worn display that contains map
navigation information. The user with Touring Machine can
interact with the AR map through a hand-held touch-sensitive
surface and a stylus [74]. In contrast, the recent AR headsets
have demonstrated remarkable improvements, especially in
user mobility. Users with lightweight AR headsets can receive
visual and audio feedback cues indicating AR objects, but
12https://www.imdb.com/title/tt0181689/
Fig. 6. Displaying virtual contents with mature technologies: Public Large
Display (Left); Pico-projector attached on top of a wearable computer (Mid-
dle), and; mini-projector inside a smartphone (Right).
other sensory dimensions such as smell and haptics are still
neglected [58]. It is worth pinpointing that AR headsets are
not the only options to access the contents from the metaverse.
When we look at the current status of AR developments, AR
overlays, and even digital entities from the metaverse, can
be delivered by various devices, including but not limited to
AR headsets [58], [75], hand-held touchscreen devices [76],
ceiling projectors [77], and tabletops [78], Pico (wearable)
projectors [79] and so on. Nevertheless, AR headsets own
advantages over other approaches, in terms of the switch of
user attention and occupying users’ hands. First, human users
have to switch their attention between physical environments
and digital content on other types of AR devices. In contrast,
AR headsets enable AR overlays displayed in front of the
user’s sight [80], [81]. Second, the user’s hands will not be
occupied by the tangible devices as the computational units
and displays are mounted on the users’ heads. Such advantages
enable users with AR headsets to seamlessly experience ‘the
metaverse through an AR lens’. More elaboration of the user
interactivity is available in Section V.
C. Mixed Reality (MR)
After explaining the two extremes of the Reality–Virtuality
Continuum [82] – AR and VR, we attempt to discuss the
relationship between the metaverse and MR. Unfortunately,
there exists no commonly agreed definition for MR, but it is
crucial to have a common term that describes the alternated
reality situated between two extremes of AR and VR. Nev-
ertheless, the vastly different definitions can be summarised
into six working definitions [55], including the “traditional”
notion of MR in the middle space of the Reality–Virtuality
Continuum [82], MR as a synonym for AR [83], MR as
a type of collaboration [84], MR as a combination of AR
and VR [85], MR as an alignment of environments [86], a
“stronger” version of AR [87].
The above six definitions have commonly appeared in the
literature related to MR. The research community views that
MR stands between AR and VR, and allows user interaction
with the virtual entities in physical environments. It is worth-
while to mention that MR objects, supported by a strong ca-
pability of environmental understandings or situational aware-
ness, can work with other tangible objects in various physical
environments. For instance, a physical screwdriver can fit turn
digital entities of screws with slotted heads in MR, demon-
strating an important feature of interoperability between digital
and physical entities. In contrast, as observed in the existing
applications [58], AR usually simply displays information
overlaid on the physical environments, without considering

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 8
Fig. 7. Two holography types: (a) The reflection-based [94] approach can
reproduce colourful holography highly similar to the real object, and; (b)
The laser-driven approach can produce a sense of touch to the user’s skin
surface [95].
such interoperability. Considering such an additional feature,
MR is viewed as a stronger version of AR in a significant
number of articles that draw more connected and collaborating
relationships among the physical spatial, user interaction, and
virtual entities [58], [69], [88], [89].
From the above discussion, albeit we are unable to draw a
definitive conclusion to MR, MR is the starting point for the
metaverse, and certain properties of the six working definitions
are commonly shared between the metaverse and MR. We
consider that the metaverse begins with the digital twins that
connect to the physical world [9]–[14]. Human users subse-
quently start content creation in the digital twins [16]–[20].
Accordingly, the digitally created contents can be reflected in
physical environments, while human users expect such digital
objects to merge with our physical surroundings across space
and time [90]. Although we cannot accurately predict how the
metaverse will eventually impact our physical surroundings,
we see the existing MR prototypes enclose some specific
goals such as pursuing scenes of realism [91], bringing senses
of presence [92], creating empathetic physical spatial [93].
These goals can be viewed as an alignment with the metaverse
advocating that multiple virtual worlds work complementary
with each other [9].
D. Large Display, Pico-Projector, Holography
Based on the existing literature, this paragraph aims to make
speculation for the ways of bringing the uniquely created
contents inside the virtual environments (ultimately metaverse)
back to the physical counterparts in the shared public space. As
the social acceptability of mobile headsets in public spaces is
still questionable [96], we lack evidence that mobile headsets
will act as the sole channel for delivering metaverse contents
into the public space. Instead, other mature technologies such
as large displays and pico-projectors may serve as a channel
to project pixels into our real world. Figure 6 depicts three
examples. Large displays13, and pico-projectors [79] allow
users without mobile headsets to view digital entities with
a high degree of realism. In addition, miniature projectors
embedded inside smartphones, e.g., MOVI Phone14, allow
13A giant 3D cat has taken over one of Tokyo’s biggest billboards: https:
//edition.cnn.com/style/article/3d-cat- billboard-tokyo/index.html
14MOVI-phone: https://moviphones.com/
content sharing anytime and anywhere. It is also worth noting
that smartphones are the most ubiquitous devices nowadays.
Finally, we discuss the possibility of holographic technology
emphasising enriched communication media exceeding the 2D
displays [97] and pursuing true volumetric displays (showing
images or videos) that show no difference from our everyday
objects. The current holographic technology can be classified
into two primary types: reflection-based and laser-driven holo-
graph15. A recent work [98] demonstrated the feasibility of
colourful volumetric display on bulky and sedentary devices,
with practical limitations of low resolution that could impact
the user perceptions to realism. However, the main advantage
of reflection-based holography is to generate the colourful
holograms with colour reproduction highly similar to real-
life objects [94] (Figure 7(a)). On the other hand, Plasma
Fairies [95] is a 3D aerial hologram that can be sensed by
the users’ skin surfaces, though the devices can only produce
plasmonic emission in a mid-air region no larger than 5 cm3
(Figure 7(b)). We conjecture that if technology breakthrough
allows such volumetric 3D objects to appear in the real world
ubiquitously, it will come as no surprise that the metaverse
can merge with our living urban, as illustrated in Figure 3
(top-right corner), and provide a strong sense of presence
to the stakeholders in urban areas. However, holographic
technology suffers from three key weaknesses in the above
works, including limited resolution, display size, as well as
device mobility. Thus, overcoming such weaknesses becomes
the critical turning point of delivering enriched 3D images in
the real world.
V. USER INTERACTIVITY
This section first reviews the latest techniques that enable
users to interact with digital entities in physical environments.
Then, we pinpoint the existing technologies that display digital
entities to human users. We also discuss the user feedback
cues as well as the haptic-driven telepresence that connects
the human users in physical environments, avatars in the meta-
verse, and digital entities throughout the advanced continuum
of extended reality.
A. Mobile Input Techniques
As the ultimate stage of the metaverse will interconnect
both the physical world and its digital twins, all human users
in the physical world can work with avatars and virtual objects
situated in both the metaverse and the MR in physical envi-
ronments, i.e., both the physical and virtual worlds constantly
impact each other. It is necessary to enable users to interact
with digital entities ubiquitously. However, the majority of the
existing metaverse only allows user interactions with the key-
boards and mice duo, which cannot accurately reflect the body
movements of the avatar [22]. Also, such bulky keyboards and
mice are not designed for mobile user interaction, and thus
enforce users to maintain sedentary postures (e.g., sitting) [58],
[69].
Albeit freehand interaction features intuitiveness due to
barehanded operations [58] and further achieve object pointing
15https://mitmuseum.mit.edu/holography- glossary

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 9
and manipulation [99], most freehand interactions rely on
computer vision (CV) techniques. Thus, accurate and real-
time recognition of freehand interaction is technically de-
manding, even the most fundamental mid-air pointing needs
sufficient computational resources [100]. Insufficient computa-
tional resources could bring latency to user actions and hence
deteriorate the user experience [101]. Apart from CV-based
interaction techniques, the research community search vastly
diversified input modality to support complicated user interac-
tion, including optical [102], IMU-driven [103], Pyroelectric
Infrared [104], electromagnetic [105], capacitive [106], and
IMU-driven user interactions [103]. Such alternative modal-
ities can capture user activities and hence interact with the
digital entities from the metaverse.
We present several existing works to illustrate the mobile
input techniques with alternative input modals, as follows.
First, the human users themselves could become the most
convenient and ready-to-use interaction surface, named as on-
body user interaction [58]. For instance, ActiTouch [106]
owns a capacitive surface attached to the user’s forearm.
The electrodes in ActiTouch turn the user’s body into a
spacious input surface, which implies that users can perform
taps on their bodies to communicate with other stakeholders
across various digital entities in the metaverse. Another similar
technique [107] enriched the set of input commands, in which
users can interact with icons, menus, and other virtual objects
as AR overlaid on the user’s arm. Additionally, such on-body
interaction can be employed as a solution for interpersonal
interactions that enable social touch remotely [108], [109].
Such on-body user interaction could enrich the communication
among human users and avatars. The latest technologies of on-
body interaction demonstrate the trend of decreasing device
size, ranging from a palm area [110]–[112] to a fingertip [113].
The user interaction, therefore, becomes more unnoticeable
than the aforementioned finger-to-arm interaction. Neverthe-
less, searching alternative input modalities does not mean that
the CV-based techniques are not applicable. The combined
use of alternative input modals and CV-based techniques can
maintain both intuitiveness and the capability of handling time-
sensitive or complicated user inputs [58]. For instance, a CV-
based solution works complementary to IMU sensors. The
CV-based technique determines the relative position between
the virtual objects user hands in mid-air, while the IMU
sensors enable subtle and accurate manipulation of virtual
objects [103].
Instead of attaching sensors to our body, another alternative
is regarded as digital textile. Digital textile integrates novel
material and conductive threads inside the usual fabrics, which
supports user interactions with 2D and 3D user interfaces
(UIs). Research prototypes such as PocketThumb [114] and
ARCord [115] convert our clothes into user interfaces with the
digital entities in MR. PocketThumb [114] is a smart fabric lo-
cated at a front trouser pocket. Users can exert taps and touches
on the fabrics to perform user interaction, e.g., positioning a
cursor during pointing tasks with 3D virtual objects in MR.
Also, ARCord [115] is a cord-based textile attached to a jacket,
and users can rub the cord to perform menu selection and
ray-casting on virtual objects in various virtual environments.
Remarkably, technology giants have invested in this area to
support the next generation of mobile user inputs. For example,
Google has launched the Jacquard project [116] that attempts
to produce smart woven at an affordable price and in a
large scale. As a result, the smart woven can merge with
our daily outfits such as jackets and trousers, supporting user
inputs anywhere and anytime. Although we cannot discuss
all types of mobile inputs due to limited space, the research
community is searching for more natural, more petite, subtle
and unnoticeable interfaces for mobile inputs and alternative
input modals in XR, e.g., Electroencephalography (EEG) and
Electromyography (EMG) [117], [118].
B. New Human Visions via Mobile Headsets
Mobile headsets, as discussed in Section IV-B, owns key
advantages such as aligned views between physical and virtual
realities, and user mobility, which can be regarded as an
emerging channel to display virtual content ubiquitously [96].
As VR mobile headsets will isolate human users from the
physical realities [60] and its potential dangers in public
spaces [119], in this section, we discuss the latest AR/MR
headsets that are designed for merging virtual contents in
physical environments.
Currently, the user immersiveness in the metaverse can be
restricted by limited Field of View (FOV) on AR/MR mobile
headsets. Narrowed FOVs can negatively influence the user
experience, usability, and task performance [80], [120]. The
MR/AR mobile headsets usually own FOVs smaller than 60
degrees. The limited FOV available on mobile headsets is far
smaller than the typical human vision. For instance, the FOV
can be equivalent to a 25-inch display 240 cm away from the
user’s view on the low-specification headsets such as Google
Glass. The first generation of Microsoft Hololens presents a
30 X 17-degree FOV, which is a similar size as a 15-inch 16:9
display located around 60 cm away from the user’s egocentric
view. We believes that the restricted view will be eventually
resolved by the advancement of display technologies, for
instance, the second generation of Microsoft Hololens owns
an enlarged display having 43 X 29-degree FOV. Moreover,
the bulky spectacle frames on MR headsets, such as Microsoft
Hololens, can occlude the users’ peripheral vision. As such,
users can reduce their awareness of incoming dangers as well
as critical situations [121]. Thus, other form factors such as
contact lens can alleviate such drawbacks. A prototypical AR
display in the form factor of contact lens [122], albeit offering
low-resolution visuals to users, can provide virtual overlays,
e.g., top, down, left, right directions in navigation tasks.
The remaining section discusses the design challenges of
presenting virtual objects through mobile headsets, how to
leverage the human visions in the metaverse. First, one design
strategy is to leverage the users’ peripheral visual field [125]
that originally aims to identify obstacles, avoid dangerous
incidents, and measure foot placements during a wide range
of locomotive activities, e.g., walking, running, driving and
other sport activities [126]. Combined with other feedback
cues such as audio and haptic feedback, users can sense the
virtual entities with higher granularity [125]. Recent works

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 10
Fig. 8. Displaying virtual contents overlaid on top of physical environments:
a restaurant (indoor, left) [123], a street (outdoor, right) [124].
also present this design strategy by displaying digital overlays
at the edge areas of the FOVs on MR/AR mobile headsets [75],
[80], [127], [128]. The display of virtual overlays at edge
areas can result in practical applications such as navigation
instructions of straight, left, and right during a navigation task
on AR maps [80]. A prominent advantage of such designs
is that the virtual overlays on the users’ peripheral visions
highly aligns with the locomotive activities. As such, users can
focus on other tasks in the physical world, without significant
interruption from the virtual entities from the metaverse. It
is important to note that other factors should be considered
together when presenting virtual overlays within the users’
visual fields, such as colour, illumination [129], content legi-
bility, readability [130], size, style [131], visual fatigue [132],
movement-driven shakiness [133]. Also, information overflow
could ruin the user ability to identify useful information.
Therefore, appropriate design of information volume and
content placements (Figure 8) is crucial to improving the
effectiveness of displaying virtual overlays extracted from the
metaverse [123], [124], [134], [135].
C. The importance of Feedback Cues
After considering the input and output techniques, the
user feedback cues is another important dimension for user
interactivity with the metaverse. We attempt to explain this
concept with the fundamental elements in 3D virtual worlds –
user interaction with virtual buttons [136]–[138]. Along with
the above discussions, virtual environments can provide highly
adaptive yet realistic environments [139], but the usability
and the sense of realism are subject to the proper design of
user feedback cues (e.g., visual, audio, haptic feedback) [140].
The key difference between touchscreen devices and virtual
environments is that touchscreen devices offer haptic feedback
cues when a user taps on a touchscreen, thus improving user
responsiveness and task performances [141]. In contrast, the
lack of haptic feedback in virtual environments can be com-
pensated in multiple simulated approaches [142], such as vir-
tual spring [143], redirected tool-mediated manipulation [144],
stiffness [145], object weighting [146]. With such simulated
haptic cues, the users can connect the virtual overlays of the
buttons) with the physical metaphors of the buttons [147]. In
other words, the haptic feedback not only works with the visual
and audio cues, and further acts as an enriched communication
signal to the users during the virtual touches (or even the
interaction) with virtual overlays in the metaverse [148]. More
importantly, such feedback cues should follow the principle of
Fig. 9. The key principles of haptic devices that support user interaction
with various tangible and virtual objects in the metaverse (Image source
from [167]).
user mobility as mentioned in Section V-A. The existing works
demonstrate various form factors exoskeletons [149], [150],
gloves [151], [152], finger addendum [153], [154], smart wrist-
bands [155], by considering numerous mechanisms including
air-jets [156], ultrasounds [157]–[159], and laser [160], [161].
In addition, the full taxonomy of mobile haptic devices is
available in [162].
After compensating the missing haptic feedback in virtual
environments, it is important to best utilise various feedback
cues and achieve multi-modal feedback cues (e.g., visual,
auditory, and haptic) [163], in order to improve the user
experiences [164], the user’s responsiveness [143], task ac-
curacy [140], [165], the efficiency of virtual object acquisi-
tion [136], [165] in various virtual environments. We also
consider inclusiveness as an additional benefit of leveraging
haptic feedback in virtual environments, i.e., the visually
impaired individuals [166]. As the prior works on the multi-
modal feedback cues do not consider the new enriched in-
stance to appear in varying scenarios inside the metaverse,
it is worthwhile to explore the combination of the feedback
modals further, and introduce new modals such as smell and
taste [63].
D. Telepresence
The discussion in previous paragraphs can be viewed as
the stimuli to achieve seamless user interaction with virtual
objects as well as other avatars representing other human users.
To this end, we have to consider the possible usage of such
stimuli that paves the path towards telepresence through the
metaverse. Apart from designing stable haptic devices [168],
the synchronisation of such stimuli is challenging. According
to the Weber-Fechner Law that describes “the minimum time
gap between two stimuli” in order to make user feels the two
stimuli are distinguishable. Therefore, the research community
employs the measures of Just Noticeable Difference (JND) to
quantify the necessary minimum time gap [169]. Considering
the benefits of including haptic feedback in virtual environ-
ments, as stated in Section V-C, the haptic stimuli should be
handled separately. As such, transmitting such a new form of

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 11
haptic data can be effectively resolved by Deadband compres-
sion techniques (60% reduction of the bandwidth) [170]. The
technique aims to serve cutaneous haptic feedback and further
manage the JND, in order to guarantee the user can precept
distinguishable haptic feedback.
Next, the network requirements of delivering haptic stimuli
would be another key challenge. The existing 4G communica-
tion technologies can barely afford AR and VR applications.
However, managing and delivering the haptic rendering for
user’s sensing the realism of virtual environments in a sub-
tle manner are still difficult with the existing 4G network.
Although 5G network features with low latency, low jitter,
and high bandwidth, haptic mobile devices, considered as
a type of machine-type communication, may not be able
to adopt in large-scale user interactivity through the current
design of 5G network designated for machine-to-machine
communication [172] (More details in Section VI. Addition-
ally, haptic mobile devices is designed for the user’s day-long
activities anywhere when the network capacity has fulfilled
the aforementioned requirements. Thus, the next important
issue is to tackle the constraints of energy and computational
resources on mobile devices [101]. Apart from reducing the al-
gorithm complexity of haptic rendering, an immediate solution
could be offloading such haptic-driven computational tasks
to adjacent devices such as cloud servers and edge devices.
More detailed information on advanced networks as well as
edge and cloud computing are available in Section XI and X,
respectively.
Although we expect that new advances in electronics and
future wireless communications will lead to real-time inter-
actions in the metaverse, the network requirements would
become extremely demanding if the metaverse will serve
unlimited concurrent users. As such, network latency could
hurt the effectiveness of such stimuli and hence the sense of
realism. To this end, a visionary concept of Tactile Internet is
coined by Fettweis [173], which advocates the redesign of the
backbone of the Internet to alleviate the negative impacts from
latency and build up ultra-reliable tactile sensory for virtual
objects in the metaverse [174]–[176]. More specifically, 1 ms
is expected as the maximum latency of Tactile Internet, which
facilitates real-time haptic feedback for the sake of various
operations during the telepresence [177]. It is important to
note that the network latency is not the only source. Other
latency sources could be caused by the devices, i.e., on-
device latency [178], [179]. For instance, the glass-to-glass
latency, representing the round-trip latency from video taken
by a smartphone camera to a virtual overlay that appeared
in a smartphone screen, is 19.18 ms [180], far exceeding the
ideal value of 1 ms for the Tactile Internet. The aggregation
of latency could further deteriorate the user perceptions with
virtual environments in the metaverse [178]. Therefore, we
call for additional research attention in this area for building
seamless yet realistic user interaction [167] with various
entities linked to the metaverse, as illustrated in Figure 9.
VI. IN TE RN ET-OF -THI NG S (IOT) AND RO BOT IC S
According to Statista [181], by 2025, the total IoT connected
devices worldwide will reach 30.9 billion, with a sharp jump
from the 13.8 billion expected in 2021. Meanwhile, the
diversity of interaction modalities is expanding. Therefore,
many observers believe that integrating IoT and AR/VR/MR
may be suitable for multi-modal interaction systems to achieve
compelling user experiences, especially for non-expert users.
The reason is that it allows interaction systems to com-
bine the real-world context of the agent and immersive AR
content [182]. To align with our focused discussion on the
metaverse, this section focuses on the virtual environments
under the spectrum of extended reality, i.e., data management
and visualisation, and human-IoT interfacing. Accordingly, we
elaborate on the impacts of XR on IoT, autonomous vehicles,
and robots/drones, and subsequently pinpoint the emerging
issues.
A. VR/AR/MR-driven human-IoT interaction
The accelerating availability of smart IoT devices in our
everyday environments offers opportunities for novel services
and applications that can improve our quality of life. However,
miniature-sized IoT devices usually cannot accommodate tan-
gible interfaces for proper user interaction [183]. The digital
entities under the spectrum of XR can compensate for the
missing interaction components. In particular, users with see-
through displays can view XR interfaces in mid-air [184].
Additionally, some bulky devices like robot arms, due to
limitations of form factors, would prefer users to control
the devices remotely, in which XR serves as an on-demand
controller [185]. Users can get rid of tangible controllers,
considering that it is impossible to bring a bundle of controllers
for numerous IoT devices. Virtual environments (AR/MR/XR)
show prominent features of visualising invisible instances
and their operations, such as WiFi [186] and user personal
data [187]. Also, AR can visualise the IoT data flow of smart
cameras and speakers to the users, thus informing users about
their risk in the user-IoT interaction. Accordingly, users can
control their IoT data via AR visualisation platforms [187].
There are several key principles to categorise the
AR/VR/MR-directed IoT interaction systems. Figure 10 shows
three models defined according to the scale and category of
the rendered AR content. Mid-air icons, menus, and virtual
3D objects allow users to control IoT devices with natural
gestures [171]. Figure 12 offers four models depicted accord-
ing to the controllability of the IoT device and the identifier
entity. In short, virtual overlays in AR/MR/XR can facilitate
data presentation and interfacing the human-IoT interaction.
Relatedly, a number of recent works have been proposed in
this direction. For example, [188] presents V.Ra, a visual and
spatial programming system that allows the users to perform
task authoring with an AR hand-held interface and attach the
AR device onto the mobile robot, which would execute the
task plan in a what-you-do-is-what-robot-does (WYDWRD)
manner. Moreover, flying drones, a popular IoT device, have
been increasingly employed in XR. In [189], multiple users
can control a flying drone remotely and work collaboratively
for searching tasks outdoors. Pinpointfly [190] presents a
hand-held AR application that allows users to edit a flying
drone’s motions and directions through enhanced AR views.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 12
Fig. 10. Three basic AR interaction models: (a) The Floating Icons model, with the user gazing at the icon. (b) The WIM model in scale mode, with a
hologram being engaged with. (c) The floating menu model, with three active items and three inactive items [171].
Similarly, SlingDrone [191] leverages MR user interaction
through mobile headsets to plan the flying path of flying
drones.
B. Connected vehicles
As nowadays vehicles are equipped with powerful computa-
tional capacity and advanced sensors, connected vehicles with
5G or even more advanced networks could go beyond the
vehicle-to-vehicle connections, and eventually connect with
the metaverse. Considering vehicles are semi-public spaces
with high mobility, drivers and passengers inside vehicles
can receive enriched media. With the above incentive, the
research community and industry are striving to advance the
progress of autonomous driving technologies in the era of AI.
Connected vehicles serves as an example of IoT devices as
autonomous vehicles could become the most popular scenarios
for our daily commute. In recent years, significant progress
has been made owing to the recently emerging technologies,
such as AR/MR [192], [193]. AR/MR play an important
role in empowering the innovation of autonomous driving.
To date, AR/MR has been applied in three directions for
autonomous driving [194]. First of all, AR/MR helps the
public (bystanders) understand how autonomous vehicles work
on the road, by offering visual cues such as the vehicle
directions. With such understandings, pedestrian safety has
been enhanced [195]. To this end, several industrial appli-
cations, such as Civil Maps16, applied AR/MR to provide a
guide for people to understand how an autonomous driving
vehicle navigates in the outdoor environment. For instance,
it shows how the vehicle detects the surroundings, vehicles,
traffic lights, pedestrians, and so on. The illustration with
AR/MR/XR or even the metaverse can build trust with the
users with connected vehicles [196]. In addition, some AR-
supported dynamic maps can also help drivers to make good
decisions when driving on the road. Second, AR/MR help
to improve road safety. For instance, virtual entities appear
in front of the windshield of vehicles, and such entities can
augment the information in the physical world to enhance
the user awareness to the road conditions. It is important to
note such virtual entities are considered as a low-cost and
convenient solution, in comparison to largely modified the
physical road infrastructure. The latest work also pinpoints
16https://civilmaps.com/
the concept of digital twins to enhance road safety, especially
for vulnerable road users [197], instead of inviting the human
users to work on risky tasks physically. For instance, the Mcity
Test Facility at the University of Michigan17 applies AR to
test the driving car. In the platform, the testing and interaction
between a real test vehicle and the virtual vehicles are created
to test driving safety. In such a MR world, an observer can see
a real vehicle passing and stopping at the intersection with the
virtual vehicles at the traffic light. Last but not least, AR/MR
have improved the vehicle navigation and user experience. For
example, WayRay18 develops an AR-based navigation system
that helps to improve road driving safety. The highlight of
this technique is that it alleviates the need for the drivers to
rely too much on gauges when driving. Surprisingly, WayRay
provides the driver with highly precise route and environment
information in real-time. Most recent research also demon-
strates the needs of shared views among connected vehicles
to enhance user safety, for instance, the view of a front car
is shared to the car(s) at the back [198]. From the above, we
see the benefits of introducing virtual entities on connected
vehicles and road traffic. Perhaps the metaverse can transform
such driving information into interesting animation without
compromising road safety.
Recent examples also shed lights on the integration between
intelligent vehicles and virtual environments. For Invisible-to-
Visible (I2V) from Nissian19 is a representative attempt to
build the metaverse platform where an AR interface is de-
signed to connect the physical and virtual worlds together such
that the information invisible to the drivers can be visible. As
shown in Figure 11, I2V employs several systems to provide
rich information from the inside and outside of vehicle. Specif-
ically, I2V first adopts the omni-sensing technology to gather
data in real-time from the traffic and the surrounding vehicles.
Meanwhile, the metaverse system seamlessly analyses the road
status from the real-time information. Based on the analysis,
I2V then identifies the driving conditions around the vehicle
immediately. Lastly, the digital twin of the vehicles, drivers,
the buildings, and the environment is created via data collected
from the omni-sensing system. In such a way, the digital twin
17https://record.umich.edu/articles/augmented- reality-u- m-improves\
protect\penalty-\@M- driverless-vehicle-testing/
18https://wayray.com/#who-we-are
19https://www.nissan-global.com/EN/TECHNOLOGY/OVERVIEW/i2v.
html

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 13
Fig. 11. (a) The I2V metaverse of Nissian for assisting driving. I2V can
connect drivers, passengers with the people all across the world.(b) The
Hyundai Mobility Adventure (HMA) showcasing the future life.
can be used to analyse the human-city interaction [69] through
the perspective of road traffic. The shared information driven
by the user activities can further connect to the metaverse.
As a result, the metaverse generates the information through
the XR interfaces, as discussed in Section IV or the vehicle
windshields. To sum up, the digital transformation with the
metaverse can deliver human users enriched media during
their commutes. In addition, I2V helps driving in two aspects.
The first is visualising the invisible environment for a more
comfortable drive. The metaverse system enables displaying
the road information and hidden obstacles, traffic congestion,
parking guidance, driving in the mountain, driving in poor
weather conditions, etc. Meanwhile, I2V metaverse system
visualises virtual human communication via MR. For instance,
it provides a chance for family members from anywhere in
the world to join the metaverse as avatars. It also provides a
tourism scenario where a local guide can join the metaverse
to guide the driver.
Furthermore, the Roborace metaverse20 is another platform
blending the physical world with a virtual world where AR
generates the virtual obstacles to interact with the track.
Hyundai Motor21 also launched ‘Hyundai Mobility Adventure
(HMA)’ to showcase the future lifestyle in the metaverse. The
HMA is a shared virtual space where various users/players,
which are represented as ‘avatars’, can meet and interact with
each other to experience mobility. Through the metaverse
platform, the participants can customise their ‘avatars’ and
imaginatively interact with each other.
C. Robots with Virtual Environments
Virtual environments such as AR/VR/MR are good solution
candidates for opening the communication channels between
robots and virtual environments, due to their prominent feature
of visualising contents [199]. Furthermore, various industrial
examples integrate virtual environments to enable human users
to understand robot operations, such as task scenario analysis
and safety analysis. Therefore, human users build trust and
confidence with the robots, leading to the paradigm shift
towards human-robot collaboration [200]. Meanwhile, to date,
research studies focus on the user perception with robots
and the corresponding interface designs with virtual environ-
ments [185], [201], [202]. Also, human users with V.Ra [188]
can collaboratively develop task plans in AR environments and
20https://roborace.com/
21https://www.hyundai.news/eu/articles/press-releases/
hyundai-vitalizes- future-mobility- in-roblox- metaverse-space.html
Fig. 12. Four interaction models proposed in [182], categorised by whether
an agent can control the IoT device through AR (c,d) or not (a,b), and whether
an IoT device (a,c) or another entity (b,d) functions as an AR identifier.
program mobile robots to interact with stationary IoTs in their
physical surroundings.
Nowadays, the emerging MR technology serves as com-
munication interfaces with humanoids in workspace [203],
with high acceptance levels to collaborative robots [204].
In our daily life, robots can potentially serve as our
friends [205] companion devices [206], services drone [207],
caring robots [208], [209], an inspector in public spaces [210],
home guardian (e.g., Amazon Astro22), sex partners [211]–
[213], and even a buddy with dogs [214], as human users
can adapt natural interactions with robots and drones [215].
It is not hard to imagine the robots will proactively serve
our society, and engage spontaneously in a wide variety of
applications and services.
The vision of the metaverse with collaborative robots is not
only limited to leveraging robots as a physical container for
avatars in the real world, and also exploring design opportuni-
ties of our alternated spatial with the metaverse. Virtual envi-
ronments in the metaverse can also become the game changer
to the user perception with collaborative robots. It is important
to note that the digital twins and the metaverse can serve as
a virtual testing ground for new robot designs. The digital
twins, i.e., digital copies of our physical environments, allow
robot and drone designers to examine the user acceptability
of novel robot agents in our physical environments. What
are the changes in user perception to our spatial environment
augmented by new robot actors, such as alternative humanoids
and mechanised everyday objects? In [216], designers evaluate
the user perceptions to the mechanised walls in digital twins of
living spaces, without actual implementation in the real world.
The mechanised walls can dynamically orchestrate with user
activities of various contexts, e.g., additional walls to separate
a user from the crowd, who prefers staying alone at works, or
lesser walls for social gatherings.
22https://www.aboutamazon.com/news/devices/
meet-astro- a-home- robot-unlike-any-other

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 14
VII. ARTI FIC IA L INT ELLIGENCE
Artificial intelligence (AI) refers to theories and tech-
nologies that enable machines to learn from experience and
perform various kinds of tasks, similar to intelligent crea-
tures [217]–[219]. AI was first proposed in 1956. In recent
years, it has achieved state-of-the-art performance in vari-
ous application scenarios, including natural language process-
ing [220], [221], computer vision [222], [223], and recom-
mender systems [224], [225]. AI is a broad concept, including
representation, reasoning, and data mining. Machine learning
is a widely used AI technique, which enables machines to
learn and improve performance with knowledge extracted from
experience. There are three categories in machine learning:
supervised learning, unsupervised learning, and reinforcement
learning. Supervised learning requires training samples to be
labelled, while unsupervised learning and reinforcement learn-
ing are usually applied on unlabelled data. Typical supervised
learning algorithms includes linear regression [226], random
forest [227], and decision tree [228]. K-means [229], principle
component analysis (PCA) [230], and singular value de-
composition (SVD) [231] are common unsupervised learning
algorithms. Popular reinforcement learning algorithms include
Q-learning [232], Sarsa [233], and policy gradient [234].
Machine learning usually requires selecting features manually.
Deep learning is involved in machine learning, which is in-
spired by biological neural networks. In deep neural networks,
each layer recieves input from the previous layers, and outputs
the processed data to the subsequent layers. Deep learning
is able to automatically extract features from a large amount
of data. However, deep learning also requires more data than
conventional machine learning algorithms to offer satisfying
accuracy. Convolutional neural network (CNN) [235], recur-
rent neural network (RNN) [236] are two typical and widely
used deep learning algorithms.
There is no doubt that the main characteristic of the emerg-
ing metaverse is the overlay of unfathomably vast amounts
of sophisticated data, which provides opportunities for the
application of AI to release operators from boring and tough
data analysis tasks, e.g., monitoring, regulating, and planning.
In this section, we review and discuss how AI is used in
the creation and operation of the metaverse. Specifically, we
classify AI applications in the metaverse into three categories:
automatic digital twin, computer agent, and the autonomy of
avatar.
A. Automatic Digital Twin
There are three kinds of digitisation, including digital
model, digital shadow, and digital twin [237]. The digital
model is the digital replication of a physical entity. There is no
interaction between the metaverse and the physical world. The
digital shadow is the digital representation of a physical entity.
Once the physical entity changes, its digital shadow changes
accordingly. In the case of a digital twin, the metaverse and the
physical world are able to influence each other. Any change
on any of them will lead to a change on the other one. In the
metaverse, we focus on this third kind of digitisation.
Fig. 13. Illustration of autonomous digital twin with deep learning.
Digital twins are digital clones with high integrity and
consciousness for physical entities or systems and keeps
interacting with the physical world [237]. These digital clones
could be used to provide classification [238], [239], recogni-
tion [240], [241], prediction [242], [243], and determination
services [244], [245] for their physical entities. Human in-
terference and manual feature selection are time-consuming.
Therefore, it is necessary to automate the process of data
processing, analysis, and training. Deep learning can automat-
ically extract knowledge from a large amount of sophisticated
data and represent it in various kinds of applications, without
manual feature engineering. Hence, deep learning has great
potential to facilitate the implementation of digital twins. Jay et
al. [246] propose a general autonomous deep learning-enabled
digital twin, as shown in Figure 13. In the training phase,
historical data from both the metaverse and physical systems
are fused together for deep learning training and testing. If the
testing results meet the requirement, the autonomous system
will be implemented. In the implementation phase, real-time
data from the metaverse and physical systems are fused for
model inference.
Smart healthcare requires interaction and convergence be-
tween physical and information systems to provide patients
with quick-response and accurate healthcare services. Hence,
the concept of digital twin is naturally applicable to smart
healthcare. Laaki et al. [247] designs A verification prototype
for remote surgery with digital twins. In this prototype, a
digital twin is created for a patient. All surgery operations
on the digital twin done by doctors will be repeated on the
patient with a robotic arm. The prototype is also compatible
with deep learning components, e.g., intelligent diagnosis and
healthy prediction. Liu et al. apply learning algorithms for
real-time monitoring and crisis warning for older adults with
their digital twins [248].
Nowadays, more IoT sensors are implemented in cities
to monitor various kinds of information and facilitate city
management. Moreover, building information models (BIM)
are getting more accurate [249]. By combining the IoT big
data and BIM, we could create digital twins with high quality
for smart cities. Such a smart-city digital twin will make urban
planning and managing easier. For example, we could learn

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 15
about the impact of air pollution and noise level on people’s
life quality [250] or test how traffic light interval impacts the
urban traffic [251]. Ruohomaki et al. create a digital twin for
an area in urban to monitor and predict the building energy
consumption. Such a system could also be used to help to
select the optimisation problem of the placement of solar
panels [252].
Industrial systems are very complex and include multiple
components, e.g., control strategy, workflow, system parame-
ter, which is hard to achieve global optimisation. Moreover,
data are heterogeneous, e.g., structured data, unstructured
data, and semi-structured data, which makes deep learning-
driven digital twin essential [253]. Min et al. design a digital
twin framework for the petrochemical industry to optimise
the production control [254]. The framework is constructed
based on workflow and expert knowledge. Then they use
historical production data to train machine learning algorithms
for prediction and optimise the whole system.
B. Computer Agent
Computer agent, also known as Non-player Character
(NPC), refers to the character not controlled by a player. The
history for NPCs in games could be traced back to arcade
games, in which the mobility patterns of enemies will be
more and more complex along with the level increasing [255].
With the increasing requirements for realism in video games,
AI is applied for NPCs to mimic the intelligent behaviour
of players to meet players’ expectations on entertainment
with high quality. The intelligence of NPCs is reflected in
multiple aspects, including control strategy, realistic character
animations, fantastic graphics, voice, etc.
The most straight and widely adopted model for NPC
to respond to players’ behaviour is finite state machines
(FSM) [256]. FSM assumes there are finite states for an object
in its lifecycle. There are four components in FSM: state,
condition, action, next state. Once the condition is met, the
object will take a new action and change its current state to the
next state. Behaviour trees and decision trees are two typical
FSM-based algorithms for NPCs to make decisions in games,
in which each node denotes a state and each edge represents
an action [257]–[260]. FSM-based strategies are very easy to
realise. However, FSM is poor at scalability, especially when
the game environment becomes complex.
Support vector machine is a classifier with the maximum
margin between different classes, which is suitable for con-
trolling NPCs in games. Pedro et al. propose a SVM-based
NPC controller in a shooter game [261]. The input is a three-
dimensional vector, including left bullets, stamina, and near
enemies. The output is the suggested behaviour, e.g., explore,
attack, or run away. Obviously, the primary drawback of such
an algorithm is limited state and behaviour classes and the
flexibility in decision-making.
Reinforcement learning is a classic machine learning algo-
rithm on decision-making problems, which enables agents to
automatically learn from the interaction experience with their
surrounding environment. The agent behaviours will be given
corresponding rewards. The desired behaviours are with a
higher reward. Due to its excellent performance, reinforcement
learning has been widely adopted in many games, e.g., shooter
games [262] and driving games [263]. It is worth noting that
the objective of NPC designing is to increase the entertainment
of the game, instead of maximising the ability of NPCs to
beat human players [264]. Hence, the reward function could
be customised according to the game objective [265]. For
example, Glavin et al. develop a skill-balancing mechanism
to dynamically adjust the skill level of NPCs according to
players performance based on reinforcement learning [266].
When the games are getting more and more complex,
from 2D to 3D, the agent state becomes countless. Deep
reinforcement learning, the combination of neural network and
reinforcement learning is proposed to solve such problems.
The most famous game based on deep reinforcement learning
is chess with AlphaGo developed by DeepMind in 2015 [267].
The state of chess is denoted as a matrix. Through the process
of neural networks, the AlphaGo outputs the action with the
highest possibility to win.
C. Autonomy of Avatar
Avatar refers to the digital representation of players in the
metaverse, where players interact with the other players or
the computer agents through the avatar [268]. A player may
create different avatars in different applications or games.
For example, the created avatar may be like a human shape,
imaginary creatures, or animals [269]. In social communica-
tion, relevant applications that require remote presence, facial
and motion characteristics reflecting the physical human are
essential [270]. Existing works in this area mainly focus on
two problems: avatar creation and avatar modelling.
To create more realistic virtual environments, a wide variety
of avatar representations are necessary. However, in most video
games, creators only rely on several specific models or allow
players to create complete avatars with only several optional
sub-models, e.g., nose, eyes, mouth, etc. Consequently, play-
ers’ avatars are highly similar.
Generative adversarial network (GAN) is a state-of-the-
art deep learning model in learning the distribution of train-
ing samples and generate data following the same distribu-
tion [271]. The core idea of GAN is the contest between a
generator network and a discriminator network. Specifically,
the generator network is used to output fake images with the
learnt data distribution, while the discriminator network inputs
the fake images and judge whether they are real. The generator
network will be trained until these fake images are not
recognised by the discriminator network. Then discriminator
network will be trained to improve its recognition accuracy.
During this procedure, these two networks learn from each
other. Finally, we got a well-performing generator network.
Several works [272]–[274] have applied GAN to automati-
cally generate 2D avatars in games. Some works [275]–[277]
further introduce real-time processing 3D mesh and textures to
generate 3D avatars. Chalas et al. develop an autonomous 3D
avatar generation application based on face scanning, instead
of 2D images [278]
Some video games allow players to leave behind their
models of themselves when players are not in the game.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 16
For example, Forza Motorsport develops Drivatars, which
learns players’ driving style with artificial intelligence [279].
When these players are not playing the game, other users can
have race with their avatars. Specifically, the system collects
players’ driving data, including road position, race line, speed,
brake, and accelerator. Drivatars learns from collected data and
creates virtual players with the same driving style. It is worth
noting that the virtual player is non-deterministic, which means
the racing results for a given virtual player may be not the
same in the same game. A similar framework is also realised
with neural network in [280].
Gesler et al. apply multiple machine learning algorithms in
the first person shooter (FPS) game to learn players’ shooting
style, including moving direction, leap moment, and acceler-
ator [281]. Through extensive experiments, they find neural
network outperforms other algorithms, including decision tree
and Naive Bayes.
For decision-making relevant games, reinforcement learning
usually outperforms other AI algorithms. Mendoncca et al.
apply reinforcement learning in fighting games [282]. They use
the same fighting data to train reinforcement learning model
and a neural network and find the reinforcement learning
model performs much better.
VIII. BL OC KC HA IN
It is expected to connect everything in the world in the
metaverse. Everything is digitised, including digital twins
for physical entities and systems, avatars for users, large-
scale, fine-grained map on various areas, etc. Consequently,
unfathomably vast amounts of data are generated. Uploading
such giant data to centralised cloud servers is impossible due
to the limited network resources [283]. Meanwhile, blockchain
techniques are developing rapidly. It is possible to apply
blockchains to the data storage system to guarantee the de-
centralisation and security in the metaverse [284], [285].
A. What is a blockchain fundamentally
Blockchain is a distributed database, in which data is stored
in blocks, instead of structured tables [286]. The architecture
of blockchain is shown in Figure 14. The generated data by
users are filled into a new block, which will be further linked
onto previous blocks. All blocks are chained in chronological
order. Users store blockchain data locally and synchronise
them with other blockchain data stored on peer devices with
a consensus model. Users are called nodes in the blockchain.
Each node maintains the complete record of the data stored
on the blockchain after it is chained. If there is an error on
one node, millions of other nodes could reference to correct
the error.
The characteristics of blockchain could be summarized as
follows:
•Decentralisation. The P2P network is formed by all
blockchain users. All users are equivalent in P2P network
and jointly maintain the data in blockchain without the
central manager. The data is stored and updated in a
decentralised manner.
Fig. 14. Illustration of blockchain.
•Anonymity. In blockchain, transactions are completed
without using the real identify ID of users. Blockchain
adopts address-based transactions based on cryptographic
algorithms, rather than personal identification. Every user
can use one or multiple anonymous addresses to access
the blockchain network. Therefore, users’ personal infor-
mation, e.g., assets and transaction information are well
protected.
•Immutability. Once the transaction is completed, it will
be packaged into blocks and synchronised to all nodes in
the network. The hash value of the current block is stored
in the next block, which means you need to regenerate
all following blocks and get recognised by all users if
you modify the the data in the block. The cost of data
tampering is significantly increased.
•Auditability. Each transaction on the blockchain is
recorded with a timestamp. Any blockchain user can trace
the transaction record through any node of the blockchain
network, and can verify the data through the hash value
of the block. A complete traceability information chian
could be formed with timestamp, which could be used
for auditing.
•Autonomy. Data can be shared in a more safe and
convenient manner on blockchian through consensus
algorithms and protocols without the participation of
authoritative third party.
According to the accessing rules, blockchain could be
classified into three categories: public blockchain, consortium
blockchain and private blockchain [287]. Public blockchain
also named permission-less blockchain, allows all anony-
mous users access the blockchain without authentication.
The data stored on the blockchain is available to all users.
All users could publish transactions on the blockchain any-
time, participate in consensus, and jointly maintain the
blockchain. The public blockchain is a kind of complete de-
centralised blockchain, owning to the open characteristic. Pub-
lic blockchain adopts proof-of-work as consensus mechanism.
Users participating in recording will be rewarded with tokens,
which further encourages users to contribute to consensus and

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 17
(a) The format of a transaction. (b) The format of a bitcoin.
Fig. 15. Illustration of the format of transaction and bitcoin.
guarantee the security of the blockchain.
Private blockchain is a kind of permissioned blockchain,
which is opposite to public blockchain, in terms of the access-
ing rule. Private blockchain is only used within enterprises.
Private blockchain is mainly used to manage database and
audit with in enterprises. Consensus mechanisms used in
private blockchain include Practical Byzantine Fault Tolerance
[288], Proof-of-Stake [289], Delegate Proof-of-Stake [290].
The performance of private blockchain is much better than
public blockchain, due to the limited nodes of blockchain in
closed environment.
Consortium blockchain, also called federated blochchain is
a tradeoff between publich blockchain and private blockchain,
which is managed by multiple organisations, instead of a single
organisation. In addition to aforementioned three consensus
algorithms adopted by private blockchian, Raft algorithm [291]
is also a common used consensus algorithm in consortium
blockchain. Consortium blockchain is often adopted in sce-
narios of transaction settlement among enterprises.
B. The ancestor: bitcoin and cryptocurrencies
Bitcoin is a cryptocurrency based on blockchain, which
is first proposed by Nakamato in 2008 [292]. Due to its
decentralisation and security, bitcoin has attracted more and
more attention from both academia and industry. The price of
bitcoin has increased by thousands times since it is released in
2009. There are also lots of cryptocurrencies proposed, e.g.,
Litecoin [293], Ripple [294], Ethereum [295], Monero [296],
etc. These cryptocurrencies are similar to Bitcoin. In Bitcoin
system, nodes connect with each other via P2P network.
Each node in P2P network keeps the complete transaction
records. In Bitcoin system, Proof-of-Work (PoW) is adopted
as consensus algorithm to select the user to record transactions
and further form extending chains. Encryption techniques are
used to guarantee the integrity, immutability, and verifiability
of the data.
A transaction consists of two parts: header and payload.
IX. CO MP UT ER VI SION
In this section, we examine the technical state of computer
vision in interactive systems and its potential for the metaverse.
Computer vision plays an important role in XR applications
and lays the foundation for achieving the metaverse. Most
XR systems capture visual information through an optical
see-through or video see-through display. This information
is processed, and results are delivered via a head-mounted
device or a smartphone, respectively. By leveraging such visual
information, computer vision plays a vital role in processing,
analysing, and understanding visuals as digital images or
videos to derive meaningful decisions and take actions. In
other words, computer vision allows XR devices to recognise
and understand visual information of users activities and
their physical surroundings, helping build more reliable and
accurate virtual and augmented environments.
Computer vision is extensively used in XR applications
to build a 3D reconstitution of the user’s environment and
locate the position and orientation of the user and device.
In Section IX-A, we review the recent research works on
3D scene localisation and mapping in indoor and outdoor
environments. Besides location and orientation, XR interactive
system also needs to track the body and pose of users.
We expect that in the metaverse, the human users will be
tracked with computer vision algorithms and represented as
avatars. With such intuition, in Section IX-B, we analyse the
technical status of human tracking and body pose estimation
in computer vision. Moreover, the metaverse will also require
to understand and perceive the user’s surrounding environment
based on scene understanding techniques. We discuss this topic
in Section IX-C. Finally, augmented and virtual worlds need
to tackle the problems related to object occlusion, motion blur,
noise, and the low-resolution of image/video inputs. Therefore,
image processing is an important domain in computer vision,
which aims to restore and enhance image/video quality for
achieving better metaverse. We will discuss the state-of-the-
art technologies in Section IX-D.
A. Visual Localisation and Mapping
In the metaverse, human users and their digital representa-
tives (i.e., avatars) will connect together and co-exist at the
intersection between the physical and digital worlds. Consid-
ering the concept of digital twins and its prominent feature
of interoperability, building such connections across physical
and digital environments requires a deep understanding of
human activities that may potentially drive the behaviours
of one’s avatar. In the physical world, we acquire spatial
information with our eyes and build a 3D reconstitution of
the world in our brain, where we know the exact location
of each object. Similarly, the metaverse needs to acquire
the 3D structure of an unknown environment and sense its
motion. To achieve this goal, simultaneous Localisation and
Mapping (SLAM) is a common computer vision technique
that estimates device motion and reconstructs an unknown
environment’s [297], [298]. A visual SLAM algorithm has to
solve several challenges simultaneously: (1) unknown space,
(2) free-moving or uncontrollable camera, (3) real-time, and
(4) robust feature tracking (drifting problem) [299]. Among
the diverse SLAM algorithms, the ORB-SLAM series, e.g.,
ORB-SLAM-v2 [300] have been shown to work well, e.g., in
the AR systems [299], [301].
23https://developer.apple.com/videos/play/wwdc2018/602

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 18
Fig. 16. Mapping before (a) and after (b) close-loop detection in ORB-
SLAM [300]. The loop trajectory is drawn in green, and the local feature
points for tracking are in red. (c) The visual SLAM demonstrate by ARCorev2
from Apple. The trajectory of loop detection is in yellow (Image source 23).
Visual SLAM algorithms often rely on three primary steps:
(1) feature extraction, (2) mapping the 2D frame to the 3D
point cloud, and (3) close loop detection.
The first step for many SLAM algorithms is to find fea-
ture points and generate descriptors [298]. Traditional feature
tracking methods, such as Scale-invariant feature transform
(SIFT) [302], detect and describe the local features in images;
however, they often too slow to run in real-time. Therefore,
most AR systems rely on computationally efficient feature
tracking methods, such as feature-based detection [303] to
match features in real-time without using GPU acceleration.
Although recently, convolutional neural networks (CNNs) have
been applied to visual SLAM and achieved promising perfor-
mance for autonomous driving with GPUs [304], it is still
challenging to apply to resource-constrained mobile systems.
With the tracked key points (features), the second step for
visual SLAM is how to map the 2D camera frames to get 3D
coordinates or landmarks, which is closely related to camera
pose estimation [305]. When the camera outputs a new frame,
the SLAM algorithm first estimates the key points. These
points are then mapped with the previous frame to estimate the
optical flow of the scene. Therefore, camera motion estimation
paves the way for finding the same key points in the new
frame. However, in some cases, the estimated camera pose
is not precise enough. Some SLAM algorithms, e.g., ORB-
SLAM [300], [306] also add additional data to refine the
camera pose by finding more key point correspondences. New
map points are generated via triangulation of the matching
key points from the connected frames. This process bundles
the 2D position of key points in the frames and the translation
and rotations between frames.
The last key step of SLAM aims to recover the camera
pose and obtain a geometrically consistent map, also called
close-loop detection [307]. As shown in Figure 16(c) for AR,
if a loop is detected, it indicates that the camera captures
previously observed views. Accordingly, the accumulated er-
rors in the camera motion can be estimated. In particular,
ORB-SLAM [300] checks whether the key points in a frame
are matched with the previously detected key points from
a different location. If the similarity exceeds a threshold, it
means the user has returned to a known place. Recently,
some SLAM algorithms also combined the camera with other
sensors, e.g., the IMU sensor, to improve the loop detection
precision [308], and some works, e.g., [309], have attempted to
fuse the semantic information to SLAM algorithms to ensure
the loop detection performance.
Although current state-of-the-art (SoTA) visual SLAM al-
gorithms already laid a solid foundation for spatial under-
standing, the metaverse needs to understand more complex
environments, especially the integration of virtual objects and
real environments. Hololens has already started getting deeper
in spatial understanding, and Apple has introduced ARKitv224
for 3D keypoint tracking, as shown in Figure 16(c). In the
metaverse, the perceived virtual universe is built in the shared
3D virtual space. Therefore, it is crucial yet challenging to
acquire the 3D structure of an unknown environment and sense
its motion. This could help to collect data for e.g., digital
twin construction, which can be connected with AI to achieve
auto conversion with the physical world. Moreover, in the
metaverse, it is important to ensure the accuracy of object
registration, and the interaction with the physical world. With
these harsh requirements, we expect the SLAM algorithms in
the metaverse to become more precise and computationally
effective to use.
B. Human Pose & Eye Tracking
In the metaverse, users are represented by avatars (see
Section XII). Therefore, we have to consider the control of
avatars in 3D virtual environments. Avatar control can be
achieved through human body and eye location and orientation
in the physical world. Human pose tracking refers to the com-
puter vision task of obtaining spatial information concerning
human bodies in an interactive environment [310]. In VR and
AR applications, the obtained visual information concerning
human pose can usually be represented as joint positions or
key points for each human body part. These key points reflect
the characteristics of human posture, which depict the body
parts, such as elbows, legs, shoulders, hands, feet, etc. [311],
[312]. In the metaverse, this type of body representation is
simple yet sufficient for perceiving the pose of a user’s body.
Tracking the position and orientation of the eye and gaze
direction can further enrich the user micro-interactions in the
metaverse. Eye-tracking enables gaze prediction, and intent
inference can enable intuitive and immersive user experiences,
which can be adaptive to the user requirement for real-time
interaction in XR environments [89], [313], [314]. In the
metaverse, it is imperative for eye tracking to operate reliably
under diverse users, locations, and visual conditions. Eye
tracking requires real-time operations within the power and
computational limitations imposed by the devices.
Achieving significant milestones of the above two tech-
niques relies on releasing several high-quality body and eye-
tracking datasets [315]–[318] combined with the recent ad-
vancement in deep learning. In the following subsections,
we review and analyse body pose and eye-tracking methods
developed for XR, and derive their potential benefits for the
metaverse.
1) Human Pose Tracking: When developing methods to
track human poses in the metaverse, we need to consider
several challenges. First, a pose tracking algorithm needs to
handle the self-occlusions of body parts. Second, the robust-
ness of tracking algorithms can impact the sense of presence,
24https://developer.apple.com/videos/play/wwdc2018/602

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 19
Fig. 17. Visual examples of pose and eye tracking. (a) body pose tracking
results from Openpose [323] and (b) eye tracking with no eye convergence
(left) and eye convergence (right) [340].
especially in multi-user scenarios. Finally, a pose tracking
algorithm needs to track the human body even in vastly
diverse illumination conditions, e.g., in the too bright or dark
scenes. Considering these challenges, most body pose tracking
methods combine the RGB sensor with infrared or depth
sensors [310], [319]–[321] to improve the detection accuracy.
Such sensor data are relatively robust to abrupt illumination
changes and convey depth information for the tracked pixel.
For XR applications, Microsoft Kinect25 and Open Natural
Interaction (OpenNI)26 are two popular frameworks for body
pose estimation.
In recent years, deep learning methods have been contin-
uously developed in the research community to extract 2D
human pose information from the RGB camera data [322]–
[324] or 3D human pose information from RGB-D sensor
data [325]–[327]. Among the SoTA methods for 2D pose
tracking, OpenPose [323] has been broadly used by researchers
to track users’ bodies in various virtual environments such
as VR [328], [329], AR [330]–[332], and metaverse [333].
For 3D pose tracking, FingerTrack [327] recently presented a
3D finger tracking and hand pose estimation method, which
displays high potential for XR applications and the metaverse.
Compared to single body pose tracking, multi-person track-
ing is more challenging. The tracking algorithm needs to
count the number of users and their positions and group them
by classes [334]. In the literature, many methods have been
proposed for VR [335], [336] and AR [337]–[339]. In the
metaverse, both single-person and multi-person body pose
tracking algorithms are needed in different circumstances. Re-
liable and efficient body pose tracking algorithms are needed
to ensure the close ties between the metaverse and the physical
world and people.
2) Eye Tracking: Eye-tracking is another challenging topic
in achieving the metaverse as the human avatars need to
‘see’ the immersive 3D environment. Eye tracking is based
on continuously measuring the distance between the pupil
centre and the refection of the cornea [341]. The angle of the
eyes converges at a certain point where the gaze intersects.
The region displayed within the angle of the eyes is called
‘vergence’ [342] – the distance changes with regard to the
angle of the eye. Intuitively, the computer vision algorithms
in eye-tracking should be able to measure the distance by
25https://developer.microsoft.com/en-us/windows/kinect/
26https://structure.io/openni
deducing from the angle of the eyes where the gaze is
fixed [340]. To measure the distance, one representative way
is to leverage infrared cameras, which can record and track
the eye movement information, as in the HMDs. In VR, the
HMD device is placed close to the eyes, making it easy to
display the vergence. However, the device cannot track the
distance owning to the 3D depth information. Therefore, depth
estimation for the virtual objects in the immersive environment
is one of the key problems.
Eye-tracking can bring lots of benefits for immersive en-
vironments in the metaverse. One of them is reducing the
computation cost in rendering the virtual environment. Eye
tracking makes it possible to only render the contents in the
view of users. As such, it can also facilitate the integration of
the virtual and real world. However, there are still challenges
in eye tracking. First of all, the lack of focus blur can lead to
an incorrect perception of the object size and distance in the
virtual environment [343]. Another challenge for eye tracking
is to ensure precise distance estimation with incomplete gaze
due to the occlusion [343]. Finally, eye tracking may lead to
motion sickness and eye fatigue [344]. In the metaverse, the
requirements for eye tracking can be much higher than tradi-
tional virtual environments. This opens up some new research
directions, such as understanding human behaviour accurately
and creating more realistic eye contact for the avatars, similar
to the physical eye contact, in the 3D immersive environment.
C. Holistic Scene Understanding
In the physical world, we understand the world by answer-
ing four fundamental questions: what is my role? What are the
contents around me? How far am I from the referred object?
What might the object be doing? In computer vision, holistic
scene understanding aims to answers these questions [345].
A person’s role is already clear in the metaverse as they are
projected through an avatar. However, the second question in
computer vision is formulated based on semantic segmentation
and object detection. Regarding the third question, we estimate
the distance to the reference objects based on our eyes in
the physical world. This way of scene perception in computer
vision is called stereo matching and depth estimation. The last
question requires us to interpret the physical world based on
our understanding. For instance, ‘a rabbit is eating a carrot’.
We need first to recognise the rabbit and the carrot and then
predict the action accordingly to interpret the scene. The
metaverse requires us to interact with other objects and users in
both the physical and virtual world. Therefore, holistic scene
understanding plays a pivotal role in ensuring the operation of
the metaverse.
1) Semantic Segmentation and Object Detection: Semantic
segmentation is a computer vision task to categorise an image
into different classes based on the per-pixel information [350],
[351], as shown in Figure 18(a). It is regarded as one of the
core techniques to understand the environment fully [352]. In
computer vision, a semantic segmentation algorithm should
efficiently and quickly segment each pixel based on the class
information. Recent deep learning-based approaches [350],
[351], [353] have shown a significant performance enhance-

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 20
Fig. 18. Visual examples for holistic scene understanding. (a) Semantic segmentation in AR environment [346]; (b) scale estimation in object detection (the
blue dots are generated by the detector) [347]; (c) Stereo depth estimation result (right) for VR [348]; (d) Deep learning-based hand action recognition based
on labels [349].
ment in urban driving datasets designed for autonomous driv-
ing. However, performing accurate semantic segmentation in
real-time remains challenging. For instance, AR applications
require semantic segmentation algorithms to run with a speed
of around 60 frames per second (fps) [354]. Therefore, seman-
tic segmentation is a crucial yet challenging task for achieving
the metaverse.
Object detection is another fundamental scene understand-
ing task aiming to localise the objects in an image or scene
and identify the class information for each object [355], as
shown in Figure 18(b). Object detection is widely used in XR
and is an indispensable task for achieving the metaverse. For
instance, in VR, face detection is a typical object detection
task, while text recognition is a common object detection
task in AR. In a more sophisticated application, AR ob-
ject recognition aims to attach a 3D model to the physical
world [347]. This requires the object detection algorithms to
precisely locate the position of objects and correctly recognise
the class. By placing a 3D virtual object and connecting it
with the physical object, users can manipulate and relocate
it. AR object detection can help build a richer and more
immersive 3D environment in the metaverse. In the following,
we analyse and discuss the SoTA semantic segmentation and
object detection algorithms for achieving the metaverse.
The early attempts of semantic segmentation mostly unitise
the feature tracking algorithms, e.g., SIFT [309] that aim
to segment the pixels based on the classification of the
handcrafted features, such as the support vector machine
(SVM) [356]. These algorithms have been applied to VR [357]
and AR [358]. However, these conventional methods suf-
fer from limited segmentation performance. Recent research
works have explored the potential of CNNs for semantic
segmentation. These methods have been successfully applied
to AR [346], [352], [354], [359]. Some works have shown the
capability of semantic segmentation for tackling the occlusion
problems in MR [360], [361]. However, as image segmentation
deals with each pixel, it leads to considerable computation and
memory load.
To tackle this problem, recent endeavours focus on real-time
semantic segmentation. Theses methods explore the image
crop/resizing [362] or efficient network design [363], [364]
or transfer learning [365], [366]. Through these techniques,
some research works managed to achieve real-time semantic
segmentation in MR [367]–[369].
In the metaverse, we need more robust and real-time se-
mantic segmentation methods to understand the pixel-wise
information in a 3D immersive world. More adaptive semantic
segmentation methods are needed because due to the diversity
and complexity of virtual and real objects, contents, and
human avatars. In particular, in the interlaced metaverse world,
the semantic segmentation algorithms also need to distinguish
the pixels of the virtual objects from the real ones. The class
information can be more complex in this condition, and the
semantic segmentation models may need to tackle unseen
classes.
Object detection in the metaverse can be classified into
two categories: detection of specific instances (e.g., face,
marker, text) and detection of generic categories (e.g., cars,
humans). Text detection methods have been broadly studied
in XR, [370], [371]. These methods have already matured
and can be directly applied to achieving the metaverse. Face
detection has also been studied extensively in recent years, and
the methods have shown to be robust in various recognition
scenarios in XR applications, e.g., [372]–[376].
In the metaverse, users are represented as avatars, and
multiple avatars can interact with each other. The face de-
tection algorithms need to detect the real faces (from the
physical world) and the synthetic faces (from the virtual
world). Moreover, the occlusion problems, sudden face pose
changes, and illumination variations in the metaverse can
make it more challenging to detect faces in the metaverse.
Another problem for face detection is the privacy problem.
Several research works have studied this problem in AR
application [377]–[379]. In the metaverse, many users can
stay in the 3D immersive environment; hence, privacy in
face detection can be more stringent. Future research should
consider the robustness of face detection, and better rules or
criteria need to be studied for face detection in the metaverse.
The detection of the generic categories has been studied
massively in recent years by the research community. Much
effort using deep learning has been focused on the detection of
multiple classes. The two-stage detector, FasterRCNN [380],
was one of the SoTA methods in the early development
stage using deep learning. Later on, the Yolo series and
SSD detectors [381]–[383] have shown wonderful detection
performance on various scenes with multiple classes. These
detectors have been successfully applied to AR [347], [384]–
[386].
From the above review, we can see that the SoTA object
detection methods have already been shown to work well for

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 21
XR. However, there are still some challenges for achieving
the metaverse. The first challenge is the smaller or tiny object
detection. This is an inevitable problem in the 3D immersive
environment as many contents co-exist in the shared space.
With variations of Field of View (FoV) of the camera, some
contents and objects will become smaller, making it hard
to detect. Therefore, the object detector in the metaverse
should be reinforced to detect these objects regardless of
the capture hardware. The second one is the data and class
distribution issues. In general, it is easy to collect large-scale
datasets with more than 100 classes; however, it is not easy
to collect datasets with a diverse scene and class distribution
in the metaverse. The last one is the computation burden for
object detection in the metaverse. The 3D immersive world
in the metaverse comprises many contents and needs to be
shared even in remote places. With the increment of class,
the computation burden is increased accordingly. To this end,
more efficient and lightweight object detection methods are
expected in the research community.
2) Stereo Depth Estimation: Depth estimation using stereo
matching is a critical task in achieving the metaverse. The
estimated distance directly determines the position of contents
in the immersive environment. The common way to estimate
depth is using a stereo camera [387], as shown in Figure 18(c).
In VR, stereo depth estimation is conducted in the virtual
space. Therefore, depth estimation estimates the absolute
distance between a virtual object to the virtual camera (first-
person view) or the referred object (third-person view). The
traditional methods first extract feature points and then us them
to compute the cost volumes, which is used to estimate the
disparity [388]. In recent years, extensive research has been
focused on exploring the potential of deep learning to estimate
depth in VR, e.g., [389], [390].
In XR, one of the critical issues is to ensure that depth
estimation is done based on both virtual and real objects. In
this way, the XR users can place the virtual objects in the
correct positions. Early methods in the literature for depth esti-
mation in AR/MR rely on the absolute egocentric depth [179],
indicating how far it is from a virtual object to the viewer. The
key techniques include “blind walking” [391], imagined blind
walking [392], and triangulation by walking [393]. Recently,
deep learning-based methods have been applied to XR [394]–
[396], showing much precise depth estimation performance.
Stereo cameras have been applied to some HMDs, e.g., the
Oculus Rift, [397]. Infrared camera sensor are also embedded
in some devices, such as HoloLens, enabling easier depth
information collection.
In the metaverse, depth estimation is a key task in ensuring
the precise positioning of objects and contents. In particular,
all users own their respective avatars, and both the digital and
real contents are connected. Therefore, depth estimation in
such a computer-generated universe is relatively challenging.
Moreover, the avatars representing human users in the physical
world are expected to experience heterogeneous activities in
real-time in the virtual world, thus requiring more sophisti-
cated sensors and algorithms to estimate depth information.
3) Action Recognition: In the metaverse, a human avatar
needs to recognise the action of the other avatars and contents.
In computer vision, understanding a person’s action is called
action recognition, which involves localising and predicting
human behaviours [400], as illustrated in Figure 18(d). In
XR, HMDs such as Hololens, usually needs to observe and
recognise the user’s actions and generate action-specific feed-
back in the 3D immersive environment. For instance, it is
often necessary to capture and analyse the user’s motion with
a camera for interaction purposes. With the advent of the
Microsoft Kinect, there have been many endeavours to capture
human body information and understand the action [321],
[401]. The captured body information is used to recognise the
view-invariant action [402], [403]. For instance, one aspect of
action recognition is finger action recognition [404].
Recently, deep learning has been applied to action recog-
nition in AR based on pure RGB image data [349], [405]
or multi-modal data via sensor fusion [406]. It has also
shown potential for emotion recognition in VR [407]. When
we dive deeper into the technical details of the success of
action recognition in XR, we find that it is important to
generate context-wise feedback based on the local and global
information of the captured pose information.
In the metaverse, action recognition can be very meaningful.
A human avatar needs to recognise the action of other avatars
or objects so that the avatar can take the correct action
accordingly in the 3D virtual spaces. Moreover, human avatars
need to emotionally and psychologically understand others and
the 3D virtual world in the physical world. More adaptive and
robust action recognition algorithms need to be explored. The
most challenging step of action recognition in the metaverse is
recognising the virtual contents across different virtual worlds.
Users may create and distribute virtual content from a virtual
world to the other. The problem of catastrophic forgetting for
AI models on multi-modal data for activity recognition should
also be tackled [408].
D. Image Restoration and Enhancement
The metaverse is connected seamlessly with the physical
environments in real-time. In such a condition, an avatar needs
to work with a physical person; therefore, it is important to
display the 3D virtual world with less noise, blur, and high-
resolution (HR) in the metaverse. In adverse visual conditions,
such as haze, low or high luminosity, or even rainy weather
conditions, the interactive systems in the metaverse still needs
to show the virtual universe.
In computer vision, these problems are studied under two
aspects: image restoration and image enhancement [409]–
[412]. Image restoration aims to reconstruct a clean image
from the degraded one (e.g., noisy, blur image). In contrast,
image enhancement focuses on improving image quality. In
the metaverse, image restoration and enhancement are much
in need. For instance, the captured body information and the
generated avatars may suffer from blur and noise when the user
moves quickly. The system thus needs to denoise and deblur
the users’ input signals and output clean visual information.
Moreover, when the users are far from the camera, the gen-
erated avatar may be in a low-resolution (LR). It is necessary
to enhance the spatial resolution and display the avatar in the
3D virtual environment with HR.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 22
Fig. 19. Visual examples for image restoration and enhancement. (a) Motion blur image and (b) no motion blurred image [398]; (c) Super-resolution image
with the comparison of HR and SR image patches [399].
1) Image Restoration: Image restoration has been shown
to be effective for VR display. For instance, [413] focuses on
colour VR based on image similarity restoration. In [398],
[414], [415], optimisation-based methods are proposed to
recover the textural details and remove the artefacts of the
virtual images in VR, as shown in Figure 19(b). These
techniques can be employed as Diminished Reality (DR) [416],
which allows human users to view the blurred scenes of the
metaverse with ‘screened contents’. Moreover, [417] examines
how image dehazing can be used to restore clean underwater
images, which can be used for marker-based tracking in AR.
Another issue is blur, which leads to registration failure in
XR. The image quality difference between the real blurred
images and the virtual contents could be apparent in the see-
through device, e.g., Microsoft Hololens. Considering this
problem, [418], [419] proposes first to blur the real images
captured by the camera and then render the virtual objects
with blur effects.
Image restoration has been broadly applied in VR and AR.
In the metaverse, colour correction, texture restoration, and
blur estimation also play important roles in ensuring a realistic
3D environment and correct interaction among human avatars.
However, it is worth exploring more adaptive yet effective
restoration methods to deal with the gap between real and
virtual contents and the correlation with the avatars in the
metaverse. In particular, the physical world, the users, and the
virtual entities are connected more closely in the metaverse
than those of AR/VR. Therefore, image restoration should be
subtly merged with the interaction system in the metaverse to
ensure effectiveness and efficiency.
2) Image Enhancement: Image enhancement, especially
image super-resolution, has been extensively studied for XR
displays. Image resolution has a considerable impact on user’s
view quality, which is related to the motion sickness caused
by HMDs. Therefore, extensive research has been focused on
optics SR e.g., [420], [421] and image SR [399], [422], [423]
for the display in VR/AR. An example of image SR for 360
images for VR is shown in Figure 19(c). Recently, [422]–
[425] applied deep learning and have achieved promising
performance on VR displays. These methods overcome the
resolution limitations that cause visible pixel artefacts in the
display.
In the metaverse, super-resolution display affects the per-
ception of the 3D virtual world. In particular, to enable a fully
immersive environment, it is important to consider the dis-
play’s image quality, for the sake of realism [91]. This requires
image super-resolution not only in optical imaging but also in
the image formation process. Therefore, future research could
consider the display resolution for the metaverse. Recently,
some image super-resolution methods, e.g., [426] have been
directly applied to HR display, and we believe these techniques
could help facilitate the technological development of the
optical and display in the metaverse. Moreover, the super-
resolution techniques in the metaverse can also be unitised
to facilitate the visual localisation and mapping, body and
pose tracking, and scene understanding tasks. Therefore, future
research could jointly learn the image restoration/enhancement
methods and the end-tasks to achieve the metaverse.
X. EDG E AN D CLO UD
With continuous, omnipresent, and universal interfaces to
information in the physical and virtual world [428], the
metaverse encompasses the reality-virtuality continuum and
allows user’s seamless experience in between. To date, the
most attractive and widely adopted metaverse interfaces are
mobile and wearable devices, such as AR glasses, headsets,
and smartphones, because they allow convenient user mobility.
However, the intensive computation required by the metaverse
is usually too heavy for mobile devices. Thus offloading is
necessary to guarantee the timely processing and user experi-
ence. The traditional cloud offloading faces several challenges:
user experienced latency, real-time user interaction, network
congestion, and user privacy. In this section, we review the
rising edge computing solution and its potential to tackle these
challenges.
A. User Experienced Latency
In the metaverse, it is essential to guarantee an immersive
feeling for the user to provide the same level of experience
as reality. One of the most critical factors that impact the
immersive feeling is the latency, e.g., motion to photon (MTP)
latency27. Researchers have found that MTP latency needs to
be below the human perceptible limit to allow users to interact
with holographic augmentations seamlessly and directly [429].
For instance, in the registration process of AR, large latency
often results in virtual objects lagging behind the intended
position [430], which may cause sickness and dizziness. As
27MTP latency is the amount of time between the user’s action and its
corresponding effect to be reflected on the display screen.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 23
Fig. 20. AR/VR network latency from the edge to the cloud [427].
such, reducing latency is critical for the metaverse, especially
in scenarios where real-time data processing is demanded,
e.g., real-time AR interaction with the physical world such as
AR surgeries [431]–[433], or real-time user interactions in the
metaverse such as multiplayer interactive exhibit in VR [434]
or multiple players’ battling in Fortnite.
As mentioned earlier, the metaverse often requires too inten-
sive computation for mobile devices and thus further increases
the latency. To compensate for the limited capacity of graphics
and chipsets in the mobile interfaces (AR glasses and VR
headsets etc.), offloading is often used to relieve the computa-
tion and memory burden at the cost of additional networking
latency [435]. Therefore a balanced tradeoff is crucial to make
the offloading process transparent to the user experience in
the virtual worlds. But it is not easy. For example, rendering
a locally navigable viewport larger than the headset’s field
of view is necessary to balance out the networking latency
during offloading [436]. However, there is a tension between
the required viewport size and the networking latency: longer
latency requires a larger viewport and streaming more content,
resulting in even longer latency [437]. Therefore, a solution
with physical deployment improvement may be more realistic
than pure resource orchestration.
Due to the variable and unpredictable high latency [438]–
[441], cloud offloading cannot always reach the optimal bal-
ance and causes long-tail latency performance, which impacts
user experience [442]. Recent cloud reachability measure-
ments have found that the current cloud distribution is able
to deliver network latency of less than 100 ms. However, only
a small minority (24 out of 184) of countries reliably meet the
MTP threshold [443] via wired networks and only China (out
of 184) meets the MTP threshold via wireless networks [444].
Thus a complementary solution is demanded to guarantee a
seamless and immersive user experience in the metaverse.
Edge computing, which computes, stores, and transmits
the data physically closer to end-users and their devices,
can reduce the user-experienced latency compared with cloud
offloading [445], [446]. As early as 2009, Satyanarayanan et
al. [439] recognized that deploying powerful cloud-like infras-
tructure just one wireless hop away from mobile devices, i.e.,
so-called cloudlet, could change the game, which is proved by
many later works. For instance, Chen et al. [447] evaluated
the latency performance of edge computing via empirical
studies on a suite of applications. They showed LTE cloudlets
could provide significant benefits (60% less latency) over the
default of cloud offloading. Similarly, Ha et al. [448] also
found that edge computing can reduce the service latency
by at least 80 ms on average compared to the cloud via
measurements. Figure 20 depicts a general end-to-end latency
comparison when moving from the edge to the cloud for an
easier understanding.
Utilising the latency advantage of edge computing, re-
searchers have proposed some solutions to improve the per-
formance of metaverse applications. For instance, EdgeXAR,
Jaguar, and EAVVE target mobile AR services. EdgeXAR
offers a mobile AR framework taking the benefits of edge
offloading to provide lightweight tracking with 6 Degree of
Freedom and hides the offloading latency from the user’s
perception [450]. Jaguar pushes the limit of mobile AR’s end-
to-end latency by leveraging hardware acceleration on edge
cloud equipped with GPUs [451]. EAVVE proposes a novel
cooperative AR vehicular perception system facilitated by
edge servers to reduce the overall offloading latency and makes
up the insufficient in-vehicle computational power [440],
[452]. Similar approaches have also been proposed for VR
services. Lin et al. [453] transformed the problem of energy-
aware VR experience to a Markov decision process and re-
alised immersive wireless VR experience using pervasive edge
computing. Gupta et al. [454] integrated scalable 360-degree
content, expected VR user viewport modelling, mmWave com-
munication, and edge computing to realise an 8K 360-degree
video mobile VR arcade streaming system with low interactive
latency. Elbamby et al. [455] proposed a novel proactive edge
computing and mmWave communication system to improve
the performance of an interactive VR network game arcade
which requires dynamic and real-time rendering of HD video
frames. As the resolution increases, edge computing will play
a more critical role to reduce the latency of 16K, 24K, or even

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 24
Fig. 21. An example MEC solution for AR applications [449].
higher resolution of the metaverse streaming.
B. Multi-access edge computing
The superior performance on reducing latency in virtual
worlds has made edge computing an essential pillar in the
metaverse’s creation in the eyes of many industry insiders.
For example, Apple uses Mac with an attached VR headset
to support 360-degree VR rendering [456]. Facebook Oculus
Quest 2 can provide VR experiences on its own without a
connected PC thanks to its powerful Qualcomm Snapdragon
XR2 chipset [457]. However, its capacity is still limited
compared with a powerful PC, and thus the standalone VR
experience comes at the cost of lower framerates and hence
less detailed VR scenes. By offloading to an edge server
(e.g., PC), users can enjoy a more interactive and immersive
experience at higher framerates without sacrificing detail. The
Oculus Air Link [458] announced by Facebook in April 2021
allows Quest 2 to offload to the edge at up to 1200 Mbps over
the home Wi-Fi network, enabling a lag-free VR experience
with better mobility. These products, however, are constrained
to indoor environments with limited user mobility.
To allow users to experience truly and fully omnipresent
metaverse, seamless outdoor mobility experience supported
by cellular networks is critical. Currently, last mile access
is still the latency bottleneck in LTE networks [459]. With
the development of 5G (promising down to 1 ms last mile
latency) and future 6G, Multi-access edge computing (MEC)
is expected to boost metaverse user experience by providing
standard and universal edge offloading services one-hop away
from the cellular-connected user devices, e.g., AR glasses.
MEC, proposed by the European Telecommunications Stan-
dards Institute (ETSI), is a telecommunication-vendor centric
edge cloud model wherein the deployment, operation, and
maintenance of edge servers is handled by an ISP operat-
ing in the area and commonly co-located with or one hop
away from the base stations [460]. Not only can it reduce
the round-trip-time (RTT) of packet delivery [461], but also
it opens a door for near real-time orchestration for multi-
user interactions [462], [463]. MEC is crucial for outdoor
metaverse services to comprehend the detailed local context
and orchestrate intimate collaborations among nearby users or
devices. For instance, 5G MEC servers can manage nearby
users’ AR content with only one-hop packet transmission and
enable real-time user interaction for social AR applications
such as ’Pok´
emon GO’ [464]. An example MEC solution
proposed by ETSI [449] is depicted in Figure 21.
Employing MEC to improve metaverse experience has ac-
quired academic attention. Dai et al. [465] designed a view
synthesis-based 360-degree VR caching system over MEC-
Cache servers in Cloud Radio Access Network (C-RAN) to
improve the QoE of wireless VR applications. Gu et al. [466]
and Liu et al. [467] both utilised the sub-6 GHz links and
mmWave links in conjunction with MEC resources to tackle
the limited resources on VR HMDs and the transmission rate
bottleneck for normal VR and panoramic VR video (PVRV)
delivery, respectively.
In reality, metaverse companies have also started to employ
MEC to improve user experience. For instance, DoubleMe, a
leading volumetric capture company, announced a proof of
concept project, Holoverse, in partnership with Telef´
onica,
Deutsche Telekom, TIM, and MobiledgeX, to test the optimal
5G Telco Edge Cloud network infrastructure for the seam-
less deployment of various services using the metaverse in
August 2021 [468]. The famous Niantic, the company which
has developed ‘Ingress’, ‘Pok´
emon GO’ and ‘Harry Potter:
Wizards Unite’, envisions building a “Planet-Scale AR”. It has
allied with worldwide telecommunications carriers, including
Deutsche Telekom, EE, Globe Telecom, Orange, SK Telecom,
SoftBank Corp., TELUS, Verizon, and Telstra, to boost their
AR service performance utilising MEC [469]. With the ad-
vancing 5G and 6G technologies, the last mile latency will
get further reduced. Hence MEC is promising to improve its
benefit on the universal metaverse experience.
C. Privacy at the edge
The metaverse is transforming how we socialise, learn, shop,
play, travel, etc. Besides the exciting changes it’s bringing,
we should be prepared for how it might go wrong. And
because the metaverse will collect more than ever user data,
the consequence if things go south will also be worse than ever.
One of the major concerns is the privacy risk [470], [471]. For
instance, the tech giants, namely Amazon, Apple, Google (Al-
phabet), Facebook, and Microsoft, have advocated password-
less authentication [472], [473] for a long time, which verifies
identity with a fingerprint, face recognition, or a PIN. The
metaverse is likely to continue this fashion, probably with
even more biometrics such as audio and iris recognition [474],
[475]. Before, if a user lost the password, the worst case is

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 25
the user lost some data and made a new one to guarantee
other data’s safety. However, since biometrics are permanently
associated with a user, once they are compromised (stolen by
an imposter), they would be forever compromised and cannot
be revoked, and the user would be in real trouble [476], [477].
Currently, the cloud collects and mines the data of end-users
and at the service provider side and thus has a grave risk of
serious privacy leakage [478]–[480]. In contrast, edge comput-
ing would be a better solution for both security and privacy
by allowing data processing and storage at the edge [481].
Edge service can also remove the highly private data from
the application during the authorization process to protect
user privacy. For instance, federated learning, a distributed
learning methodology gaining wide attention, trains and keeps
user data at local devices and updates the global model via
aggregating local models [482]. It can run on the edge servers
owned by the end users and conduct large-scale data mining
over distributed clients without demanding user private data
uploaded other than local gradients updates. This solution
(train at the edge and aggregate at the cloud) can boost the
security and privacy of the metaverse. For example, the eye-
tracking or motion tracking data collected by the wearables
of millions of users can be trained in local edge servers
(ideally owned by the users) and aggregated via a federated
learning parameter server. Hence, users can enjoy services
such as visual content recommendations in the metaverse
without leaking their privacy.
Due to the distinct distribution and heterogeneity charac-
teristics, edge computing involves multiple trust domains that
demand mutual authentication for all functional entities [483].
Therefore, edge computing requires innovative data security
and privacy-preserving mechanisms to guarantee its benefit.
Please refer to Section XVIII for more details.
D. Versus Cloud
As stated above, the edge wins in several aspects: lower
latency thanks to its proximity to the end-users, faster local or-
chestration for nearby users’ interactions, privacy-preservation
via local data processing. However, when it comes to long-
term, large-scale metaverse data storage and economic op-
erations, the cloud is still leading the contest by far. The
primary reason is that the thousands of servers in the cloud
datacenter can store much more data with better reliability
than the edge. This is critical for the metaverse due to its
unimaginably massive amount of data. As reasoned by High
Fidelity [484], the metaverse will be 1,000 times the size of
earth 20 years from now, assuming each PC on the planet only
needs to store and serve and simulate a much smaller area than
a typical video game. For this reason, robust cloud service is
essential for maintaining a shared space for thousands or even
millions of concurrent users in such a big metaverse.
Besides, as the Internet bandwidth and user-device capacity
increase, the metaverse will continue expansion and thus
demand expanding computation and storage capacity. It is
much easier and more economical to install additional servers
at the centralised cloud warehouses than the distributed and
space-limited edge sites. Therefore, the cloud will still play
a vital role in the metaverse era. On the other hand, edge
computing can be a complementary solution to enhance real-
time data processing and local user interaction while the cloud
maintains the big picture.
To optimise the interaction between the cloud and the edge,
an efficient orchestrator is a necessity to meet diversified and
stringent requirements for different processes in the meta-
verse [485]–[487]. For example, the cloud runs extensive data
management for latency-tolerant operations while the edge
takes care of real-time data processing and exchange among
nearby metaverse users. The orchestrator in this context can
help schedule the workload assignment and necessary data
flows between the cloud and the edge for better-integrated
service to guarantee user’s seamless experience. For example,
edge services process real-time student discussions in a virtual
classroom at a virtual campus held by the cloud. Or, like
mentioned in Section X-C, the edge stores private data such as
eye-tracking traces, which can leak user’s interests to various
types of visual content, while the cloud stores the public visual
content.
Several related works have been proposed lately to explore
the potential of edge cloud collaborations for the metaverse.
Suryavansh et al. [488] compared hybrid edge and cloud with
baselines such as only edge and only cloud. They analyzed
the impact of variation of WAN bandwidth, cost of the cloud,
edge heterogeneity, and found that the hybrid edge cloud
model performs the best in realistic setups. On the other hand,
Younis et al. and Zhang et al. proposed solutions for AR
and VR, respectively. More specifically, Younis et al. [489]
proposed a hybrid edge cloud framework, MEC-AR, for
MAR with a similar design to Figure 21. In MEC-AR, MEC
processes incoming edge service requests and manages the
AR application objects. At the same time, the cloud provides
an extensive database for data storage that cannot be cached
in MEC due to memory limits. Zhang et al. [490] focused on
the three main requirements of VR-MMOGs, namely stringent
latency, high bandwidth, and supporting a large number of
simultaneous players. They correspondingly proposed a hybrid
gaming architecture that places local view change updates
and frame rendering on the edge and global game state
updates on the cloud. As such, the system cleverly distributes
the workload while guaranteeing immediate responses, high
bandwidth, and user scalability.
In summary, edge computing is a promising solution to
complement current cloud solutions in the metaverse. It can 1)
reduce user experienced latency for metaverse task offloading,
2) provide real-time local multi-user interaction with better
mobility support, and 3) improve privacy and security for
the metaverse users. Indeed, the distribution and heterogeneity
characteristics of edge computing also bring additional chal-
lenges to fully reach its potential. We briefly outline several
challenges in Section XVIII.
XI. NE TWORK
By design, a metaverse will rely on pervasive network
access, whether to execute computation-heavy tasks remotely,
access large databases, communicate between automated sys-
tems, or offer shared experiences between users. To address

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 26
the diverse needs of such applications, the metaverse will rely
heavily on future mobile networking technologies, such as 5G
and beyond.
A. High Throughput and Low-latency
Continuing on the already established trends of real-time
multimedia applications, the metaverse will require massive
amounts of bandwidth to transmit very high resolution con-
tent in real-time. Many interactive applications consider the
motion-to-photon latency, that is the delay between an action
by the user and its impact on-screen [491], as one of the
primary drivers of user experience.
The throughput needs of future multimedia applications
are increasing exponentially. The increased capabilities of 5G
(up to 10Gb/s [492]) have opened the door to a multitude
of applications relying on the real-time transmission of large
amounts of data (AR/VR, cloud gaming, connected vehicles).
By interconnecting such a wide range of technologies, the
metaverse’s bandwidth requirements will be massive, with
high-resolution video flows accounting for the largest part of
the traffic, followed by large amounts of data and metadata
generated by pervasive sensor deployments [493]. In a shared
medium such as mobile networks, the metaverse will not
only require a significant share of the available bandwidth,
but also likely compete with other applications. As such, we
expect the metaverse’s requirements to exceed 5G’s available
bandwidth [435]. Latency requirements highly depend on the
application. In the case of highly interactive applications such
as online and cloud gaming, 130 ms is usually considered as
the higher threshold [494], while some studies exhibit drops
in user performance for latencies as low as 23 ms [495]. Head-
mounted displays such as see-through AR or VR, as well
as haptic feedback devices exhibit motion-to-photon latency
requirements down to the millisecond to preserve the user’s
immersion [496], [497].
Many factors contribute to the motion-to-photon latency,
among which the hardware sensor capture time (e.g., frame
capture time, touchscreen presses [498]), and the computation
time. For applications requiring latency in the order of the
millisecond, the OS context switching frequency (often set
between 100Hz and 1500Hz [499]), and memory allocation
and copy times between different components (e.g. copy
between CPU and GPU memory spaces) also significantly
affect the overall motion-to-photon latency [500]. In such con-
strained pipeline, network operations introduce further latency.
Although 5G promised significant latency improvements, re-
cent measurement studies show that the radio access network
(RAN) itself displays very similar latency to 4G, while most
of the improvements come from the communication between
the gNB and the operator core network [501]. However, it
is important to note that most 5G networks are implemented
in Non Standalone (NSA) mode, where only the RAN to the
gNB use 5G radio, while the operator core network remains
primarily 4G. Besides, despite standardising RAN latency to
4 ms for enhanced Mobile Broadband (eMBB) and 0.5 ms
for Ultra-Reliable Low-Latency Communication (uRRLC –
still not implemented) [502], the communication between the
Large Data
Downloads
Voice
Communication
Live Video
Streaming
Clound and Online
Gaming
General
IoT
Smart Homes
Connected Vehicles
Smart Cities AR/VR
Enhanced Mobile
Broadband
(eMBB)
Ultra Reliable Low
Latency Communications
(URLLC)
massive Machine Type
Communications
(mMTC)
Robots and
Drones
Fig. 22. Metaverse applications and 5G service classes.
gNB and the core network account for most of the round
trip latency (between 10 and 20 ms), with often little control
from the ISP [501]. As such, unless servers are directly
connected to the 5G gNB, the advantages of edge computing
over cloud computing may be significantly limited [503], espe-
cially in countries with widespread cloud deployments [504].
Another consideration for reduced latency could be for content
providers to control the entire end-to-end path [505], by
reaching inside the ISP using to network virtualization [506].
Such a vision requires commercial agreements between ISPs
and content providers that would be more far-reaching than
peering agreements between AS. One of the core condition
for the metaverse to succeed will be the complete coordination
of all actors (application developers, ISPs, content providers)
towards ensuring a stable, low-latency and high throughput
connection.
At the moment, 5G can therefore barely address the la-
tency requirements of modern multimedia applications, and
displays latency far too high for future applications such as
see-through AR or VR. The URLLC service class promises
low latency and high reliability, two often conflicting goals,
with a standardised 0.5 ms RAN latency. However, URLLC
is still currently lacking frameworks encompassing the entire
network architecture to provide latency guarantees from client
to server [507]. As such, no URLLC has so far been com-
mercially deployed. Besides, we expect uRRLC to prioritize
applications for which low-latency is a matter of safety, such
as healthcare, smart grids, or connected vehicles, over enter-
tainment applications such as public-access AR and VR. The
third service class provided by the 5G specification is massive
Machine Type Communication (mMTC). This class targets
specifically autonomous machine-to-machine communication
to address the growing number of devices connected to the
Internet [508]. Numerous applications of the metaverse will
require mMTC to handle communication between devices
outside of the users’ reach, including smart buildings and
smart cities, robots and drones, and connected vehicles. Future
mobile networks will face significant challenges to efficiently
share the spectrum between billions of autonomous devices
and human-type applications [509], [510]. We summarize the

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 27
application of these service classes in Figure 22 Network
slicing will also be a core enabler of the metaverse, by
providing throughput, jitter, and latency guarantees to all
applications within the metaverse [511]. However, similar to
URLLC, deploying network slicing in current networks will
most likely target mission-critical applications, where network
conditions can significantly affect the safety of the equipment
or the users [512], [513]. Besides, network slicing still needs to
address the issue of efficiently orchestrating network resources
to map the network slices with often conflicting requirements
to the finite physical resources [514]. Finally, another feature
of 5G that may significantly improve both throughput and
latency is the usage of new frequency bands. The Millimeter
wave band (24GHz-39GHz) allows for wide channels (up
to 800MHz) providing large throughput while minimizing
latency below 1 ms. mmWave frequencies suffer from low
range and obstacle penetration. As such, mmWave has been
primarily used through dense base station deployments in
crowded environments such as the PyeongChang olympics
in 2018 (Korea) or Narita airport (Japan) [515]. Such dense
deployments allowed to serve a significantly higher number
of users simultaneously, while preserving high throughput and
low latency at the RAN.
B. Human- and user-centric networking
The metaverse is a user-centric application by design. As
such, every component of the multiverse should place the
human user at its core. In terms of network design, such
consideration can take several forms, from placing the user
experience at the core of traffic management, to enabling user-
centric sensing and communication.
To address these issues, the network community has been
increasingly integrating metrics of user experience in network
performance measures, under the term Quality of Experience
(QoE). QoE aims to provide a measurable way to estimate
the user’s perception of an application or a service [516].
Most studies tend to use the term QoE as a synonym for
basic Quality of Service (QoS) measures that may affect the
user experience (e.g., latency, throughput). However, several
works attempt to formalise the QoE through various models
combining network- and application-level metrics. Although
these models represent a step in the right direction, they are
application-specific, and can be affected by a multitude of
factors, whether human, system, or context [517]. Measuring
QoE for a cloud gaming application run on a home video game
console such as Sony PS Now28 is significantly different from
a mobile XR application running on a see-through headset.
Besides, many studies focus on how to estimate the video
quality as close as possible to the user’s perception [518],
[519], and most do not consider other criteria such as usability
or the subjective user perception [520]. The metaverse will
need to integrate such metrics to handle user expectations and
proactively manage traffic to maximise the user experience.
Providing accurate QoE metrics to assess the user experi-
ence is critical for user-centric networked applications. The
next step is to integrate QoE in how the network handles
28https://www.playstation.com/en-us/ps-now/
traffic. QoE can be integrated at various levels on the network.
First, the client often carries significant capabilities in sensing
the users, their application usage, and the application’s context
of execution. Besides, many applications such as AR or live
video streaming may generate significant upload traffic. As
such, it makes sense to make the client responsible for man-
aging network traffic from an end-to-end perspective [521],
[522]. The server-side often carries more computing power,
and certain applications are download-heavy, such as 360
video or VR content streaming. In this case, the server may use
the QoE measurements communicated by the client to adapt
the network transmission accordingly. Such approach has
been used for adapting the quality of video streaming based
on users’ preferences [523], using client’s feedback [524].
Finally, it is possible to use QoE measures to handle traffic
management in core network, whether through queuing poli-
cies [525], [526], software defined network [527], or network
slicing [528]. To address the stringent requirements leading
to a satisfying user experiences, the metaverse will likely
require to skirt the traditional layered approach to networks.
The lower network layers may communicate information on
network available resources for the application layer to adapt
the amount of data to transmit, while measurement of QoE
at application-level may be considered by the lower layers to
adapt the content transmission [521].
Making networks more human-centric also means consider-
ing human activities that may affect nework communication.
Mobility and handover are one of the primary factor affecting
the core network parameters’ stability. Handover have always
been accompanied with a transient increase in latency [529].
Although many works attempt to minimise handover latency
in 5G [530], [531], such latency needs to be accounted for
when designing ultra-low-latency services in mobile scenarios.
The network conditions experienced by a mobile user are
also directly related to the heterogeneity of mobile operator
infrastructure deployment. A geographical measurement study
of 4G latency in Hong Kong and Helsinki over multiple
operators showed that mobile latency was significantly im-
pacted by both the ISP choice and the physical location of
the user [532]. Overall, user mobility significantly affects the
network parameters that drive the user experience, and should
be accounted for in the design of user-centric applications.
Another aspect of human-centric networking lies within the
rise of embodied sensors. In recent years, sensor networks
have evolved from fixed environment sensors to self-arranging
sensor networks [533]. Many of such sensors were designed
to remain at the same location for extended durations, or
in controlled mobility [534]. In parallel, embodied sensors
have long been thought to sense only the user. However, we
are now witnessing a rise in embodied sensors sensing the
entire environment of the user, raising the question of how
such sensors may communicate in the already-crowded com-
munication landscape. Detecting and aggregating redundant
information between independent sensors may be critical to
release important resources on the network [535].

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 28
Latency
Throughput
Jitter
Mobility QoE
Interaction UsageCongestion Control
User Network Application
Generated Data
Fig. 23. Network- and User-aware applications in the metaverse. A synergy
between the traditional network layers and the application-level measures of
user experience allow for maximising the user experience given the actual
network conditions.
C. Network-aware applications
In the previous section, we saw how the transmission
of content should be driven by QoE measurements at the
application layer. While this operation enables a high accuracy
in estimating user experience by combining network metrics
with application usage measures, the lower network layers
only have limited control on the content to be transmitted. In
many applications of the metaverse, it would make more sense
for the application layer to drive the amount of data to transmit,
as well as the priority of the content to the lower network
layers [435]. Network-aware applications were proposed in
the late 1990s to address such issues [536], [537]. Many
framework were proposed, for both fixed and mobile net-
works [538]. More recently, network-aware applications have
been proposed for resource provisioning [539], distributed
learning optimization [540], and content distribution [541],
[542].
With the rapid deployment of 5G, there is a renewed interest
in network-aware applications [543]. 5G enabled many user-
centric applications to be moved to the cloud, such as cloud
gaming, real-time video streaming, or cloud VR. These appli-
cations rely extensively on the real-time transmission of video
flows, which quality can be adapted to the network conditions.
The 5G specification includes network capability exposure,
where the gNB can communicate the RAN conditions to the
user equipment [502]. In edge computing scenarios where the
edge server is located right after the gNB, the user equipment
is thus made aware of the conditions of the entire end-to-end
path. When the server is located further down the network,
network capability exposure stills addresses one of the most
variable components of the end-to-end path, providing valu-
able informations to drive the transmission. Such information
from the physical and access layer can then be propagated
to the network layer, where path decisions may be taken
according to the various networks capabilities, the transport
layer to proactively address potential congestion [544], and
the application layer to reduce or increase the amount of data
to transmit and thus maximise the user experience [545].
Figure 23 summarises how a synergy between user-centric
and network-aware applications can be established to maxi-
mize the user experience. The application communicates QoE
and application usage metrics to the lower layers in order
to adapt the transmission and improve the user experience.
In parallel, the network layers communicate the network
conditions to the application, which in turns regulates the
amount of content to transmit on the network, for instance,
by reducing the resolution of a video stream.
XII. AVATAR
The term Avatar is originated from the Hindu concept that
describes the incarnation of a Hindu god, appearing as humans
or animals in the ordinary world29. Avatars appear in a broad
spectrum of digital worlds. First, it has been commonly used
as profile pictures in various chatrooms (e.g., ICQ), forums
(e.g., Delphi), blogs (e.g., Xanga), as well as social networks
(e.g., Facebook, Figure 24(a)). Moreover, game players, with
very primitive metaverse examples such as AberMUD and
Second Life, leverage the term avatar to represent themselves.
Recently, game players or participants in virtual social net-
works can modify and edit the appearance of their avatars,
with nearly unlimited options [546], for instance, Fortnite, as
shown in Figure 24(b). Also, VR games, such as VR Chat
(Figure 24(c)), allow users to scan their physical appearance,
and subsequently choose their virtual outfits, to mimic the
users’ real-life appearances. Figure 24(d) shows that online
meetings, featured with AR, enable users to convert their
faces into various cartoon styles. Research studies have also
attempted to leverage avatars as one’s close friends, coaches,
or an imaginary self to govern oneself and goal setting such
as learning and nutrition [547], [548].
Under the domain of computer science and technology,
avatars denote the digital representation of users in virtual
spaces, as above mentioned, and other physical embodied
agents, e.g., social robots, regardless of form sizes and
shapes [549]. This section focuses the discussion on the digital
representationsn. However, it is worthy of pinpointing that
the social robots could be a potential communication channel
between human users and virtual entities across the real world
and the metaverse, for instance, robots can become aware of
the user’s emotions and interact with the users appropriately in
a conversation [550], or robots can serve as service providers
as telework (telepresence workplace) in physical worlds [551].
The digital representation of a human user aims to serve as
a mirrored self to represent their behaviours and interaction
with other users in the metaverse. The design and appear-
ance of avatars could impact the user perceptions, such as
senses of realism [552] and presence [553], trust [554], body
ownership [555], and group satisfaction [556], during various
social activities inside the metaverse, which are subject to a
bundle of factors, such as the details of the avatar’s face [557]
and the related micro-expression [558], the completeness of
the avatar’s body [553], the avatar styles [559], representa-
tion [560], colour [561] and positions [562], fidelity [563],
the levels of detail in avatars’ gestures [564], shadow [565],
the design of avatar behaviours [554], synchronisation of
the avatar’s body movements [566], Walk-in-Place move-
ments [567], ability of recognising the users’ self motions
reflected on their avatars [568], cooperation and potential
29https://www.merriam-webster.com/dictionary/avatar

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 29
Fig. 24. Several real-life examples of avatars, as a ‘second-identity’ on a wide spectrum of virtual worlds: (a) Facebook Avatar – users can edit their own
avatars in social media; (b) Fortnite – a multiplayer game that allows game players to create and edit their own worlds; (c) VR Chat – a VR game, and; (d)
Memoji – virtual meetings with cartoonised faces during FaceTime on Apple iOS devices, regarded as an example of AR.
glitches among multiple avatars [569], and to name but a few.
As such, avatars has the key role of shaping how the virtual
social interaction performs in the multi-user scenarios inside
the metaverse [546]. However, the current computer vision
techniques are not ready to capture and reflect the users’
emotions, behaviours and their interaction in real-time, as
mentioned in Section IX, Therefore, additional input modality
can be integrated to improve the granularity of avatars. For
instance, the current body sensing technology is able to enrich
the details of the avatar and reflect the user’s reactions in
real-time. In [570], an avatar’s pupillary responses can reflect
its user’s heartbeat rate. In the virtual environments of VR
Chat, users in the wild significantly rely on body sensing
technology (i.e., sensors attached on their body) to express
their body movements and gestural communication, which
facilitate non-verbal user interaction (i.e., voice, gestures, gaze,
and facial expression) emulating the indispensable part of real-
life communication [571].
When avatars become more commonplace in vastly di-
versified virtual environments, the studies of avatars should
go beyond the sole design aspects as above. We briefly
discuss six under-explored issues related to the user interaction
through avatars with virtual environments – 1) in-the-wild user
behaviours, 2) the avatar and their contexts of virtual environ-
ments, 3) avatar-induced user behaviours, 4) user privacy, 5)
fairness, and 6) connections with physical worlds. First, as
discussed in prior sections, metaverse could become indepen-
dent virtual venues for social gatherings and other activities.
The user behaviours in the wild (i.e., outside laboratories),
on behalf of the users’ avatars, need further investigation,
and the recently emerging virtual worlds could serve as a
testing bed for further studies. For instance, it is interest-
ing to understand the user behaviours, in-group dynamics,
between-group competitions, inside the virtual environments
encouraging users to earn NFTs through various activities.
Second, We foresee that users with avatars will experience
various virtual environments, representing diversified contexts.
The appearance of avatars should fit into such contexts. For
instance, avatars should behave professionally to gain trust
from other stakeholders in virtual work environments [572].
Third, it is necessary to understand the changes and dynamics
of user behaviours induced by the avatars in virtual environ-
ments. A well-known example is the Proteus Effect [573]
that describes the user behaviours within virtual worlds are
influenced by the characteristics of our avatar. Similarly,
supported by the Self-perception theory, user’s behaviours in
virtual environments are subjects to avatar-induced behavioural
and attitudinal changes through a shift in self-perception [574].
Furthermore, when the granularity of the avatars can be truly
reflected by advancing technologies, avatar designers should
consider privacy-preserving mechanisms to protect the identity
of the users [575]. Next, the choices of avatars should represent
a variety of populations. The current models of avatars may
lead to biased choices of appearances [576], for instance, a tall
and white male [577]. Avatar designers should offer a wide
range of choices that enables the population to equally choose
and edit their appearance in virtual environments. Finally,
revealing metaverse avatars in real-world environments are
rarely explored. Revealing avatars in the real world is able to
enhance the presence (i.e., co-presence of virtual humans in
the real world [578]), especially when certain situations prefer
the physical presence of an avatar that represents a specific per-
son, e.g., lectures [579]. Interaction designers should explore
various ways of displaying the avatar on tangible devices (three
examples as illustrated in Figure 6) as well as social robots.
XIII. CO NT EN T CREATIO N
This section aims to describe the existing authoring sys-
tems that support content creation in XR, and then discuss
censorship in the metaverse and a potential picture of creator
culture.
A. Authoring and User Collaboration
In virtual environments, authoring tools enable users to
create new digital objects in intuitive and creative manners.
Figure 25 illustrates several examples of XR/AR/VR authoring
systems in the literature. In VR [17], [580]–[582], the immer-
sive environments provides virtual keyboards and controllers
that assist users in accomplishing complicated tasks, e.g.,
constructing Functional Reactive Programming (FRP) diagram
as shown in Figure 25(a). In addition, re-using existing patterns
can speed up the authoring process in virtual environments,
such as a presentation (Figure 25(b)). Also, users can leverage
smart wearables to create artistic objects, e.g., smart gloves in
Figure 25(c). Combined with the above tools, users can design
interactive AI characters and their narratives in virtual environ-
ments (Figure 25(d)). In AR or MR, users can draw sketches
and paste overlays on physical objects and persons in their
physical surroundings [584], [585], [587]–[589]. Augmenting

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 30
Fig. 25. Authoring systems with various virtual environments across extended reality (e) & (h), VR (a) – (d), and AR (f) – (g): (a) FlowMatic [580], (b)
VR nuggets with patterns [581], (c) HandPainter [17] for VR artistic painting, (d) Authoring Interactive VR narrative [582], (e) Corsican Twin [583] as an
example of digital twins, (f) PintAR [584] for low-fidelity AR sketching, (g) Body LayARs [585] creates AR emojis according to the detected faces, (h)
Creating medium-fidelity AR/VR experiences with 360 degree theatre [586].
the physical environments can be achieved by drawing a new
sketch in mid-air [584], [587], e.g., Figure 25(f), detecting
the contexts with pre-defined AR overlays ((Figure 25(g)),
recording the motions of real-world objects to simulate their
physical properties in AR [590], inserting physical objects
in AR (Figure 25(h)), or even using low-cost objects such
papers [591] and polymer clay [588].
Although the research community is increasingly interested
in XR/AR/VR authoring systems [592], such authoring tools
and platforms mainly assist users in creating and inserting
content without high technological barriers. Additionally, it
is important to note that AI can play the role of automatic
conversion of entities from the physical world to virtual
environments (Section VII). As such, UI/UX designers and
other non-coders feel more accessible to content creation in
virtual environments, on top of virtual world driven by the
AI-assisted conversion. Nevertheless, to build the metaverse
at scale, three major bottlenecks exist: 1) organising the new
contents in interactive and storytelling manners [593], 2)
allowing collaborative works among multiple avatars (i.e., hu-
man users) [594], and 3) user interaction supported by multiple
heterogeneous devices [595]. To the best of our knowledge,
only limited work attempts to resolve the aforementioned
bottleneck, and indicate the possibility of role-based collab-
orative content creation [18], [586], [596]. As depicted by
Speichers et al. [586], the peer users can act in different
roles and work collaboratively in virtual environments, such
as wizards,observers,facilitators,AR and VR users as content
creators, and so on. Similarly, Nebeling et al. consider three
key roles of directors,actors, and cinematographers to create
complex immersive scenes for storytelling scenarios in virtual
environments.
Although we cannot speculate all the application scenar-
ios of the authoring techniques and solutions, human users
can generate content in various ways, i.e., user-generated
content, in the metaverse. It is important to note that such
authoring systems and their digital creation are applicable
to two apparent use cases. First, remote collaboration on
physical tasks [597] and virtual tasks [598] enable users
to give enriched instructions to their peers, and accordingly
create content for task accomplishment remotely. Second,
the content creation can facilitate the video conference or
equivalent virtual venues for social gathering, which are
the fundamental function of the metaverse. Since 2020, the
unexpected disruption by the global pandemic has sped up
the digital transformation, and hence virtual environments
are regarded as an alternative for virtual travelling, social
gathering and professional conferencing [599], [600]. Online
Lectures and remote learning are some of the most remarkable
yet impactful examples, as schools and universities suspend
physical lessons globally. Students primarily rely on remote
learning and obtaining learning materials from proprietary
online platforms. Teachers choose video conferencing as the
key reaching point with their students under this unexpected

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 31
circumstance. However, such online conferences would require
augmentations to improve their effectiveness [601]. XRStudio
demonstrates the benefits from the additions of virtual over-
lays (AR/VR) in video conferencing between instructors and
students. Similarly, digital commerce relies heavily on online
influencers to stimulate sales volumes. Such online influencers
share user-generated content via live streaming, for instance,
tasting and commenting on foods online [602], gain attention
and interactions with viewers online. According to the above
works, we foresee that the future of XR authoring systems
can serve to augment the participants (e.g., speakers) during
their live streaming events. The enriched content, supported by
virtual overlays in XR, can facilitate such remote interaction.
The speakers can also invite collaborative content creations
with the viewers. The metaverse could serve as a medium to
knit the speakers (the primary actor of user-generated content)
and the viewers virtually onto a unified landscape.
B. Censorship
Censorship is a common way of suppressing ideas and in-
formation when certain stakeholders, regardless of individuals
or groups, as well as authorities may find such ideas and in-
formation are objectionable, dangerous, or detrimental [603]–
[605]. In the real world, censorship brings limited access to
specific websites, controlling the dissemination of information
electronically, restricting the information disclosed to the pub-
lic, facilitating religious beliefs and creeds, and reviewing the
contents to be released, so as to guarantee the user-generated
contents would not violate rules and norms in a particular
society, with the potential side effects of sacrificing freedom
of speech or certain digital freedom (e.g., discussions on
certain topics) [606]. Several censorship techniques (e.g., DNS
manipulation and HTTP(S)-layer interference) are employed
digitally [603]–[609]: 1) entire subnets are blocked by using
IP-filtering techniques; 2) certain sensitive domain is limited
to block the access of specific websites; 3) certain keywords
become the markers of targeting certain sensitive traffic, 4)
Specific contents and pages are specified as the sensitive or
restricted categories, perhaps with manual categorisations.
Other prior works of censorship in the Internet and social
networks have reflected the censorship employed in Iran [605],
Egypt, Sri Lanka, Norway [609], Pakistan [607], Syria [603]
and other countries in the Arab world [608]. The majority
of these existing works leverages the probing approaches –
the information being censored is identified by the events of
requests of generating new content and subsequently the actual
blocking of such requests. Although the probing approaches
allow us to become more aware of censorship in particular
regions, it poses two key limitations: 1) limited observation
size (i.e., limited scalability) and 2) difficult identification of
the contents being censored (i.e., primarily by inference or
deduction).
Once the metaverse becomes a popular place for content
creations, numerous user interaction traces and new content
will be created. For instance, Minecraft has been regarded as a
remarkable virtual world in which avatars have a high degree
of freedom to create new user-generated content. Minecraft
also supports highly diversified users who intend to meet
and disseminate information in such virtual worlds. In 2020,
Minecraft acted as a platform to hold the first library for
censored information, named The Uncensored Library30, with
the emphasis of ‘A safe haven for press freedom, but the
content you find in these virtual rooms is illegal’. Analogue
to the censorship employed on the Internet, we conjecture
that similar censorship approaches will be exerted in the
metaverse, especially when the virtual worlds in the metaverse
grow exponentially, for instance, blocking the access of certain
virtual objects and virtual environments in the metaverse. It is
projected that censorship may potentially hurt the interoper-
ability between virtual worlds, e.g., will the users’ logs and
their interaction traces be eradicated in one censored virtual
environment? As such, do we have any way of preserving the
ruined records? Alternatively, can we have any instruments
temporarily served as a haven for sensitive and restricted
information? Also, other new scenarios will appear in the
virtual 3D spaces. For example, censorship can be applied
to restrict certain avatar behaviours, e.g., removal of some
keywords in their avatars’ speeches, forbidding avatars’ body
gestures, and other non-verbal communication means [610].
Although we have no definitive answer to the actual imple-
mentation of the censorship in the metaverse and the effective
solutions to alleviate such impacts, we advocate a compre-
hensive set of metrics to reflect the degree of censorship in
multitudinous virtual worlds inside the metaverse, which could
serve as an important lens for the metaverse researchers to
understand the root cause(s) and its severity and popularity of
the metaverse censorship. The existing metrics for the Internet,
namely Censored Planet, perform a global-scale censorship
observatory that helps to bring transparency to censorship
practices, and supports the human rights of Internet users
through discovering key censorship events.
C. Creator Culture
The section on content creation ends with a conjecture of
creator culture, as we can only construct our argument with
the existing work related to creators and digital culture to
outline a user-centric culture on a massive scale inside the
metaverse. First, as every participant in the metaverse would
engage in creating virtual entities and co-contribute to the new
assets in the metaverse, we expect that the aforementioned
authoring systems should remove barriers for such co-creation
and co-contribution. In other words, the digital content cre-
ation will probably let all avatars collaboratively participate
in the processes, instead of a small number of professional
designers [611]. Investigating the design space of author-
ing journeys and incentive schemes designated for amateur
and novice creators to actively participate in the co-creation
process could facilitate the co-creation processes [612]. The
design space should further extend to the domain of human-
AI collaboration, in which human users and AI can co-create
instances in the metaverse [613]. Also, one obvious incentive
could be token-based rewards. For instance, in the virtual
environment Alien Worlds, coined as a token-based pioneer of
30https://www.uncensoredlibrary.com/en

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 32
the metaverse, allows players’ efforts, through accomplishing
missions with their peers, to be converted into NFTs and hence
tangible rewards in the real world.
It is projected that the number of digital contents in the
metaverse will proliferate, as we see the long-established
digital music and arts [614], [615]. For instance, Jiang et
al. [17] offer a virtual painting environment that encourages
users to create 3D paintings in VR. Although we can assume
that computer architectures and databases should own the
capacity to host such growing numbers of digital contents,
we cannot accurately predict the possible outcomes when the
accumulation of massive digital contents exceed the capacity
of the metaverse – the outdated contents will be phased out or
be preserved. This word capacity indicates the computational
capacity of the metaverse, and the iteration of the virtual space.
An analogy is that real-world environments cannot afford
an unlimited number of new creations due to resource and
space constraints. For example, an old street painting will be
replaced by another new painting.
Similarly, the virtual living space containing numerous
avatars (and content creators) may add new and unique con-
tents into their virtual environments in iterative manners. In
virtual environments, the creator culture can be further en-
hanced by establishing potential measurements for the preser-
vation of outdated contents, for instance, a virtual museum to
record the footprint of digital contents [616], [617]. The next
issue is how the preserved or contemporaneous digital contents
should appear in real-world environments. Ideally, everyone in
physical environments can equally access the fusing metaverse
technology, sense the physical affordances of the virtual en-
tities [618], and their contents in public urban spaces [619].
Also, the new virtual culture can influence the existing culture
in the real world, for instance, digital cultures can influence
working relationships in workspaces [620], [621].
XIV. VIRTUAL EC ONOMY
Evident in Figure 26, this section first introduces readers
to the economic governance required for the virtual worlds.
Then, we discuss the metaverse industry’s market structure
and details of economic support for user activities and content
creation discussed in the previous section.
A. Economic Governance
Throughout the past two decades, we have observed several
instances where players have created and sustained in-game
economic systems. The space theme game EVE quintessen-
tially distinguishes itself from others with a player-generated
sophisticated cobweb of an economic system, where play-
ers also take up some roles in economic governance, as
demonstrated by their monthly economic reports31. This is
not to say, however, metaverse developers can simply mimic
EVE’s success and delegate all economic governance to their
users. For one, one of the main underlying difficulties of
realising cryptocurrency as a formal means of transaction is its
association with potential deflationary pressure. Specifically,
whereas players control currency creation in EVE32, cryp-
31https://bit.ly/3o49mgM
32https://bit.ly/3u6PiLP
Fig. 26. A breakdown of sub-topics discussed in the section of Virtual
Economy, where they can be separated into two strands depending on whether
they are related to real or the virtual world. Amongst them, internal/ external
economic governance forms the bedrock of the virtual economy. Building
upon, the section discusses the metaverse industry’s market concentration in
the real world and commerce, specifically trading in the virtual world.
tocurrency is characterised by a steady and relatively slow
money supply growth due to how the ‘mining’ process is
set up. Unlike the current world we reside in, where central
banks can adjust money supply through monetary instruments
and other financial institutions can influence money supply
by creating broad money, cryptocurrency in its nascent form
simply lacks such a mechanism. Consequently, the quantity
theory of money entails that if money velocity is relatively
stable in the long term, one is justified to be concerned
about deflationary pressure as the money supply fails to
accommodate the growing amount of transactions in a thriving
metaverse [622]. Though some may posit that issuing new
cryptocurrency is a viable remedy to address the relatively
static money supply, such a method will only be viable if the
new currency receives sufficient trust to be recognised as a
formal currency. To achieve such an end, users of the meta-
verse community will have to express some level of acceptance
towards the new currency, either endogenously motivated or
through developers’ intervention. However, suppose an official
conversion rate between the newly launched cryptocurrency
and the existing one was to be enforced by developers. In
that case, they could find themselves replaying the failure of
bimetallism as speculators in the real world are incentivised to
exploit any arbitrage, leading to ‘bad’ crypto drives out ‘good’
crypto under Gresham’s Law [623]. Therefore, to break this
curse, some kind of banking system is needed to enable money
creation through fractional reserve banking [622] instead of
increasing the monetary base. Meaning that lending activities
in the metaverse world can increase the money supply. There
are already several existings platforms such as BlockFi that
allow users to deposit their cryptocurrency and offer an interest
as reward. Nevertheless, the solution does not come with no
hitch, as depositing cryptocurrency with some establishments
can go against the founding ideas of decentralisation [622].
Alternative to introducing a banking system, others have pro-
posed different means to stabilise cryptocurrency. An example
can be stabilisation through an automatic rebasing process
to national currency or commodity prices [624]. A pegged

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 33
cryptocurrency is not an imaginary concept in nowadays
world. A class of cryptocurrency known as stablecoin that
pegs to sovereign currencies already exists, and one study have
shown how arbitrage in one of the leading stablecoins, Tether,
has produced a stabilising effect on the peg [625]. Even more,
unlike the potential vulnerability of stablecoins to changes in
market sentiment on the sufficiency of collateral to maintain
the peg [625], a commonly recognised rebasing currency may
circumvent such hitch as it does not support a peg through the
use of collateral. Nonetheless, it is worth mentioning that there
has not yet been a consensus on whether cryptocurrency’s
deflationary feature should be considered as its shortcom-
ing nor the extent of deflationary pressure will manifest in
cryptocurrency in future. Additionally, another major doubt
on cryptocurrency becoming a standard form of means of
transaction arises from its highly speculative attribute. Thus,
developers should consider the economic governance required
to tune cryptocurrency into a reliable and robust currency to
be adopted by millions of metaverse users. Similarly, we have
also noticed the need of internal governance in areas such as
algorithmic fairness [626], [627], which we will discuss in
detail in Section XV-C.
Furthermore, another potential scope for economic gover-
nance emerges at a higher level: governments in our real world.
As we will show in the next section, degrees of competition
between metaverse companies can affect consumer welfare.
Therefore, national governments or even international bodies
should be entrusted to perform their roles in surveilling for
possible collusion between these firms as they do in other
business sectors. In extreme cases, the governments should
also terminate mergers and acquisitions or even break apart
metaverse companies to safeguard the welfare of consumers,
as the social ramification being at stake (i.e., control over a
parallel world) is too great to omit. That being said, economic
governance at (inter) national level is not purely regressive
towards the growth of metaverse business. Instead, state inter-
vention will play a pivotal role in buttressing cryptocurrency’s
status as a trusted medium of exchange in the parallel world.
This is because governments’ decisions can markedly shape
market sentiment. This is seen in the two opposing instances
of Turkey’s restriction33 on cryptocurrency payment and El
Salvador’s recognition of Bitcoin as legal tender34, which both
manifest as shocks to the currency market. Therefore, even
in lack of centralised control, governments’ assurances and
involvements in cryptocurrency that promise political stability
towards the currency can in return brings about stability in the
market as trust builds in. Indeed, government involvement is
a positive factor for trust in currency valued by interviewees
in a study [628]. Though it may not wholly stabilise the mar-
ket, it removes the uncertainty arising from political factors.
Furthermore, national and international bodies’ consents will
also be essential for financial engineering, such as fractional
reserve banking for cryptocurrency. Building such external
governance is not a task starting from scratch; One can learn
from past regulations on cryptocurrency and related literature
33https://reut.rs/3AEuttF
34https://cnb.cx/39COl4m
discussions [629], [630]. Nonetheless, the establishment of the
cryptocurrency banking system has another fallibility in ro-
bustness as authorities can face tremendous hardship in acting
as lender of last resort to forestall the systematic collapse of
this new banking system [631], which only increases their
burden on top of tackling illegal activities associated with
decentralised currency [632].
B. Oligopolistic Market
Fig. 27. Historical trend of Google’s annual advertising revenue34.
Observing the dominance of big tech companies in our real
world, it is no surprise for individuals like Tim Sweeney,
founder of Epic Games, to call for an ‘open metaverse’35.
With the substantial cost involved in developing a metaverse,
however, whether a shift in the current paradigm to a less con-
centrated market for metaverse will take place is questionable.
Specifically, empirical findings have shown that sunk cost is
positively correlated to an industry’s barriers to entry [633]. In
the case of the metaverse, sunk cost may refer to companies’
irretrievable costs invested in developing a metaverse system.
In fact, big corporate companies like Facebook and Microsoft
have already put their skins in the game36,37. Hence, unless
the cost of developing and maintaining a metaverse world
capable of holding millions of users drastically decreases in
the future either due to institutional factors or simply plain-
vanilla technological progress, late coming startups with a
lack of financing will face significant hardship in entering the
market. With market share concentrated at the hands of a few
leading tech companies, the metaverse industry can become
an oligopolistic market. Though it is de jure less extreme
when compared to having our parallel world dominated by
a gargantuan monopoly, the incumbent oligopolies can still
wield great power, especially at the third stage of metaverse
development (i.e., the surreality). With tech giants like Alpha-
bet generating a revenue of 147 billion dollars from Google’s
advertisements alone38 in real-life (Figure 27) shows Google’s
historical growth of advertising revenue), the potential scope
for profit in a metaverse world at the last stage of development
34https://bit.ly/3o2wGeM
35https://bit.ly/3Cwaj5w
36https://www.washingtonpost.com/technology/2021/08/30/
what-is- the-metaverse/
37https://bit.ly/3kCFOVi
38https://cnb.cx/3kztchN

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 34
cannot be neglected. The concern about “From the moment
that we wake up in the morning, until we go to bed, we’re
on those handheld tablets”39 does expose not only privacy
concerns but also the magnitude of the business potential of
owning and overseeing such a parallel world (as demonstrated
in Figure 28). However, an oligopolistic market is not entirely
malevolent. Letting alone its theoretical capability of achieving
a Pareto efficient outcome, we indeed see more desirable
outcomes specifically for rivalling tech giants’ consumers in
recent years40. Such a trend is accompanied by the rise of
players who once were outsiders to a particular tech area
but with considerable financial strength decidedly challenge
established technology firms. Therefore, despite leading tech
companies like the FANG group (Facebook, Amazon, Netflix,
and Alphabet) may prima facie be the most prominent players
in making smooth transitions to a metaverse business, it does
not guarantee they will be left uncontested by other industrial
giants which root outside of tech industry. In addition, eco-
nomic models on oligopolistic markets also provide theoretical
bedrocks for suggesting a less detrimental effect of the market
structure on consumers’ welfare provided that products are
highly differentiated and firms do not collude [634]. The
prior is already evident at the current stage of metaverse
development. Incumbent tech players, though recognising
metaverse’s diversity in scope, have approached metaverse in
differentiated manners. Whereas Fortnite inspired Sweeney’s
vision of metaverse41, Mark Zuckerberg’s recent aim was to
test out VR headsets for work42. It is understandable that given
metaverse’s uncertainties and challenges, companies choose
to approach it in areas where they hold expertise first and
eventually converges to similar directions. Having different
starting points may still result in differentiation in how each
company’s metaverse manifests. In addition, the use of differ-
ent hardware such as AR glasses and VR headsets by different
companies can also contribute to product differentiation. The
latter, however, will largely depend on economic governance,
albeit benevolent intentions held by some firms43.
C. Metaverse commerce
As an emerging concept, metaverse commerce refers to
trading taking place in the virtual world, including but not
limited to user-to-user and business-to-user trade. As com-
merce takes place digitally, the trading system can largely
borrow from the established e-commerce system we enjoy
now. For instance, with a net worth of 48.56 Billion USD50,
eBay is a quintessential example of C2C e-commerce for
the metaverse community to transplant from. Nonetheless,
39https://wapo.st/3EKDns8
40https://econ.st/3i03Sjq
41https://bit.ly/3EJGsIS
42https://cnet.co/2XG0ovg
43https://econ.st/2ZpMwpL
44https://earth2.io/
45https://bit.ly/3i0ElXw
46https://www.battlepets.finance/#/pet-shop
47https://bit.ly/39vMfDp
48https://opensea.io/collection/music
49https://bit.ly/3EHJPA5
50https://www.macrotrends.net/stocks/charts/EBAY/ebay/net-worth
Fig. 28. A scenario of a virtual world filled where advertisements are
ubiquitous. Hence demonstrating how companies in the metaverse industry,
especially when the market is highly concentrated, could possibly flood
individuals’ metaverse experiences with advertisements. The dominant player
in the metaverse could easily manipulate the user understanding of ‘good’
commerce.
metaverse commerce is not tantamount to the existing e-
commerce. Not only do the items traded differs, which will
be elaborated in the next section, but the main emphasis of
metaverse commerce is also interoperability: users’ feasibility
to carry their possessions across different virtual worlds51. The
system of the metaverse is not about creating one virtual world,
but many. Namely, users can travel around numerous virtual
worlds to gain different immersive experiences as they desire.
Therefore, as individuals can bring their possessions when
they visit another country for vacation, developers should also
recreate such experiences in the digital twin. At the current
stage, most video games, even those offered by the same
providers, do not proffer players with full interoperability
from one game to another. Real-life, however, does offer
existing games with some elements of interoperability, albeit
in lesser forms. To illustrate, games like Monster Hunter and
Pok´
emon allow players to transfer their data from Nintendo
3DS to Nintendo Switch52,53 . Nevertheless, such transfers
tend to be unilateral (e.g., from older to the newer game)
and lacks an immersive experience as they typically take
place outside the actual gameplay. Another class of games
arguably reminiscent of interoperability can be games with
downloadable contents (DLC) deriving from purchases of
other games from the same developer. A case in point could be
Capcom’s ‘Monster Hunter Stories 2”s bonus contents, where
players of the previous Capcom’s game ‘Monster Hunter Rise’
can receive an in-game outfit that originated in ‘Monster
Hunter Rise’54. However, having some virtual item bonus that
resembles users’ virtual properties in another game is not the
same as complete interoperability. An additional notable case
for interoperability for prevailing games is demonstrated in
51https://bit.ly/3CAbwbZ
52https://bit.ly/3hSzRll
53https://bit.ly/3AzUZEp
54https://bit.ly/3Cvjymo

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 35
Fig. 29. A collection of various virtual objects currently traded online: (a) Plots of land in London offered on Earth 2, an virtual replica of our planet Earth44,
(b) A virtual track roller listed on OpenSea45, (c) Virtual pet on Battle Pets46, (d) CryptoKitties47 , (e) Sound tracks listed on OpenSea48, (f) Custom made
virtual avatar on Fiverr49.
Minecraft: gamers can keep their avatars’ ‘skin’55 and ‘cape’56
when logging onto different servers, which can be perceived
as a real-world twin of metaverse players travelling between
different virtual worlds. After inspecting all three types of
existing game functions that more or less link to the notion
of interoperability, one may become aware of the lack of user
freedom as a recurring theme. Notably, inter-game user-to-
user trade is de facto missing, and the type of content, as
well as the direction of flow of contents between games,
are strictly set by developers. More importantly, apart from
the Minecraft case, there is a lack of smoothness in data
transfer as it is not integrated as part of a natural gaming
experience. That is, the actions of transferring or linking game
data is not as natural as real life behaviour of carrying or
selling goods from one place to another. Therefore, metaverse
developers should factor in the shortcomings of existing games
in addressing interoperability and promote novel solutions.
While potentially easier for metaverse organised by a sole
developer, such solutions may be more challenging to arrive
at for smaller and individual developers in a scenario of
‘open metaverse’. As separate worlds can be built in the
absence of a common framework, technical difficulties can
impede users’ connections between different virtual spaces,
let alone the exchange of in-game contents. With that being
said, organisations like the Open Metaverse Interoperability
Group have sought to connect individual virtual spaces with
a common protocol57. Hence, perhaps like the emergence of
TCP/IP protocol (i.e., a universal protocol), we need common
grounds of some sort to work on for individual metaverse
developers.
55https://minecraft.fandom.com/wiki/Skin
56https://minecraft.fandom.com/wiki/Cape#Obtaining
57https://omigroup.org/home/
D. Virtual Objects Trading
As briefly hinted in the preceding section, virtual objects
trading is about establishing a trading system for virtual
objects between different stakeholders in the metaverse. Since
human kinds first began barter trading centuries ago, trading
has been an integral part of our mundane lives. Hence, the real-
world’s digital twins should also reflect such eminent physical
counterparts. Furthermore, the need for a well-developed trad-
ing system only deepens as we move from the stage of digital
twins to digital natives, where user-created virtual contents
begin to boom. Fortunately, the existence of several real-life
exemplars sheds light on the development of the metaverse
trading system. Trading platforms for Non-Fungible Tokens
(NFTs), such as OpenSea and Rarible, allow NFT holders
to trade with one another at ease, similar to trading other
conventional objects with financial values. As demonstrated
in Figure 29, a wide range of virtual objects are being traded
at the moment. Some have gone further by embedding NFT
trading into games: Battle Pets58 and My DeFi Pet59 allow
players to nurture, battle and trade their virtual pets with
others. Given the abundance of real-life NFT trading examples,
metaverse developers can impose these structures in the virtual
world to create a marketplace for users to exchange their
virtual contents. In addition, well-known real-life auctioning
methods for goods with some degrees of common values such
as Vickrey-Clarke-Groves mechanism [635] and Simultane-
ous Multiple Round Auction [636] can also be introduced
in the virtual twin for virtual properties like franchises for
operating essential services in virtual communities such as
providing lighting for one’s virtual home. However, similar
to difficulties encountered with metaverse commerce, existing
trading systems also need to be fine-tuned to accommodate
58https://www.battlepets.finance/#/
59https://yhoo.it/3kxSNrD

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 36
the virtual world better. One potential issue can be trading
across different virtual worlds. Particularly, an object created
in world A may not be compatible in world B, especially when
different engines power the two worlds. Once again, as virtual
object trading across different worlds is intertwined with
interoperability, the call for a common framework becomes
more salient. At the current stage, some have highlighted
inspirations for constructing an integrated metaverse system
can be obtained from retrospecting existing technologies such
as the microverse architecture60,61. In Figure 30, we conjecture
how tradings between two different virtual worlds may look
like.
With more virtual objects trades at the digital natives stage
and more individuals embracing a lifestyle of digital nomad,
the virtual trading market space should also be competent
in safeguarding ownership of virtual objects. In spite of
the fact that a NFT cannot be appropriated by other users
from the metaverse communities, counterfeits can always
be produced. Specifically, after observing a user-generated
masterpiece listed on the virtual trading platform, individuals
with mischievous deeds may attempt to produce counterfeits
of it and claim for its originality. NFT-related defraud is not
eccentric, as reports have shown several cases where buyers
were deluded into thinking they were paying for legitimate
pieces from famous artists, where trading platforms lack
sufficient verification62,63. This can be particularly destructive
to a metaverse community given the type of goods being
traded. Unlike necessities traded in real life: such as staples,
water and heating, where a significant proportion of values of
these goods derive from their utilitarian functions to support
our basic needs, virtual objects’ values can depend more
on their associated social status. In other words, the act of
possessing some rare NFTs in the virtual world may be
similar to individuals consumption of Veblen goods [637]
like luxurious clothing and accessories. Therefore, the objects’
originality and rareness become a significant factor for their
pricing. Hence, a trading market flooded with feigned items
will deter potential buyers. With more buyers’ concerns about
counterfeit items and consequently becoming more reserved
towards offering a high price, genuine content creators are
disincentivised. This coincides with George Akerlof’s ‘market
of lemon’, leading to undesirable market distortion [638].
Given the negative consequences, the question to be asked
is: which stakeholder should be responsible for resolving such
a conundrum? Given that consumers tend not to possess the
best information and capacity to validate listed items, they
should not be forced to terminate their metaverse experience to
conduct an extensive search of the content creator’s credibility
in real life. Similarly, content creators are not most capable
of protecting themselves from copyright infringement as they
may be unable to price in their loss through price discrimina-
tion and price control [639]. Therefore, metaverse developers
should address the ownership issue to upkeep the market order.
60https://spectrum.ieee.org/open-metaverse?utm campaign=post-teaser&
utm content=1kp270f8
61https://microservices.io/
62https://bit.ly/3CwcZ3c
63https://www.cnn.com/style/article/banksy-nft- fake-hack/index.html
Fig. 30. Our conjecture of how virtual object trading may look like. This
figure shows two users from different virtual worlds entering a trading space
through portals (the two ellipse-shaped objects), where they trade a virtual
moped.
So far, some studies have attempted to address art forgery with
the use of neural networks by examining particular features of
an artwork [640], [641]. Metaverse developers may combine
conventional approaches by implementing a more stringent
review process before a virtual object is cleared for listing
as well as utilising neural networks to flag items that are
highly similar to items previously listed on the platform, which
may be achieved by building upon current achievements in
applications of neural networks in related fields [642], [643].
XV. SOC IA L ACC EP TABILITY
This section discusses a variety of design factors influ-
encing the social acceptability of the metaverse. The factors
include privacy threats, user diversity, fairness, user addiction,
cyberbullying, device acceptability, cross-generational design,
acceptability of users’ digital copies (i.e., avatars), and green
computing (i.e., design for sustainability).
A. Privacy threats
Despite the novel potentials which could be enabled by
the metaverse ecosystem, it will need to address the issue
of potential privacy leakage in the earlier stage when the
ecosystem is still taking its shape, rather than waiting for
future when the problem is so entrenched in the ecosystem
that any solution to address privacy concerns would require
redesign from scratch. An example of this issue is the third-
party cookies based advertisement ecosystem, where the initial
focus was to design for providing utilities. The entire revenue
model was based on cookies which keep track of users in order
to provide personalised advertisements, and it was too late
to consider privacy aspects. Eventually, they were enforced
by privacy regulations like GDPR, and the final nail to the
coffin came from Google’s decision to eliminate third-party
cookies from Chrome by 2022, which have virtually killed
the third-party cookies based advertisement ecosystem. Also,
we have some early signs of how society might react to

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 37
the ubiquitous presence of technologies that would enable
the metaverse from the public outcry against the Google
Glass, when their concerns (or perceptions) are not taken
into account. Afterwards, many solutions were presented to
respect of the privacy of bystanders and non-users [644],
[645]. However, all of them rely on the good intentions of
the device owners because there is no mechanism, either legal
or technical, in place to verify whether the bystanders’ privacy
was actually respected. Coming up with a verifiable privacy
mechanism would be one of the foremost problems to be
solved in order to receive social acceptability.
Another dimension of privacy threat in the context of social
acceptability comes from the privacy paradox, where users
willingly share their own information, as demonstrated in
Figure 31. For the most part, users do not pay attention to
how their public data are being used by other parties, but show
very strong negative reactions when the difference between
the actual use of their data and the perceived use of data
become explicit and too contrast. For example, many people
shared their data on Facebook willingly. Still, the Facebook
and Cambridge Analytica Data Scandal triggered a public
outcry to the extent that Facebook was summoned by the U.S.
Congress and the U.K. Parliament to hearings, and Cambridge
Analytica went bankrupt soon after. One solution would be
not to collect any user’s data at all. However it will greatly
diminish the potential innovations which the ecosystem could
enable. Another solution which has also been advocated by
world leaders like the German chancellor Angela Merkel, is
to enable user-consented privacy trading, where users can sell
their personal data in return for benefits, either monetary or
otherwise. Researchers have already provided their insights on
the economics of privacy [646], and the design for an efficient
market for privacy trading [647], [648]. This approach will
enable the flow of data necessary for potential innovations,
and at the same time, it will also compensate users fairly
for their data, thereby paving the path for broader social
acceptability [649].
B. User Diversity
As stated in a visionary design of human-city interac-
tion [69], the design of mobile AR/MR user interaction in city-
wide urban should consider various stakeholders. Similarly, the
metaverse should be inclusive to everyone in the community,
regardless of race, gender, age and religion, such as children,
elderly, disabled individuals, and so on. In the metaverse,
various contents can appear and we have to ensure the contents
are appropriate to vastly diversified users. In addition, it is
important to consider personalised content display in front of
users [124], and promote the fairness of the recommendation
systems, in order to minimise the biased contents and thus
impact the user behaviours and decision making [650] (More
details in Section XV-C). The contents in virtual worlds can
lead to higher acceptance by delivering factors of enjoyment,
emotional involvement, and arousal [651]. ‘How to design the
contents to maximise the acceptance level under the consid-
eration of user diversity’, i.e., design for user diversification,
would be a challenging question.
Fig. 31. The figure pictorially depicts uncontrolled data flow to every activity
under the metaverse. The digitised world with MR and the user data are
collected in various activities (Left); Subsequently, the user data is being sold
to online advertising agents without the user’s prior consents (Right).
C. Fairness
Numerous virtual worlds will be built in the metaverse, and
perhaps every virtual world has its separate rules to govern
the user behaviours and their activities. As such, the efforts
of managing and maintaining such virtual worlds would be
enormous. We expect that autonomous agents, support by
AI (Section VII), will engage in the role of governance in
virtual worlds, to alleviate the demands of manual workload.
It is important to pinpoint that the autonomous agents in
virtual worlds rely on machine learning algorithms to react
to the dynamic yet constant changes of virtual objects and
avatars. It is well-known that no model can perfectly describe
the real-world instance, and similarly, an unfair or biased
model could systematically harm the user experiences in the
metaverse. The biased services could put certain user groups
in disadvantageous positions.
On social networks, summarising user-generated texts by
algorithmic approaches can make some social groups being
under-represented. In contrast, fairness-preserving summari-
sation algorithms can produce overall high-quality services
across social groups [652]. This real-life example sheds light
on the design of the metaverse. For this reason, the metaverse
designers, considering the metaverse as a virtual society,
should include the algorithmic fairness as the core value of
the metaverse designs [627], and accordingly maintain the
procedural justices when we employ algorithms and computer
agents to take managerial and governance roles, which requires
a high degree of transparency to the users and outcome
control mechanisms. In particular, outcome controls refer to
the users’ adjustments to the algorithmic outcomes that they
think is fair [626]. Unfavourable outcomes to individual users
or groups could be devastating. This implies the importance of
user perceptions to the fairness of such machine learning algo-
rithms, i.e., perceived fairness. However, leaning to perceived
fairness could fall into another trap of outcome favorability
bias [653]. Additionally, metaverse designers should open

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 38
channels to collect the voices of diversified community groups
and collaboratively design solutions that lead to fairness in the
metaverse environments [627].
D. User Addiction
Excessive use with digital environments (i.e., user addic-
tions) would be an important issue when the metaverse be-
comes the most prevalent venue for people to spend their time
in the virtual worlds. In the worst scenario, users may leverage
the metaverse to help them ‘escaping’ from the real world, i.e.,
escapism [651]. Prior works have found shreds of evidence of
addictions to various virtual cyberspaces or digital platforms
such as social networks [654], mobile applications [655],
smartphones [656], VR [657], AR [658], and so on. User
addictions to cyberspaces could lead to psychological issues
and mental disorders, such as depression, loneliness, as well
as user aggression [659], albeit restrictions on screentime had
been widely employed [660]. Knowing that the COVID-19
pandemic has prompted a paradigm shift from face-to-face
meetings or social gatherings to various virtual ways, most
recent work has indicated that the prolonged usage of such
virtual meetings and gatherings could create another problem
– abusive use or addiction to the Internet [661].
Therefore, we have questioned whether ‘the metaverse will
bring its users to the next level of user addiction’. We discuss
the potential behaviour changes through reviewing the existing
AR/VR platforms, based not-at-all on evidence. First, VR
Chat, known as a remarkable example of metaverse virtual
worlds, can be considered as a pilot example of addiction to
the metaverse64. Meanwhile, VR researchers studied the rela-
tionship among such behavioural addiction in VR, root causes,
and corresponding treatments [662]. Also, AR games, e.g.,
Pokemon Go, could lead to the behaviour changes of massive
players, such as spending behaviours, group-oriented actions
in urban areas, dangerous or risky actions in the real world,
and such behaviour changes could lead to discernible impacts
on the society [663], [664]. A psychological view attempts
to support the occurrence of user addiction, which explains
the extended self of a user, including person’s mind, body,
physical possessions, family, friends, and affiliation groups,
encourages user to explore the virtual environments and pur-
sue rewards, perhaps in an endless reward-feedback loop, in
virtual worlds [665]. We have to pinpoint that we raise the
issues of addictions of immersive environments (AR/VR) here,
aiming at provoking debates and drawing research attentions.
In the metaverse, the users could experience super-realism
that allows users to experience various activities that highly
resemble the real world. Also, the highly realistic virtual
environments enable people to try something impossible in
their real life (e.g., replicating an event that are immoral in our
real life [666] or experiencing racist experience [667] ), with a
bold assumption that such environments can further exacerbate
the addictions, e.g., longer usage time. Further studies and
observation of in-the-wild user behaviours could help us to
understand the new factors of user addiction caused by the
super-realistic metaverse.
64https://www.worldsbest.rehab/vrchat-addiction/
E. Cyberbullying
Cyberbullying refers to the misbehaviours such as sending,
posting, or sharing negative, harmful, false, or malevolent
content about victims in cyberspaces, which frequently occurs
on social networks [668]. We also view the metaverse as
gigantic cyberspace. As such, another unignorable social threat
to the ecosystem could be cyberbullying in the metaverse.
The metaverse would not be able to run in long terms, and
authorities will request to shut down some virtual worlds in
the metaverse, according to the usual practice – shutdown the
existing cyberbullying cyberspace65. Moreover, considering
the huge numbers of virtual worlds, the metaverse would
utilise cyberbullying detection approaches are driven by al-
gorithms [669]. The fairness of such algorithms [670] will
become the crucial factors to deliver perceived fairness to the
users in the metaverse. After identifying any cyberbullying
cases, mitigation solutions, such as care and support, virtual
social supports, and self-disclosures, should be deployed effec-
tively in virtual environments [671], [672]. However, recognis-
ing cyberbullying in the game-alike environment is far more
complicated than social networks. For instance, the users’
misbehaviour can be vague and difficult to identify [673].
Similarly, 3D virtual worlds inside the metaverse could further
complicate the scenarios and hence make difficult detection of
cyberbullying at scale.
F. Other Social Factors
First, social acceptability to the devices connecting people
with the metaverse needs further investigation, which refers to
the acceptability of the public or bystanders’ to such devices,
e.g., mobile AR/VR headsets [96]. Additionally, the user safety
of mobile headsets could negatively impact the users and their
adjacent bystanders, causing breakdowns of user experience
in virtual worlds [119]. To the best of our knowledge, we
only found limited studies on social acceptability to virtual
worlds [674], but not digital twins as well as the metaverse.
Moreover, the gaps in cross-generation social networks also
indicate that Gen Z adults prefer Instagram, Snapchat and
Tiktok over Facebook. Rather, Facebook retains more users
from Gen X and Y [675]. Until now, social networks have
failed to serve all users from multiple demographics in one
platform. From the failed case, we have to prepare for the user
design of cross-generational virtual worlds, especially when
we consider the metaverse with the dynamic user cohorts in a
unified landscape.
Besides, we should consider the user acceptability of the
avatars, the digital copy of the users, at various time points. For
instance, once a user passes away, what is the acceptability of
the user’s family members, relatives, or friends to the avatars?
This question is highly relevant to the virtual immortality that
describes storing a person’s personality and their behaviours
as a digital copy [676]. The question could also shape the
future of Digital Humanity [677] in the metaverse, as we are
going to iterate the virtual environments, composed of both
virtual objects and avatars, as separate entities from the real
65https://www.change.org/p/shut-down-cyberbullying\protect\penalty-\
@M-website- ask-fm- in-memory- of-izzy- dix-12- other-teens-globally

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 39
Fig. 32. when the metaverse is enabled by numerous technologies and sensors,
the highly digitalized worlds, regardless of completely virtual (left, a malicious
avatar as a camouflage of garbage next to a garbage bin) or merged with the
physical world (right, an adjacent attack observes a user’s interaction with
immersive environments, similar to shoulder surfing attack.), will be easily
monitored (or eavesdrop) by attackers/hackers.
world, e.g., should we allow the new users talking with a
two-centuries-long avatar representing a user probably passed
away?
Furthermore, the metaverse, regarded as a gigantic digital
world, will be supported by countless computational devices.
As such, the metaverse can generate huge energy consumption
and pollution. Given that the metaverse should not deprive
future generations, the metaverse designers should not ne-
glect the design considerations from the perspective of green
computing. Eco-friendliness and environmental responsibility
could impact the user affection and their attitudes towards
the metaverse, and perhaps the number of active users and
even the opposers [678]. Therefore, sourcing and building the
metaverse with data analytics on the basis of sustainability
indices would become necessary for the wide adoption of the
metaverse [679], [680].
Finally, we briefly mention other factors that could impact
the user acceptability to the metaverse, such as in-game
injuries, unexpected horrors, user isolation, accountability and
trust (More details in Section XVII), identity theft/leakage, vir-
tual offence, manipulative contents inducing user behaviours
(e.g., persuasive advertising), to name but a few [681], [682].
XVI. PRI VACY AN D SEC UR IT Y
Internet-connected devices such as wearables allow mon-
itoring and collect users’ information. This information can
be interpreted in multiple ways. In most situations, such as
in smart homes, we are not even aware of such ubiquitous
and continuous recordings, and hence, our privacy can be at
risk in ways we cannot foresee. These devices can collect
several types of data: personal information (e.g., physical,
cultural, economic), users’ behaviour (e.g., habits, choices),
and communications (e.g., metadata related to personal com-
munications). In many situations, users accept the benefits in
comparison with the possible privacy and security risks of
using these smart devices or services [470]. For example, GPS
positioning is used to search for nearby friends [683]. In the
case of VR - the primary device used to display the metaverse
- the new approaches to enable more immersive environments
(e.g., haptic devices, wearables to track fine-grained users’
movements) can threaten the users in new ways.
The metaverse can be seen as a digital copy of what we
see in our reality, for example, buildings, streets, individuals.
However, the metaverse can also build things that do not exist
in reality, such as macro concerts with millions of spectators
(Figure 32). The metaverse can be perceived as a social
microcosmos where players (individuals using the metaverse)
can exhibit realistic social behaviour. In this ecosystem, the
privacy and security perceptions of individuals can follow
the real behaviours. In this section, we will elaborate on the
privacy and security risks that individuals can face when using
the metaverse. We start with an in-depth analysis of the users’
behaviour in the metaverse and the risks they can experience,
such as invasion of privacy or continuous monitoring, and
privacy attacks that individuals can suffer in the metaverse
such as deep-fakes and alternate representations. Second, we
evaluate how designers and developers can develop ethical
approaches in the metaverse and protect digital twins. Finally,
we focus on the biometric information that devices such as
VR headsets and wearables can collect about individuals when
using the metaverse.
A. Privacy behaviours in the metaverse
In the metaverse, individuals can create avatars using similar
personal information, such as gender, age, name, or entirely
fictional characters that do not resemble the physical appear-
ance or include any related information with the real person.
For example, in the game named Second Life – an open-world
social metaverse - the players can create their avatars with
full control over the information they want to show to other
players.
However, due to the nature of the game, any player can
monitor the users’ activities when they are in the metaverse
(e.g., which places they go, whom they talk to). Due to the
current limitations of VR and its technologies, users cannot
be fully aware of their surroundings in the metaverse and who
is following them. The study by [470] shows that players do
behave similarly in the metaverse, such as Second Life, and
therefore, their privacy and security behaviours are similar to
the real one. As mentioned above, players can still suffer from
extortion, continuous monitoring, or eavesdropping when their
avatars interact with other ones in the metaverse.
A solution to such privacy and security threats can be the use
of multiple avatars and privacy copies in the metaverse [470].
The first technique focuses on creating different avatars with
different behaviour and freedom according to users’ pref-
erences. These avatars can be placed in the metaverse to
confuse attackers as they will not know which avatar is the
actual user. The avatars can have different configurable (by
the user) behaviours. For example, when buying an item in
the metaverse, the user can generate another avatar that buys
a particular set of items, creating confusion and noise to the

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 40
attacker who will not know what the actual avatar is. The
second approach creates temporary and private copies of a
portion of the metaverse (e.g., a park). In this created and
private portion, attackers can not eavesdrop on the users. The
created copy from the main fabric of the metaverse will or
not create new items (for example, store items). Then in the
case the private portion use resources from the main fabric, the
metaverse API should address the merge accordingly from the
private copy to the main fabric of the metaverse. For example,
if the user creates a private copy of a department store, the
bought items should be updated in the store of the main fabric
when the merge is complete. This will inherently create several
challenges when multiple private copies of the same portion of
the metaverse are being used simultaneously. Techniques that
address the parallel use of items in the metaverse should be
implemented to avoid inconsistencies and degradation of the
user experience (e.g., the disappearance of items in the main
fabric because they are being used in a private copy). Finally,
following the creation of privacy copies, the users can also be
allowed to create invisible copies of their avatar so they can
interact in the metaverse without being monitored. However,
this approach will suffer from similar challenges as the private
copies when the resources of the main fabric are limited or
shared.
In these virtual scenarios, the use of deep-fakes and alternate
representations can have a direct effect on users’ behaviours,
e.g., Figure 33. In the metaverse, the generated virtual worlds
can open potential threats to privacy more substantial than
in the real world. For example, ‘deep-fakes’ can have more
influence in users’ privacy behaviours. Users can have trouble
differentiating authentic virtual subjects/objects from deep-
fakes or alternate representations aiming to ‘trick’ users. The
attackers can use these techniques to create a sense of urgency,
fear, or other emotions that lead the users to reveal personal
information. For example, the attacker can create an avatar
that looks like a friend of the victim to extract some personal
information from the latter. In other cases, the victim’s security
can be at stake, such as physically (in the virtual world)
assaulting the victim. Finally, other more advanced techniques
can use techniques, such as dark patterns, to influence users
into unwanted or unaware decisions by using prior logged
observations in the metaverse. For example, the attacker can
know what the users like to buy in the metaverse, and he/she
will design a similar virtual product that the user will buy
without noticing it is not the original product the user wanted.
Moreover, machine learning techniques can enable a new way
of chatbots and gamebots in the metaverse. These bots will
use the prior inferred users’ traits (e.g., personality) to create
nudged [684] social interactions in the metaverse.
B. Ethical designs
As we mentioned above, the alternate representations and
deep-fakes that attackers can deliver in the metaverse should be
avoided. First, we discuss how the metaverse can be regulated
and even the governance possibilities in the metaverse.
For example, Second Life is operated in the US, and there-
fore, it follows US regulations in terms of privacy and security.
Fig. 33. One possible undesirable outcome in the metaverse – occupied
by advertising content. The physical world (left) is occupied with full of
advertising content in immersive views (right). This may apply to the users
without a premium plan, i.e., free users. Users have to pay to remove such
unwanted content. More importantly, if the digital content appears in the real
world with the quality of super-realism, the human users may not be able to
distinguish the tangible content in the real world. User perceptions in the real
world can be largely distorted by the dominant players in the metaverse.
However, the metaverse can achieve worldwide proportions
creating several challenges to protect the users in such a broad
spectrum. The current example of Second Life shows an in-
world (inside the metaverse) regulation and laws. In this envi-
ronment, regulations are enforced using code and continuous
monitoring of players (e.g., chat logs, conversations). The
latter can help the metaverse developers to ban users after
being reported by others. However, as we can observe, this
resembles some governance. This short of governance can
interfere with the experience in the metaverse, but without
any global control, the metaverse can become anarchy and
chaos. This governance will be in charge of decisions such as
restrictions of a particular player that has been banned.
In the end, we can still face the worldwide challenges of
regulations and governance in the metaverse to have some
jurisdiction over the virtual world. We can foresee that the
following metaverse will follow previous approaches in terms
of regulations (according to the country the metaverse oper-
ates) and a central government ruled by metaverse developers
(using code and logs).
Some authors [470] have proposed the gradual implemen-
tation of tools to allow groups to control their members
similarly as a federated model. Users in the metaverse can
create neighbours with specific rules. For example, users can
create specific areas where only other users with the same
affinities can enter. Technologies such as blockchain can also
allow forcing users of the metaverse to not misbehave accord-
ing to some guidelines, with the corresponding punishment
(maybe by democratic approaches). However, the regulations
of privacy and security and how to enforce them are out of
this section’s scope.
1) Digital twins protection: Digital twins are virtual objects
created to reflect physical objects. These digital objects do

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 41
not only resemble the physical appearance but can also the
physical performance or behaviour of real-world assets. Digital
twins will enable clones of real-world objects and systems.
Digital twins can be the foundation of the metaverse, where
digital objects will behave similarly to physical ones. The
interactions in the metaverse can be used to improve the
physical systems converging in a massive innovation path and
enhanced user experience.
In order to protect digital twins, the metaverse has to ensure
that the digital twins created and implemented are origi-
nal [685]. In this regard, the metaverse requires a trust-based
information system to protect the digital twins. Blockchain is a
distributed single chain, where the information is stored inside
cryptographic blocks [286]. The validity of each new block
(e.g., creation of a new digital twin) is verified by a peer-
to-peer network before adding the new record to the chain.
Several works [686]–[688] propose the use of blockchain sys-
tems to protect the digital twins in the metaverse. In [688], the
authors propose a blockchain-based system to store health data
electronically (e.g., biometric data) health records that digital
twins can use. As we have seen with recent applications, they
can enable new forms of markets in the digital ecosystems
such as non-fungible token (NFT) [687]. The latter allows
digital twins creators to sell their digital twins as unique assets
by using the blockchain.
2) Biometric data: The metaverse uses data from the
physical world (e.g., users’ hand movements) to achieve an
immersive user [689]. For example, different sensors attached
to users (e.g., gyroscope to track users’ head movements)
can control their avatar more realistically. Besides VR head-
mounted displays, wearables, such as gloves and special suits,
can enable new interaction approaches to provide more real-
istic and immersive user experiences in the metaverse. These
devices can allow users to control their avatar using gestures
(e.g., glove-based hand tracking devices) and render haptic
feedback to display more natural interactions. The goal of
capturing such biometric information is to integrate this mixed
modality (input and output) to build a holistic user experience
in the metaverse, including avatars’ interactions with digital
assets such as other avatars [689].
However, all these biometric data can render more im-
mersive experiences whilst opening new privacy threats to
users. Moreover, as previously commented, digital twins used
real-world data such as users’ biometric data (e.g., health
monitoring and sport activities) to simulate more realistic
digital assets in the metaverse. Therefore, there exists a need to
protect such information against attacks while still accessible
for digital twins and other devices (e.g., wearables that track
users’ movements).
XVII. TRU ST A ND AC CO UN TABILITY
As the advancements in the Internet, Web technologies, and
XR converge to make the idea of the metaverse technically
feasible. And the eventual success would depend on how
likely are users willing to adopt it, which further depends
on the perceived trust and the accountability in the event of
unintended consequences.
A. Trust and Information
Socrates did not want his words to go fatherless into
the world, transcribed onto tablets or into books that could
circulate without their author, to travel beyond the reach of
discussion and questions, revision, and authentication. So, he
talked and augured with others on the streets of Athens, but he
wrote and published nothing. The problems to which Socrates
pointed are acute in the age of recirculated “news”, public
relations, global gossip, and internet connection. How can
rumours be distinguished from the report, fact from fiction,
reliable source from disinformation, and trust-teller from de-
ceiver? These problems have already been proven to be the
limiting factor for the ubiquitous adoption of social networks
and smart technologies, as evident from the migration of users
in many parts of the world from supposedly less trustworthy
platforms (i.e., WhatsApp) to supposedly higher trustworthy
platforms (i.e., Signal) [690]. For the same reason, in order
for the convergence of XR, social networks, and the Internet
to be truly evolved to the metaverse, one of the foremost
challenges would to establish a verifiable trust mechanism. A
metaverse universe also has the potential to solve many social
problems, such as loneliness. For example, Because of the
COVID-19 pandemic, or the lifestyle of the elderly in urban
areas, the elderly people were forced to cancel the activities for
their physical conditions or long distances. However, elderly
people are almost most venerable to online scams/frauds,
which makes coming up with solutions for the trust mechanism
quite imperative.
As in the metaverse universe, users are likely to devote
more time to their journeys in immersive environments, and
they would leave themselves vulnerable by exposing their
actions to other (unknown) parties. This can present another
limiting factor. Some attempts have been to address this
concern by exploiting the concept of “presence”, i.e., giving
users “place illusion” defined as the sensation of being there,
and “plausibility presence” defined as the sensation that the
events occurring in the immersive environment are actually
occurring [691]. However, it remains to be seen how effective
this approach is on a large scale.
Another direction towards building trust could be from the
perspective of situational awareness. Research on trust in au-
tomation suggests that providing insight into how automation
functions via situational awareness display improve trust [692].
XR can utilise the same approach of proving such information
to the user’s view in an unobtrusive manner in the immersive
state.
Dependability is also considered as an important aspect
of trust. Users should be able to depend on XR technolo-
gies to handle their data in a way they expect to. Re-
cent advances in trusted computing have paved a path for
hardware/crypto-based trusted execution environments (TEEs)
in mobile devices. These TEEs provide for secure and isolated
code execution and data processing (cryptographically sealed
memory/storage), as well as remote attestation (configuration
assertions). The critical operations on user’s data can be done
through TEEs. However, the technology is yet to be fully
developed to be deployed in XR devices while ensuring real-

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 42
time experience.
On the flip side, there is also a growing concern of over-
trust. Users tend to trust products from big brands far too
easily, and rightly so, since human users have often relied
on using reputation as predominant metric to decide whether
to trust a product/service from the given brand. However,
in the current data-driven economy where user’s information
are a commodity, even big brands have been reported to
engage in practices aimed to learn about the user as much as
possible [693], i.e., Google giving access of users’ emails to
the third parties [694]. This concern is severe in XR, because
XR embodies human-like interactions, and their misuses by
the third parties can cause significant physiological trauma to
users. The IEEE Global Initiative on Ethics of Autonomous
and Intelligent Systems recommends that upon entering any
virtual realm, users should be provided a “hotkey” tutorial on
how to rapidly exit the virtual experience, and information
about the nature of algorithmic tracking and mediation within
any environment [695]. Technology designers, standardisa-
tion bodies and regulatory bodies will also need to consider
addressing these issues under consideration for a holistic
solution.
B. Informed Consent
In the metaverse system, a great amount of potentially
sensitive information is likely to leave the owner’s sphere of
control. As in the physical world during face-to-face communi-
cation, we place trust since we can check the information and
undertakings others offer, similarly we will need to develop
the informed consent mechanism which will allow avatars, i.e.,
the virtual embodiment of users, to place their trust. Such a
consent mechanism should be allowed consent to be given or
refused in the light of information that should be checkable.
However, the challenges arise from the fact that avatars may
not capture the dynamics of a user’s facial expression in real-
time, which are important clues to place trust in face-to-face
communications.
Another challenge that the metaverse universe will need
to address is how to handle sensitive information of the
minors since minors constitute a wide portion of increasingly
sophisticated and tech-savvy XR users. They are traditionally
less aware of the risks involved in the processing of their data.
From a practical standpoint, it is often difficult to ascertain
whether a user is a child and, for instance, valid parental
consent has been given. Service providers in the metaverse
should accordingly review the steps they are taking to pro-
tect children’s data on a regular basis and consider whether
they can implement more effective verification mechanisms,
other than relying upon simple consent mechanisms. Devel-
oping a consent mechanism for metaverse can use general
recommendations issued by the legal bodies, such as Age
Appropriate Design Code published by the UK Information
Commissioner’s Office.
Designing a consent mechanism for users from venera-
ble populations will also require additional consideration.
Vulnerable populations are those whose members are not
only more likely to be susceptible to privacy violations, but
Fig. 34. What is the user responsibilities with their digital copies as the
avatars? For instance, the avatar’s autonomous actions damage some properties
in the metaverse.
whose safety and well-being are disproportionately affected by
such violations, are likely to suffer discrimination because of
their physical/mental disorder, race, gender or sex, and class.
Consent mechanisms should not force those users to provide
sensitive information which upon disclosing may further harm
users [696].
Despite an informed consent mechanism already in place,
it may not always lead to informed choice form presenting
to users. Consent forms contain technical and legal jargon
and are often spread across many pages, which users rarely
read. Oftentimes, users go ahead to the website contents with
the default permission settings. An alternative way would
be to rely on the data-driven consent mechanism that learns
user’s privacy preferences, change permission settings for data
collection accordingly, and also considers that user’s privacy
preferences may change over time [697], [698].
C. Accountability
Accountability is likely to be one of the major keys to
realising the full potential of the metaverse ecosystem. De-
spite the technological advances making ubiquitous/pervasive
computing a reality, many of the potential benefits will not
be realised unless people are comfortable with and embrace
the technologies, e.g., Figure 34. Accountability is crucial for
trust, as it relates to the responsibilities, incentives, and means
for recourse regarding those building, deploying, managing,
and using XR systems and services.
Content moderation policies that detail how platforms and
services will treat user-generated content are often used in
traditional social media to hold users accountable for the
content which they generate. As outlined in Section XII, in
the metaverse universe, users are likely to interact with each
other through their avatars, which already obfuscates the user’s
identity to a certain extent. Moreover, recent advances in multi-
modal machine learning can be used for machine-generated
3D avatars [699]. The metaverse content moderation will
foremost need to distinguish where the given avatar embodies

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 43
a human user, or is simply an auto-troll, since the human users
are entitled to the freedom of expression, barring the cases
of violent/extremist content, hate speech, or other unlawful
content. In recent years, the content moderation of a popular Q
& A website, Quora, has received significant push-backs from
users primarily based in the United States, since the U.S. based
users are used to the freedom of expression in an absolute
sense, and expect the same in the online world. One possible
solution could be to utilise the constitutional rights extended to
users in a given location to design the content moderation for
that location. However, in the online world, users often cross
over the physical boundary, thus making constitutional rights
as the yardstick to design content moderation also challenging.
Another aspect of accountability in the metaverse universe
comes from the fact how users’ data are handled since XR
devices inherently collect more sensitive information like the
user’s locations and their surroundings than the traditional
smart devices. Privacy protection regulations like GDPR rely
on the user’s consent, and ’Right to be forgotten’, to address
this problem. But, oftentime, users are not entirely aware of
potential risks and invoke their “Right to be forgotten” mostly
after some unintended consequences have already occurred.
To tackle this issue, the metaverse universe should promote
the principle of data minimisation, where only the minimum
amount of user’s data necessary for the basic function are
collected, and the principle of zero-knowledge, where the
systems retain the user’s data only as long as it is needed [700].
Another direction worth exploring is utilising blockchain tech-
nology to operationalise the pipeline for data handling which
always follows the fixed set of policies that have been already
consented to. The users can always keep track of their data,
i.e., keep track of decision provenance [701].
In traditional IT systems, auditing has often be used as a
way to ensure the data controllers are accountable to their
stakeholders [702]. Auditors are often certified third parties
which do not have a conflict of interest with the data con-
trollers. In theory, auditing can be used in the metaverse as
well. However, it faces the challenge regarding how to audit
secondary data which were created from the user’s data, but
it is difficult to establish the relationship between a given
secondary data and the exact primary data, thus making it
challenging for the auditor to verify whether the wishes of
the users which withdrew their consent, later on, were indeed
respected. The current data protection regulation like GDPR
explicitly focuses on personally identifiable data and does not
provide explicit clarity on the secondary data. This issue also
relates the data ownership in the metaverse, which is still under
debate.
Apart from the data collection, stakes are even higher in
the metaverse, since unintended consequences could cause
not only psychological damage, but also physical harm [703].
For example, the projection of digital overlays by the user’s
XR mobile headsets may contain critical information, such as
manholes or the view ahead, which may cause life-threatening
accidents. Regulatory bodies are still debating how to set up
liabilities for incidents that are triggered by machines taking
away user’s full attention. In 2018, a self-Driving Uber Car
which had a human driver killed a pedestrian in Arizona [704].
The accident could have been avoided if the human operator’s
full attention was on the driving. However, mandating full
human attention all the time also diminishes the role of
these assistive technologies. Regulatory bodies will need to
consider broader contexts in the metaverse to decide whether
the liability in such scenarios belong to the user, the device
manufacturer, or any other third parties.
XVIII. R ES EA RCH AGE NDA AND GRAND C HA LL EN GE S
We have come a long way since the days of read-only online
content on desktop computers. The boundary between virtual
and physical environments has become more blurred than ever
before. As such, we are currently in the midst of the most
significant wave of digital transformation, where the advent
of emerging technology could flawlessly bind the physical and
digital twins together and eventually reach the Internet featured
with immersive and virtual environments.
As mentioned in Section I, the migration towards such an
integration of physical and virtual consists of three stages:
digital twins, digital natives, and the metaverse. As such, our
immersive future with the metaverse requires both efforts to
technology development and ecosystem establishment. The
metaverse should own perpetual, shared, concurrent, and
3D virtual spaces that are concatenated into a perceived
virtual universe. We expect the endless and permanent
virtual-physical merged cyberspace to accommodate an un-
limited number of users, not only on earth, but eventual im-
migrants living on other planets (e.g., the moon and the mars),
would inter-planetary travel and communication develop [705].
Technology enablers and their technical requirements are
therefore unprecedentedly demanding. The metaverse also
emphasises the collection of virtual worlds and the rigorous
activities in such collective virtual environments where human
users will spend their time significantly. Thus, a complete set
of economic and social systems will be formed in the meta-
cyberspace, resulting in new currency markets, capital markets,
commodity markets, cultures, norms, regulations, and other
societal factors.
Figure 35 illustrates a vision of building and upgrading
cyberspace towards the metaverse in the coming decade(s).
It is worthwhile to mention that the fourteen focused areas
pinpointed in this survey are interrelated, e.g., [450] leverages
IoT, CV, Edge, Network, XR, and user interactivity in its
application design. Researchers and practitioners should view
all the areas in a holistic view. For instance, the metaverse
needs to combine the virtual world with the real world, and
even the virtual world is more realistic than the real world. It
has to rely on XR-driven immersive technologies to integrate
with one or more technologies, such as edge and cloud (e.g.,
super-realism and zero-latency virtual environments at scale),
avatar and user interactivity (e.g., motion capture and gesture
recognition seamlessly with XR), artificial intelligence and
computer vision for scene understanding between MR and the
metaverse and the creation of creating digital twins at scale,
Edge and AI (Edge AI) together for privacy-preserving AI
applications in the metaverse, and to name but a few.
In the remaining of this section, we highlight the high-level
requirements of the eight focused technologies for actualising

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 44
Fig. 35. The figure depicts a future roadmap for metaverse three-stage developments towards the surreality, i.e., the concept of duality and the final stage
of co-existence of physical-virtual reality. The technology enablers and ecosystem drivers help achieve self-sustaining, autonomous, perpetual, unified and
endless cyberspace.
the metaverse. Accordingly, we pinpoint the six ecosystem as-
pects that could lead to the long-term success of the metaverse.
Extended Reality. The metaverse moves from concept to
reality, and VR/AR/MR is a necessary intermediate stage.
To a certain extent, virtual environments are the technical
foundation of the metaverse. The metaverse is a shared virtual
space that allows individuals to interact with each other in the
digital environment. Users exist in such a space as concrete
virtual images, just like living in a world parallel to the
real world. Such immersive technologies will shape the new
form of immersive internet. VR will allow users to obtain a
more realistic and specific experience in the virtual networked
world, making the virtual world operation more similar to the
real world. Meanwhile, AR/MR can transform the physical
world. As a result, the future of our physical world is more
closely integrated with the metaverse.
More design and technical considerations should address the
scenarios when digital entities have moved from sole virtual
(VR) to physical (MR) environments. Ideally, MR and the
metaverse advocate full integration of virtual entities with the
physical world. Hence, super-realistic virtual entities merging
with our physical surroundings will be presented everywhere
and anytime through large displays, mobile headsets, or holog-
raphy. Metaverse users with digital entities can interplay and
inter-operate with real-life objects. As such, XR serves as
a window to enable users to access various technologies,

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 45
e.g., AI, computer vision, IoT sensors and other five focused
technologies, as below discussed.
User Interactivity. Mobile techniques for user interaction
enable users to interact with digital overlays through the
lens of XR. Designing mobile techniques in body-centric,
miniature-sized and subtle manners can achieve invisible com-
puting interfaces for ubiquitous user interaction with virtual
environments in the metaverse. Additionally, multi-modal feed-
back cues and especially haptic feedback on mobile techniques
allow users to sense the virtual entities with improved senses
of presence and realism with the metaverse, as well as to work
collaboratively with IoT devices and service robots.
On the other hand, virtual environments (VR/AR/MR) are
enriched and somehow complex, which can only give people
a surreal experience of partial senses, but cannot realise the
sharing and interaction of all senses. Brain-Computer Interface
(BCI) technology, therefore, stands out. Brain-computer inter-
face technology refers to establishing a direct signal channel
between the human brain and other electronic devices, thereby
bypassing language and limbs to interact with electronic
devices. Since all human senses are ultimately formed by
transmitting signals to the brain, if brain-computer interface
technology is used, in principle, it will be able to fully simulate
all sensory experiences by stimulating the corresponding areas
of the brain. Compared with the existing VR/AR headsets,
a brain-computer interface directly connected to the human
cerebral cortex (e.g., Neuralink66) is more likely to become
the best device for interaction between players and the virtual
world in the future meta-universe era.
IoT and Robotics. IoT devices, autonomous vehicles and
Robots leverage XR systems to visualise their operations and
invite human users to co-participate in data management
and decision-making. Therefore, presenting the data flow in
comfortable and easy-to-view manners are necessary for the
interplay with IoTs and robots. Meanwhile, appropriate de-
signs of XR interfaces would fundamentally serve as a medium
enabling human-in-the-loop decision making. To the best of
our knowledge, the user-centric design of immersive and
virtual environments, such as design space of user interfaces
with various types of robotics, dark patterns of IoT and
robotics, subtle controls of new robotic systems and so on, are
in their nascent stage. Therefore, more research studies can be
dedicated to facilitating the metaverse-driven interaction with
IoT and robots.
Artificial Intelligence. The application of artificial intel-
ligence, especially deep learning, makes great progress in
automation for operators and designers in the metaverse, and
achieves higher performance than conventional approaches.
However, applying artificial intelligence to facilitate users’
operation and improve the immerse experience is lacking. Ex-
isting artificial intelligence models are usually very deep and
require massive computation capabilities, which is unfriendly
for resource-constrained mobile devices. Hence, designing
light but efficient artificial intelligence models is necessary.
Blockchain. With the increasing demand for decentralised
content creation in the metaverse, NFT is playing a more
66https://neuralink.com/
critical role. NFT enables the created properties to be traded
with customised values. However, the research on NFT is still
in the early phase. Currently, most NFT solutions are based
on Ethereum. Hence, the drawbacks, e.g., slow confirmation
and high transaction cost, are naturally inherited. Further-
more, blockchain adopts the proof of work as the consensus
mechanism, which requires participants to spend effort on
puzzles to guarantee data security. However, the verification
process for encrypted data is not as fast as conventional
approaches. Hence, faster proof of work to accelerate the data
accessing speed and scalability is a challenge to be solved.
Currently, more than 60$ is required to mine an NFT token,
which is obviously too much for small-scale transactions.
Anonymity is another challenge. Most NFT schemes adopt
pseudo-anonymity, instead of strict anonymity, which may lead
to privacy leakage.
Computer Vision. Computer vision allows computing de-
vices to understand the visual information of the user’s ac-
tivities and their surroundings. To build more a reliable and
accurate 3D virtual world in the metaverse, computer vision
algorithms need to tackle the following challenges. Firstly, in
the metaverse, an interaction system needs to understand more
complex environments, in particular, the integration of virtual
objects and physical world. Therefore, we expect more precise
and computationally effective spatial and scene understanding
algorithms to use soon for the metaverse.
Furthermore, more reliable and efficient body and pose
tracking algorithms are needed as metaverse is closed con-
nected with the physical world and people. Lastly, in the meta-
verse, colour correction, texture restoration, blur estimation
and super-resolution also play important roles in ensuring a
realistic 3D environment and correct interaction with human
avatars. However, it is worth exploring more adaptive yet
effective restoration methods to deal with the gap between
real and virtual contents and the correlation with avatars in
the metaverse.
Edge and Cloud. The last mile latency especially for
mobile users (wireless connected) is still the primary latency
bottleneck, for both Wi-Fi and cellular networks, thus the
further latency reduction of edge service relies on the im-
provement of the last mile transmission, e.g., 1 ms promised
by 5G, for seamless user experience with the metaverse.
Also, MEC involves multiple parties such as vendors, ser-
vice providers, and third-parties. Thus, multiple adversaries
may be able to access the MEC data and steal or tamper the
sensitive information. Regarding security, in the distributed
edge environment at different layers, even a small portion com-
promised edge devices could lead to harmful results for the
whole edge ecosystem and hence the metaverse services, e.g.,
feature inference attack in federated learning by compromising
one of the clients.
Network. The major challenges related to the network itself
are highly related to the typical performance indicators of
mobile networks, namely latency and throughput, as well
as the jitter, critical in ensuring a smooth user experience.
User mobility and embodied sensing will further complicate
this task. Contrary to the traditional layered approach to
networks, where minimal communication happens between

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 46
Fig. 36. Two perspectives: 1) “In the Metaverse, nobody knows that you are
a cat.” is analogue to “On the Internet, no one knows that you are a dog”. 2)
Metaverse can become a new horizon in Animal-Computer Interaction (ACI),
e.g., a virtual environment as ‘Kittiverse’. How to prepare the metaverse going
beyond human users (or human avatars)?
layers, addressing the strict requirements of user experience in
the metaverse will require two-way communication between
layers. 5G and its successors will enable the gNB to communi-
cate network measurements to the connected user equipment,
that can be forwarded to the entire protocol stack up to the
application to adapt the transmission of content. Similarly, the
transport layer, where congestion control happens, may signal
congestion to the application layer. Upon reception of such
information, the application may thus reduce the amount of
data to transmit to meet the throughput, bandwidth, and latency
requirements. Similarly, QoE measurements at the application
layer may be forwarded to the lower layers to adapt the
transmission of content and improve the user experience.
Avatar. Avatars serve as our digital representatives in the
metaverse. Users would rely on the avatars to express them-
selves in virtual environments. Although the existing technol-
ogy can capture the features of our physical appearance and
automatically generate an avatar, the ubiquitous and real-time
controls of avatars with mobile sensors are still not ready for
mobilising our avatars in the metaverse. Additional research
efforts are required to enhance the Micro-expression and
non-verbal expression of the avatars. Moreover, current gaps
in understanding the design space of avatars, its influences
to user perception (e.g., super-realism and alternated body
ownership), and how the avatars interact with vastly diversified
smart devices (IoTs, intelligent vehicles, Robots), should be
further addressed. The avatar design could also go farther
than only human avatars. We should consider the following
situations (Figure 36): either human users employ their pets
as avatars in the metaverse, or when human users and their
pets (or other animals) co-exist in the metaverse and hence
enjoy their metaverse journey together.
Meanwhile, the ethical design of avatars and their corre-
sponding behaviours/representation in cyberspace would also
be a complicated issue. The metaverse could create a grey
area for propagating offensive messages, e.g., race and could
raise debate and prompt a new perspective to our identity.
An avatar creates a new identity of oneself in the metaverse,
potentially raises a debate, and prompts new thinking of human
life. That is, the digital clone of humanity in the metaverse
will live forever. Thus, even if the physical body, in reality, is
annihilated, you in the digital world will continue to live in the
meta-universe, retaining your personality, behavioural logic,
and even memories in the real world. If this is the case, the
metaverse avatars bring technical and design issues and ethical
issues of the digital self. Is the long-lasting avatar able to fulfil
human rights and obligations? Can it inherit my property? Is
it still the husband of the father and wife of my child in the
real world?
Content Creation. Content Creation should not be lim-
ited to professional designers – it is everyone’s right in the
metaverse. Considering various co-design processes, such as
Participatory design, would encourage all stakeholders in the
metaverse to create the digital world together. Investigating
the Motivations and Incentives would enable the participa-
tory design to push the progress of content creation in the
metaverse. More importantly, the design and implementation
of automatic and decentralised governance of censorship
are unknown. Also, we should consider the establishment
of creator cultures with cultural diversity, cross-generational
contents, and preservation of phase-out contents (i.e., digital
heritage).
Virtual Economy. When it comes to the currency for the
metaverse, the uncertainty revolves around the extent to which
cryptocurrency can be trusted to function as money, as well
as the innovation required to tailor it for the virtual world.
Moreover, as the virtual world users will also be residents
of the real world, the twin virtual and real economies will
inevitably be intertwined and should not be treated as two
mutually exclusive entities. Therefore, a holistic perspective
should be adopted when examining what virtual economy truly
means for the metaverse ecosystem.
Areas to be considered holistically include individual
agent’s consumption behaviours in the virtual and real world
as well as how aggregate economic activities in the two worlds
can affect each other. In addition, a virtual world that is
highly similar to the real world can potentially be used as a
virtual evaluation sandbox to test out new economic policies
before we implement them in real life. Hence, to harness such
merit, we need a conversion mechanism that optimally sets
up the computer-mediated sandbox to properly simulate the
reality with an accurate representation of economic agents’
incentives.
Social Acceptability. Social acceptability is the reflection
of metaverse users’ behaviours, representing collective judge-
ments and opinions of actions and policies. The factors of
social acceptability, such as privacy threats, user diversity,
fairness, and user addiction, would determine the sustainability
of the metaverse. Furthermore, as the metaverse would impact
both physical and virtual worlds, complementary rules and
norms should be enforced across both worlds.
On the other hand, we presume the existing factors of
social acceptability can be applied to the metaverse. However,
manual matching of such factors to the enormous metaverse
cyberspace will be tedious and not affordable. And examining
such factors case by case is also tedious. Automatic adoption

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 47
of rules and norms and subsequently the evaluation with social
acceptability, to understand the collective opinions, would
rely on many autonomous agents in the metaverse. Therefore,
designing such agents at scale in the metaverse wold become
an urgent issue.
More importantly, as the metaverse will be integrated into
every aspect of our life, everyone will be impacted by this
emerging cyberspace. Designing strategies and technologies
for fighting cybercrime and reporting abuse would be crucial
to improving the enormous cyberspace’s social acceptability.
Security and Privacy. As for security, the highly digitised
physical world will require users frequently to authenticate
their identities when accessing certain applications and ser-
vices in the metaverse, and XR-mediated IoTs and mechanised
everyday objects. Additionally, protecting digital assets is the
key to secure the metaverse civilisations at scale. In such
contexts, asking textual passwords for frequent metaverse
applications would be a huge hurdle to streamline the authen-
tications with innumerable objects. The security researchers
would consider new mechanisms to enable application au-
thentications with alternative modalities, such as biometric
authentication, which is driven by muscle movements, body
gestures, eye gazes, etc. As such, seamless authentication
could happen with our digitised journey in various physi-
cal contexts – as convenient as opening a door. However,
such authentication system still requirements improvements in
multitudinous dimensions, especially security levels, detection
accuracy and speed, as well as the device acceptability.
On the other hand, countless records of user activities
and user interaction traces will remain in the metaverse.
Accordingly, the accumulated records and traces would cause
privacy leakages in the long term. The existing consent forms
for accessing every website in 2D UIs would make users
overwhelmed. Users with virtual 3D worlds cannot afford such
frequent and recurring consent forms. Instead, it is necessary
to design privacy-preserving machine learning to automate
the recognition of user privacy preference for dynamic yet
diversified contexts in the metaverse.
The creation and management of our digital assets such
as avatars and digital twins can also have great challenges
when protecting users against digital copies. These copies can
be created to modify users’ behaviour in the metaverse and
for example share more personal information with ‘deep-fake’
avatars.
Trust and Accountability. The metaverse, i.e., convergence
of XR and the Internet, expands the definition of personal
data, including biometrically-inferred data, which is prevalent
in XR data pipelines. Privacy regulations alone can not be
the basis of the definition of personal data, since they can
not cope up with the pace of innovation. One of the major
grand challenges would be to design a principled framework
that can define personal data while keeping up with potential
innovations.
As human civilisation has progressed from the past towards
the future, it has accommodated the rights of minorities,
though after many sacrifices. It is analogous to how the
socio-technical systems on the world wide web have evolved,
wherein the beginning, norms dictated acceptable or unaccept-
able actions, and these norms were decided by the democratic
majority. As the metaverse ecosystem evolves, it must consider
the rights of minorities and vulnerable communities from the
beginning, because unlike in traditional socio-technical sys-
tems, potential mistreatment would have far more disastrous
consequences, i.e., the victims might feel being mistreated as
if they were in the real world.
XIX. CON CL UD IN G NOTE S
On a final note, technology giants such as Apple and Google
have ambitious plans for materialising the metaverse. With
the engagement of emerging technologies and the progressive
development and refinement of the ecosystem, our virtual
worlds (or digital twins) will look radically different in the
upcoming years. Now, our digitised future is going to be more
interactive, more alive, more embodied and more multimedia,
due to the existence of powerful computing devices and intel-
ligent wearables. However, there exist still many challenges to
be overcome before the metaverse become integrated into the
physical world and our everyday life.
We call for a holistic approach to build the metaverse, as
we consider the metaverse will occur as another enormous
entity in parallel to our physical reality. By surveying the most
recent works across various technologies and ecosystems, we
hope to have provided a wider discussion within the metaverse
community. Through reflecting on the key topics we discussed,
we identify the fundamental challenges and research agenda
to shape the future of metaverse in the next decades(s).
REF ER EN CES
[1] Judy Joshua. Information Bodies: Computational Anxiety in Neal
Stephenson’s Snow Crash. Interdisciplinary Literary Studies, 19(1):17–
47, 2017. Publisher: Penn State University Press.
[2] Anders Bruun and Martin Lynge Stentoft. Lifelogging in the wild:
Participant experiences of using lifelogging as a research tool. In
INTERACT, 2019.
[3] William Burns III. Everything you know about the metaverse is wrong?,
Mar 2018.
[4] Kyle Chayka. Facebook wants us to live in the metaverse, Aug 2021.
[5] Cond´
e Nast. Kevin kelly.
[6] Nvidia omniverse™ platform, Aug 2021.
[7] Paul Milgram, Haruo Takemura, Akira Utsumi, and Fumio Kishino.
Augmented reality: a class of displays on the reality-virtuality continuum.
In Hari Das, editor, Telemanipulator and Telepresence Technologies,
volume 2351, pages 282 – 292. International Society for Optics and
Photonics, SPIE, 1995.
[8] Neda Mohammadi and John Eric Taylor. Smart city digital twins. 2017
IEEE Symposium Series on Computational Intelligence (SSCI), pages 1–
5, 2017.
[9] Michael W. Grieves and J. Vickers. Digital twin: Mitigating unpredictable,
undesirable emergent behavior in complex systems. 2017.
[10] Thomas Bauer, Pablo Oliveira Antonino, and Thomas Kuhn. Towards
architecting digital twin-pervaded systems. In Proceedings of the 7th
International Workshop on Software Engineering for Systems-of-Systems
and 13th Workshop on Distributed Software Development, Software
Ecosystems and Systems-of-Systems, SESoS-WDES ’19, page 66–69.
IEEE Press, 2019.
[11] Abhishek Pokhrel, Vikash Katta, and Ricardo Colomo-Palacios. Digital
twin for cybersecurity incident prediction: A multivocal literature review.
In Proceedings of the IEEE/ACM 42nd International Conference on
Software Engineering Workshops, ICSEW’20, page 671–678, New York,
NY, USA, 2020. Association for Computing Machinery.
[12] `
Eric Pairet, Paola Ard´
on, Xingkun Liu, Jos´
e Lopes, Helen Hastie,
and Katrin S. Lohan. A digital twin for human-robot interaction. In
Proceedings of the 14th ACM/IEEE International Conference on Human-
Robot Interaction, HRI ’19, page 372. IEEE Press, 2019.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 48
[13] P. Cureton and Nick Dunn. Digital twins of cities and evasive futures.
2020.
[14] F. V. Langen. Concept for a virtual learning factory. 2017.
[15] Aaron Bush. Into the void: Where crypto meets the metaverse, Jan 2021.
[16] S. Viljoen. The promise and limits of lawfulness: Inequality, law, and
the techlash. International Political Economy: Globalization eJournal,
2020.
[17] Ying Jiang, Congyi Zhang, Hongbo Fu, Alberto Cannav`
o, Fabrizio
Lamberti, Henry Y K Lau, and Wenping Wang. HandPainter - 3D
Sketching in VR with Hand-Based Physical Proxy. Association for
Computing Machinery, New York, NY, USA, 2021.
[18] Michael Nebeling, Katy Lewis, Yu-Cheng Chang, Lihan Zhu, Michelle
Chung, Piaoyang Wang, and Janet Nebeling. XRDirector: A Role-Based
Collaborative Immersive Authoring System, page 1–12. Association for
Computing Machinery, New York, NY, USA, 2020.
[19] Balasaravanan Thoravi Kumaravel, Cuong Nguyen, Stephen DiVerdi,
and Bj¨
orn Hartmann. TutoriVR: A Video-Based Tutorial System for
Design Applications in Virtual Reality, page 1–12. Association for
Computing Machinery, New York, NY, USA, 2019.
[20] Jens M ¨
uller, Roman R¨
adle, and Harald Reiterer. Virtual Objects as
Spatial Cues in Collaborative Mixed Reality Environments: How They
Shape Communication Behavior and User Task Load, page 1245–1249.
Association for Computing Machinery, New York, NY, USA, 2016.
[21] Richard L. Daft and Robert H. Lengel. Organizational information
requirements, media richness and structural design. Management Science,
32:554–571, 1986.
[22] Haihan Duan, Jiaye Li, Sizheng Fan, Zhonghao Lin, Xiao Wu, and Wei
Cai. Metaverse for social good: A university campus prototype. ACM
Multimedia 2021, abs/2108.08985, 2021.
[23] John Zoshak and Kristin Dew. Beyond Kant and Bentham: How Ethical
Theories Are Being Used in Artificial Moral Agents. Association for
Computing Machinery, New York, NY, USA, 2021.
[24] Semen Frish, Maksym Druchok, and Hlib Shchur. Molecular mr
multiplayer: A cross-platform collaborative interactive game for scientists.
In 26th ACM Symposium on Virtual Reality Software and Technology,
VRST ’20, New York, NY, USA, 2020. Association for Computing
Machinery.
[25] Moreno Marzolla, Stefano Ferretti, and Gabriele D’Angelo. Dynamic
resource provisioning for cloud-based gaming infrastructures. Computers
in Entertainment, 10(1):4:1–4:20, December 2012.
[26] Jeff Terrace, Ewen Cheslack-Postava, Philip Levis, and Michael J.
Freedman. Unsupervised Conversion of 3D Models for Interactive
Metaverses. In 2012 IEEE International Conference on Multimedia and
Expo, pages 902–907, July 2012. ISSN: 1945-788X.
[27] Amit Goel, William A. Rivera, Peter J. Kincaid, Waldemar Karwowski,
Michele M. Montgomery, and Neal Finkelstein. A research framework
for exascale simulations of distributed virtual world environments on high
performance computing (HPC) clusters. In Proceedings of the Symposium
on High Performance Computing, HPC ’15, pages 25–32, San Diego, CA,
USA, April 2015. Society for Computer Simulation International.
[28] Rebecca S. Portnoff, Sadia Afroz, Greg Durrett, Jonathan K. Kummer-
feld, Taylor Berg-Kirkpatrick, Damon McCoy, Kirill Levchenko, and Vern
Paxson. Tools for Automated Analysis of Cybercriminal Markets. In
Proceedings of the 26th International Conference on World Wide Web,
WWW ’17, pages 657–666, Republic and Canton of Geneva, CHE, April
2017. International World Wide Web Conferences Steering Committee.
[29] Christine Webster, Franc¸ois Garnier, and Anne Sedes. Empty Room, an
electroacoustic immersive composition spatialized in virtual 3D space,
in ambisonic and binaural. In Proceedings of the Virtual Reality
International Conference - Laval Virtual 2017, VRIC ’17, pages 1–7, New
York, NY, USA, March 2017. Association for Computing Machinery.
[30] Vlasios Kasapakis and Damianos Gavalas. User-Generated Content in
Pervasive Games. Computers in Entertainment, 16(1):3:1–3:23, Decem-
ber 2017.
[31] Kim Nevelsteen and Martin Wehlou. IPSME- Idempotent Pub-
lish/Subscribe Messaging Environment. In Proceedings of the Interna-
tional Workshop on Immersive Mixed and Virtual Environment Systems
(MMVE ’21), MMVE ’21, pages 30–36, New York, NY, USA, July 2021.
Association for Computing Machinery.
[32] Iain Oliver, Alan Miller, and Colin Allison. Mongoose: Throughput
Redistributing Virtual World. In 2012 21st International Conference
on Computer Communications and Networks (ICCCN), pages 1–9, July
2012. ISSN: 1095-2055.
[33] Mary K. Young, John J. Rieser, and Bobby Bodenheimer. Dyadic
interactions with avatars in immersive virtual environments: high fiving.
In Proceedings of the ACM SIGGRAPH Symposium on Applied Percep-
tion, SAP ’15, pages 119–126, New York, NY, USA, September 2015.
Association for Computing Machinery.
[34] Ariel Vernaza, V. Ivan Armuelles, and Isaac Ruiz. Towards to an
open and interoperable virtual learning enviroment using Metaverse at
University of Panama. In 2012 Technologies Applied to Electronics
Teaching (TAEE), pages 320–325, June 2012.
[35] Yungang Wei, Xiaoran Qin, Xiaoye Tan, Xiaohang Yu, Bo Sun, and
Xiaoming Zhu. The Design of a Visual Tool for the Quick Customization
of Virtual Characters in OSSL. In 2015 International Conference on
Cyberworlds (CW), pages 314–320, October 2015.
[36] Gema Bello Orgaz, Mar´
ıa D. R-Moreno, David Camacho, and David F.
Barrero. Clustering avatars behaviours from virtual worlds interactions.
In Proceedings of the 4th International Workshop on Web Intelligence &
Communities, WI&amp;C ’12, pages 1–7, New York, NY, USA, April
2012. Association for Computing Machinery.
[37] Gema Bello-Orgaz and David Camacho. Comparative study of text
clustering techniques in virtual worlds. In Proceedings of the 3rd
International Conference on Web Intelligence, Mining and Semantics,
WIMS ’13, pages 1–8, New York, NY, USA, June 2013. Association
for Computing Machinery.
[38] Amirreza Barin, Igor Dolgov, and Zachary O. Toups. Understanding
Dangerous Play: A Grounded Theory Analysis of High-Performance
Drone Racing Crashes. In Proceedings of the Annual Symposium on
Computer-Human Interaction in Play, pages 485–496. Association for
Computing Machinery, New York, NY, USA, October 2017.
[39] Suzanne Beer. Virtual Museums: an Innovative Kind of Museum Survey.
In Proceedings of the 2015 Virtual Reality International Conference,
VRIC ’15, pages 1–6, New York, NY, USA, April 2015. Association
for Computing Machinery.
[40] Yungang Wei, Xiaoye Tan, Xiaoran Qin, Xiaohang Yu, Bo Sun, and
Xiaoming Zhu. Exploring the Use of a 3D Virtual Environment in Chinese
Cultural Transmission. In 2014 International Conference on Cyberworlds,
pages 345–350, October 2014.
[41] Hiroyuki Chishiro, Yosuke Tsuchiya, Yoshihide Chubachi, Muham-
mad Saifullah Abu Bakar, and Liyanage C. De Silva. Global PBL for
Environmental IoT. In Proceedings of the 2017 International Conference
on E-commerce, E-Business and E-Government, ICEEG 2017, pages
65–71, New York, NY, USA, June 2017. Association for Computing
Machinery.
[42] Didier Sebastien, Olivier Sebastien, and Noel Conruyt. Providing
services through online immersive real-time mirror-worlds: The Immex
Program for delivering services in another way at university. In Pro-
ceedings of the Virtual Reality International Conference - Laval Virtual,
VRIC ’18, pages 1–7, New York, NY, USA, April 2018. Association for
Computing Machinery.
[43] Frederico M. Schaf, Suenoni Paladini, and Carlos Eduardo Pereira.
3D AutoSysLab prototype. In Proceedings of the 2012 IEEE Global
Engineering Education Conference (EDUCON), pages 1–9, April 2012.
ISSN: 2165-9567.
[44] Liane Tarouco, Barbara Gorziza, Ygor Corrˆ
ea, ´
Erico M. H. Amaral, and
Tha´
ısa M¨
uller. Virtual laboratory for teaching Calculus: An immersive
experience. In 2013 IEEE Global Engineering Education Conference
(EDUCON), pages 774–781, March 2013. ISSN: 2165-9567.
[45] Elif Ayiter. Further Dimensions: Text, Typography and Play in the
Metaverse. In 2012 International Conference on Cyberworlds, pages
296–303, September 2012.
[46] Elif Ayiter. Azimuth to Cypher: The Transformation of a Tiny (Virtual)
Cosmogony. In 2015 International Conference on Cyberworlds (CW),
pages 247–250, October 2015.
[47] Rui Prada, Helmut Prendinger, Panita Yongyuth, Arturo Nakasoneb,
and Asanee Kawtrakulc. AgriVillage: A Game to Foster Awareness of
the Environmental Impact of Agriculture. Computers in Entertainment,
12(2):3:1–3:18, February 2015.
[48] Ben Falchuk, Shoshana Loeb, and Ralph Neff. The Social Metaverse:
Battle for Privacy. IEEE Technology and Society Magazine, 37(2):52–61,
June 2018. Conference Name: IEEE Technology and Society Magazine.
[49] Johanna Ylipulli, Jenny Kangasvuo, Toni Alatalo, and Timo Ojala. Chas-
ing Digital Shadows: Exploring Future Hybrid Cities through Anthropo-
logical Design Fiction. In Proceedings of the 9th Nordic Conference
on Human-Computer Interaction, NordiCHI ’16, pages 1–10, New York,
NY, USA, October 2016. Association for Computing Machinery.
[50] John David N. Dionisio, William G. Burns III, and Richard Gilbert. 3D
Virtual worlds and the metaverse: Current status and future possibilities.
ACM Computing Surveys, 45(3):34:1–34:38, July 2013.
[51] Brendan Kelley and Cyane Tornatzky. The Artistic Approach to
Virtual Reality. In The 17th International Conference on Virtual-Reality

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 49
Continuum and its Applications in Industry, VRCAI ’19, pages 1–5, New
York, NY, USA, November 2019. Association for Computing Machinery.
[52] Diego Martinez Plasencia. One step beyond virtual reality: connecting
past and future developments. XRDS: Crossroads, The ACM Magazine
for Students, 22(1):18–23, November 2015.
[53] Luis A. Ib’ ˜
nez and Viviana Barneche Naya. Cyberarchitecture: A
Vitruvian Approach. In 2012 International Conference on Cyberworlds,
pages 283–289, September 2012.
[54] Richard Skarbez, Missie Smith, and Mary C. Whitton. Revisiting
milgram and kishino’s reality-virtuality continuum. Frontiers in Virtual
Reality, 2:27, 2021.
[55] Maximilian Speicher, Brian D. Hall, and Michael Nebeling. What is
Mixed Reality?, page 1–15. Association for Computing Machinery, New
York, NY, USA, 2019.
[56] I. Sutherland. The ultimate display. 1965.
[57] Minna Pakanen, Paula Alavesa, Niels van Berkel, Timo Koskela, and
Timo Ojala. “nice to see you virtually”: Thoughtful design and evaluation
of virtual avatar of the other user in ar and vr based telexistence systems.
Entertainment Computing, 40:100457, 2022.
[58] Lik-Hang Lee and Pan Hui. Interaction methods for smart glasses: A
survey. IEEE Access, 6:28712–28732, 2018.
[59] Yuta Itoh, Tobias Langlotz, Jonathan Sutton, and Alexander Plopski.
Towards indistinguishable augmented reality: A survey on optical see-
through head-mounted displays. ACM Comput. Surv., 54(6), July 2021.
[60] Jonathan W. Kelly, L. A. Cherep, A. Lim, Taylor A. Doty, and Stephen B
Gilbert. Who are virtual reality headset owners? a survey and comparison
of headset owners and non-owners. 2021 IEEE Virtual Reality and 3D
User Interfaces (VR), pages 687–694, 2021.
[61] S. Singhal and M. Zyda. Networked virtual environments - design and
implementation. 1999.
[62] Huaiyu Liu, Mic Bowman, and Francis Chang. Survey of state melding
in virtual worlds. ACM Comput. Surv., 44(4), September 2012.
[63] Takuji Narumi, Shinya Nishizaka, Takashi Kajinami, Tomohiro
Tanikawa, and Michitaka Hirose. Augmented Reality Flavors: Gustatory
Display Based on Edible Marker and Cross-Modal Interaction, page
93–102. Association for Computing Machinery, New York, NY, USA,
2011.
[64] D. Schmalstieg and T. Hollerer. Augmented Reality - Principles and
Practice. Addison-Wesley Professional, 2016.
[65] J. LaViola et al. 3D User Interfaces: Theory and Practice. Addison
Wesley, 2017.
[66] Steven K. Feiner, Blair MacIntyre, Marcus Haupt, and Eliot Solomon.
Windows on the world: 2d windows for 3d augmented reality. In UIST
’93, 1993.
[67] Jeffrey S. Pierce, Brian C. Stearns, and R. Pausch. Voodoo dolls:
seamless interaction at multiple scales in virtual environments. In SI3D,
1999.
[68] Jeffrey S. Pierce and R. Pausch. Comparing voodoo dolls and homer:
exploring the importance of feedback in virtual environments. Proceed-
ings of the SIGCHI Conference on Human Factors in Computing Systems,
2002.
[69] Lik-Hang Lee, Tristan Braud, S. Hosio, and Pan Hui. Towards aug-
mented reality-driven human-city interaction: Current research and future
challenges. ArXiv, abs/2007.09207, 2020.
[70] Tobias Langlotz, Stefan Mooslechner, Stefanie Zollmann, Claus Degen-
dorfer, Gerhard Reitmayr, and Dieter Schmalstieg. Sketching up the
world: in situ authoring for mobile augmented reality. Personal and
Ubiquitous Computing, 16:623–630, 2011.
[71] Tobias Langlotz, Claus Degendorfer, Alessandro Mulloni, Gerhard
Schall, Gerhard Reitmayr, and Dieter Schmalstieg. Robust detection
and tracking of annotations for outdoor augmented reality browsing.
Computers & Graphics, 35:831 – 840, 2011.
[72] Blair MacIntyre, Enylton Machado Coelho, and Simon J. Julier. Esti-
mating and adapting to registration errors in augmented reality systems.
Proceedings IEEE Virtual Reality 2002, pages 73–80, 2002.
[73] B. MacIntyre and E. M. Coelho. Adapting to dynamic registration
errors using level of error (loe) filtering. Proceedings IEEE and ACM
International Symposium on Augmented Reality (ISAR 2000), pages 85–
88, 2000.
[74] Steven K. Feiner, Blair MacIntyre, Tobias H¨
ollerer, and Anthony Web-
ster. A touring machine: Prototyping 3d mobile augmented reality systems
for exploring the urban environment. In SEMWEB, 1997.
[75] Lik-Hang Lee, Kit-Yung Lam, Yui-Pan Yau, Tristan Braud, and Pan
Hui. Hibey: Hide the keyboard in augmented reality. 2019 IEEE
International Conference on Pervasive Computing and Communications
(PerCom, pages 1–10, 2019.
[76] Philippe Wacker, Adrian Wagner, Simon Voelker, and Jan O. Borchers.
Heatmaps, shadows, bubbles, rays: Comparing mid-air pen position
visualizations in handheld ar. Proceedings of the 2020 CHI Conference
on Human Factors in Computing Systems, 2020.
[77] Chun Xie, Yoshinari Kameda, Kenji Suzuki, and Itaru Kitahara. Large
scale interactive ar display based on a projector-camera system. Proceed-
ings of the 2016 Symposium on Spatial User Interaction, 2016.
[78] Joan Sol Roo, Renaud Gervais, J´
er´
emy Frey, and Martin Hachet.
Inner garden: Connecting inner states to a mixed reality sandbox for
mindfulness. Proceedings of the 2017 CHI Conference on Human Factors
in Computing Systems, 2017.
[79] Jeremy Hartmann, Yen-Ting Yeh, and Daniel Vogel. Aar: Augmenting
a wearable augmented reality display with an actuated head-mounted
projector. In Proceedings of the 33rd Annual ACM Symposium on User
Interface Software and Technology, UIST ’20, page 445–458, New York,
NY, USA, 2020. Association for Computing Machinery.
[80] Isha Chaturvedi, Farshid Hassani Bijarbooneh, Tristan Braud, and Pan
Hui. Peripheral vision: A new killer app for smart glasses. In Proceedings
of the 24th International Conference on Intelligent User Interfaces,
IUI ’19, page 625–636, New York, NY, USA, 2019. Association for
Computing Machinery.
[81] Ting Zhang, Yu-Ting Li, and Juan P. Wachs. The effect of embodied
interaction in visual-spatial navigation. ACM Trans. Interact. Intell. Syst.,
7(1), December 2016.
[82] P. Milgram and F. Kishino. A taxonomy of mixed reality visual displays.
IEICE Transactions on Information and Systems, 77:1321–1329, 1994.
[83] Pedro Lopes, Sijing You, Alexandra Ion, and Patrick Baudisch. Adding
Force Feedback to Mixed Reality Experiences and Games Using Electri-
cal Muscle Stimulation, page 1–13. Association for Computing Machin-
ery, New York, NY, USA, 2018.
[84] Derek Reilly, Andy Echenique, Andy Wu, Anthony Tang, and W. Keith
Edwards. Mapping out Work in a Mixed Reality Project Room, page
887–896. Association for Computing Machinery, New York, NY, USA,
2015.
[85] Masaya Ohta, Shunsuke Nagano, Hotaka Niwa, and Katsumi Yamashita.
[poster] mixed-reality store on the other side of a tablet. In Proceedings
of the 2015 IEEE International Symposium on Mixed and Augmented
Reality, ISMAR ’15, page 192–193, USA, 2015. IEEE Computer Society.
[86] Joan Sol Roo, Renaud Gervais, Jeremy Frey, and Martin Hachet.
Inner Garden: Connecting Inner States to a Mixed Reality Sandbox for
Mindfulness, page 1459–1470. Association for Computing Machinery,
New York, NY, USA, 2017.
[87] Ya-Ting Yue, Yong-Liang Yang, Gang Ren, and Wenping Wang. Scenec-
trl: Mixed reality enhancement via efficient scene editing. In Proceedings
of the 30th Annual ACM Symposium on User Interface Software and
Technology, UIST ’17, page 427–436, New York, NY, USA, 2017.
Association for Computing Machinery.
[88] Laura Malinverni, Julian Maya, Marie-Monique Schaper, and Narcis
Pares. The World-as-Support: Embodied Exploration, Understanding and
Meaning-Making of the Augmented World, page 5132–5144. Association
for Computing Machinery, New York, NY, USA, 2017.
[89] Aaron L Gardony, Robert W Lindeman, and Tad T Bruny´
e. Eye-
tracking for human-centered mixed reality: promises and challenges. In
Optical Architectures for Displays and Sensing in Augmented, Virtual, and
Mixed Reality (AR, VR, MR), volume 11310, page 113100T. International
Society for Optics and Photonics, 2020.
[90] David Lindlbauer and Andy D. Wilson. Remixed Reality: Manipulating
Space and Time in Augmented Reality, page 1–13. Association for
Computing Machinery, New York, NY, USA, 2018.
[91] Cha Lee, Gustavo A. Rincon, Greg Meyer, Tobias H¨
ollerer, and Doug A.
Bowman. The effects of visual realism on search tasks in mixed reality
simulation. IEEE Transactions on Visualization and Computer Graphics,
19(4):547–556, 2013.
[92] Antoine Lassagne, Andras Kemeny, Javier Posselt, and Frederic Meri-
enne. Performance evaluation of passive haptic feedback for tactile hmi
design in caves. IEEE Transactions on Haptics, 11(1):119–127, 2018.
[93] Martijn J.L. Kors, Gabriele Ferri, Erik D. van der Spek, Cas Ketel,
and Ben A.M. Schouten. A breathtaking journey. on the design of an
empathy-arousing mixed-reality game. In Proceedings of the 2016 Annual
Symposium on Computer-Human Interaction in Play, CHI PLAY ’16,
page 91–104, New York, NY, USA, 2016. Association for Computing
Machinery.
[94] Irene V ´
azquez-Mart´
ın, J. Mar´
ın-S´
aez, Marina G´
omez-Climente,
D. Chemisana, M. Collados, and J. Atencia. Full-color multiplexed
reflection hologram of diffusing objects recorded by using simultaneous
exposure with different times in photopolymer bayfol® hx. Optics and
Laser Technology, 143:107303, 2021.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 50
[95] Evan Ackerman. Femtosecond lasers create 3-d midair plasma displays
you can touch, Jun 2021.
[96] Valentin Schwind, Jens Reinhardt, Rufat Rzayev, Niels Henze, and
Katrin Wolf. Virtual reality on the go? a study on social acceptance
of vr glasses. In Proceedings of the 20th International Conference on
Human-Computer Interaction with Mobile Devices and Services Adjunct,
MobileHCI ’18, page 111–118, New York, NY, USA, 2018. Association
for Computing Machinery.
[97] T. Kubota. Creating a more attractive hologram. Leonardo, 25:503 –
506, 1992.
[98] W. Rogers and D. Smalley. Simulating virtual images in optical trap
displays. Scientific Reports, 11, 2021.
[99] Zhi Han Lim and Per Ola Kristensson. An evaluation of discrete and
continuous mid-air loop and marking menu selection in optical see-
through hmds. Proceedings of the 21st International Conference on
Human-Computer Interaction with Mobile Devices and Services, 2019.
[100] Lik-Hang Lee, Tristan Braud, Farshid Hassani Bijarbooneh, and Pan
Hui. Tipoint: detecting fingertip for mid-air interaction on computational
resource constrained smartglasses. Proceedings of the 23rd International
Symposium on Wearable Computers, 2019.
[101] Lik-Hang Lee, Tristan Braud, Farshid Hassani Bijarbooneh, and Pan
Hui. Ubipoint: towards non-intrusive mid-air interaction for hardware
constrained smart glasses. Proceedings of the 11th ACM Multimedia
Systems Conference, 2020.
[102] Aakar Gupta and Ravin Balakrishnan. Dualkey: Miniature screen text
entry via finger identification. Proceedings of the 2016 CHI Conference
on Human Factors in Computing Systems, 2016.
[103] Yizheng Gu, Chun Yu, Zhipeng Li, Weiqi Li, Shuchang Xu, Xiaoying
Wei, and Yuanchun Shi. Accurate and low-latency sensing of touch
contact on any surface with finger-worn imu sensor. Proceedings of
the 32nd Annual ACM Symposium on User Interface Software and
Technology, 2019.
[104] J. Gong, Y. Zhang, X. Zhou, and X. D. Yang. Pyro: Thumb-tip gesture
recognition using pyroelectric infrared sensing proc of the 30th annual
acm symp. on user interface soft. and tech. (uist ’17). pages 553–563,
2017.
[105] Farshid Salemi Parizi, Eric Whitmire, and Shwetak N. Patel. Auraring:
Precise electromagnetic finger tracking. Proc. ACM Interact. Mob.
Wearable Ubiquitous Technol., 3:150:1–150:28, 2019.
[106] Yang Zhang, Wolfgang Kienzle, Yanjun Ma, Shiu S. Ng, Hrvoje Benko,
and Chris Harrison. Actitouch: Robust touch detection for on-skin ar/vr
interfaces. Proceedings of the 32nd Annual ACM Symposium on User
Interface Software and Technology, 2019.
[107] Cheng Zhang, Abdelkareem Bedri, Gabriel Reyes, Bailey Bercik,
Omer T. Inan, Thad Starner, and Gregory D. Abowd. Tapskin: Rec-
ognizing on-skin input for smartwatches. Proceedings of the 2016 ACM
International Conference on Interactive Surfaces and Spaces, 2016.
[108] Taku Hachisu, Baptiste Bourreau, and Kenji Suzuki. Enhancedtouchx:
Smart bracelets for augmenting interpersonal touch interactions. Pro-
ceedings of the 2019 CHI Conference on Human Factors in Computing
Systems, 2019.
[109] Kenji Suzuki, Taku Hachisu, and Kazuki Iida. Enhancedtouch: A smart
bracelet for enhancing human-human physical touch. Proceedings of the
2016 CHI Conference on Human Factors in Computing Systems, 2016.
[110] Pui Chung Wong, Kening Zhu, and Hongbo Fu. Fingert9: Leverag-
ing thumb-to-finger interaction for same-side-hand text entry on smart-
watches. Proceedings of the 2018 CHI Conference on Human Factors in
Computing Systems, 2018.
[111] Mohamed Soliman, Franziska Mueller, Lena Hegemann, Joan Sol Roo,
Christian Theobalt, and J¨
urgen Steimle. Fingerinput: Capturing expressive
single-hand thumb-to-finger microgestures. Proceedings of the 2018 ACM
International Conference on Interactive Surfaces and Spaces, 2018.
[112] Da-Yuan Huang, Liwei Chan, Shuo Yang, Fan Wang, Rong-Hao
Liang, De-Nian Yang, Yi-Ping Hung, and Bing-Yu Chen. Digitspace:
Designing thumb-to-fingers touch interfaces for one-handed and eyes-
free interactions. Proceedings of the 2016 CHI Conference on Human
Factors in Computing Systems, 2016.
[113] Zheer Xu, Pui Chung Wong, Jun Gong, Te-Yen Wu, Aditya Shekhar
Nittala, Xiaojun Bi, J¨
urgen Steimle, Hongbo Fu, Kening Zhu, and Xing-
Dong Yang. Tiptext: Eyes-free text entry on a fingertip keyboard.
Proceedings of the 32nd Annual ACM Symposium on User Interface
Software and Technology, 2019.
[114] David Dobbelstein, Christian Winkler, Gabriel Haas, and Enrico
Rukzio. Pocketthumb: a wearable dual-sided touch interface for cursor-
based control of smart-eyewear. Proc. ACM Interact. Mob. Wearable
Ubiquitous Technol., 1:9:1–9:17, 2017.
[115] Konstantin Klamka and Raimund Dachselt. Arcord: Visually aug-
mented interactive cords for mobile interaction. Extended Abstracts of the
2018 CHI Conference on Human Factors in Computing Systems, 2018.
[116] Ivan Poupyrev, Nan-Wei Gong, Shiho Fukuhara, Mustafa Emre Karago-
zler, Carsten Schwesig, and Karen E. Robinson. Project jacquard: Inter-
active digital textiles at scale. Proceedings of the 2016 CHI Conference
on Human Factors in Computing Systems, 2016.
[117] Kirill A. Shatilov, Dimitris Chatzopoulos, Lik-Hang Lee, and Pan Hui.
Emerging exg-based nui inputs in extended realities: A bottom-up survey.
ACM Trans. Interact. Intell. Syst., 11(2), July 2021.
[118] Young D. Kwon, Kirill A. Shatilov, Lik-Hang Lee, Serkan Kumyol,
Kit-Yung Lam, Yui-Pan Yau, and Pan Hui. Myokey: Surface electromyo-
graphy and inertial motion sensing-based text entry in ar. In 2020 IEEE
International Conference on Pervasive Computing and Communications
Workshops (PerCom Workshops), pages 1–4, 2020.
[119] Emily Dao, Andreea Muresan, Kasper Hornbæk, and Jarrod Knibbe.
Bad Breakdowns, Useful Seams, and Face Slapping: Analysis of VR Fails
on YouTube. Association for Computing Machinery, New York, NY, USA,
2021.
[120] Kevin Arthur. Effects of field of view on task performance with
head-mounted displays. Conference Companion on Human Factors in
Computing Systems, 1996.
[121] Long Qian, Alexander Plopski, Nassir Navab, and Peter Kazanzides.
Restoring the awareness in the occluded visual field for optical see-
through head-mounted displays. IEEE Transactions on Visualization and
Computer Graphics, 24:2936–2946, 2018.
[122] Andrew Lingley, Muhammad Umair Ali, Y. Liao, Ramin Mirjalili,
Maria Klonner, M. Sopanen, Sami Suihkonen, Tueng Shen, Brian P. Otis,
H. Lipsanen, and Babak A. Parviz. A single-pixel wireless contact lens
display. Journal of Micromechanics and Microengineering, 21:125014,
2011.
[123] Kit Yung Lam, Lik Hang Lee, Tristan Braud, and Pan Hui. M2a:
A framework for visualizing information from mobile web to mobile
augmented reality. In 2019 IEEE International Conference on Pervasive
Computing and Communications (PerCom, pages 1–10, 2019.
[124] Kit Yung Lam, Lik-Hang Lee, and Pan Hui. A2w: Context-aware
recommendation system for mobile augmented reality web browser. In
ACM International Conference on Multimedia, United States, October
2021. Association for Computing Machinery (ACM).
[125] Alexander Marquardt, Christina Trepkowski, Tom David Eibich, Jens
Maiero, and Ernst Kruijff. Non-visual cues for view management in
narrow field of view augmented reality displays. 2019 IEEE International
Symposium on Mixed and Augmented Reality (ISMAR), pages 190–201,
2019.
[126] Tsontcho Sean Ianchulev, Don S. Minckler, H. Dunbar Hoskins, Mark
Packer, Robert L. Stamper, Ravinder D. Pamnani, and Edward Koo.
Wearable technology with head-mounted displays and visual function.
JAMA, 312 17:1799–801, 2014.
[127] Yu-Chih Lin, Leon Hsu, and Mike Y. Chen. Peritextar: utilizing
peripheral vision for reading text on augmented reality smart glasses.
Proceedings of the 24th ACM Symposium on Virtual Reality Software
and Technology, 2018.
[128] Lik-Hang Lee, Tristan Braud, Kit-Yung Lam, Yui-Pan Yau, and Pan
Hui. From seen to unseen: Designing keyboard-less interfaces for text
entry on the constrained screen real estate of augmented reality headsets.
Pervasive Mob. Comput., 64:101148, 2020.
[129] Mingqian Zhao, Huamin Qu, and Michael Sedlmair. Neighborhood
perception in bar charts. Proceedings of the 2019 CHI Conference on
Human Factors in Computing Systems, 2019.
[130] Michele Gattullo, Antonio E. Uva, Michele Fiorentino, and Joseph L.
Gabbard. Legibility in industrial ar: Text style, color coding, and
illuminance. IEEE Computer Graphics and Applications, 35:52–61, 2015.
[131] Daniel Boyarski, Christine Neuwirth, Jodi Forlizzi, and Susan Harkness
Regli. A study of fonts designed for screen display. Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems, 1998.
[132] Alexis D. Souchet, St´
ephanie Philippe, Floriane Ober, Aur´
elien
L´
eveque, and Laure Leroy. Investigating cyclical stereoscopy effects over
visual discomfort and fatigue in virtual reality while learning. 2019 IEEE
International Symposium on Mixed and Augmented Reality (ISMAR),
pages 328–338, 2019.
[133] Yuki Matsuura, Tsutomu Terada, Tomohiro Aoki, Susumu Sonoda,
Naoya Isoyama, and Masahiko Tsukamoto. Readability and legibility
of fonts considering shakiness of head mounted displays. Proceedings of
the 23rd International Symposium on Wearable Computers, 2019.
[134] Masayuki Nakao, Tsutomu Terada, and Masahiko Tsukamoto. An
information presentation method for head mounted display considering

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 51
surrounding environments. Proceedings of the 5th Augmented Human
International Conference, 2014.
[135] Kohei Tanaka, Y. Kishino, Masakazu Miyamae, T. Terada, and
S. Nishio. An information layout method for an optical see-through
head mounted display focusing on the viewability. 2008 7th IEEE/ACM
International Symposium on Mixed and Augmented Reality, pages 139–
142, 2008.
[136] Mitchell L Gordon and Shumin Zhai. Touchscreen haptic augmentation
effects on tapping, drag and drop, and path following. In Proceedings
of the 2019 CHI Conference on Human Factors in Computing Systems,
pages 1–12, 2019.
[137] C Doerrer and R Werthschuetzky. Simulating push-buttons using a
haptic display: Requirements on force resolution and force-displacement
curve. In Proc. EuroHaptics, pages 41–46, 2002.
[138] Carlos Bermejo, Lik Hang Lee, Paul Chojecki, David Przewozny, and
Pan Hui. Exploring button designs for mid-air interaction in virtual
reality: A hexa-metric evaluation of key representations and multi-modal
cues. Proc. ACM Hum.-Comput. Interact., 5(EICS), May 2021.
[139] Jindˇ
rich Adolf, Peter K´
an, Benjamin Outram, Hannes Kaufmann,
Jarom´
ır Doleˇ
zal, and Lenka Lhotsk´
a. Juggling in vr: Advantages of
immersive virtual reality in juggling learning. In 25th ACM Symposium
on Virtual Reality Software and Technology, VRST ’19, New York, NY,
USA, 2019. Association for Computing Machinery.
[140] Adam Faeth and Chris Harding. Emergent effects in multimodal
feedback from virtual buttons. ACM Transactions on Computer-Human
Interaction (TOCHI), 21(1):3, 2014.
[141] Eve Hoggan, Stephen A. Brewster, and Jody Johnston. Investigating the
effectiveness of tactile feedback for mobile touchscreens. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems,
CHI ’08, page 1573–1582, New York, NY, USA, 2008. Association for
Computing Machinery.
[142] Jennifer L Tennison and Jenna L Gorlewicz. Non-visual perception of
lines on a multimodal touchscreen tablet. ACM Transactions on Applied
Perception (TAP), 16(1):1–19, 2019.
[143] Anatole L´
ecuyer, J-M Burkhardt, Sabine Coquillart, and Philippe
Coiffet. ” boundary of illusion”: an experiment of sensory integration
with a pseudo-haptic system. In Proceedings IEEE Virtual Reality 2001,
pages 115–122. IEEE, 2001.
[144] Patrick L. Strandholt, Oana A. Dogaru, Niels C. Nilsson, Rolf Nordahl,
and Stefania Serafin. Knock on wood: Combining redirected touching and
physical props for tool-based interaction in virtual reality. In Proceedings
of the 2020 CHI Conference on Human Factors in Computing Systems,
CHI ’20, page 1–13, New York, NY, USA, 2020. Association for
Computing Machinery.
[145] Evan Pezent, Ali Israr, Majed Samad, Shea Robinson, Priyanshu
Agarwal, Hrvoje Benko, and Nick Colonnese. Tasbi: Multisensory
squeeze and vibrotactile wrist haptics for augmented and virtual reality.
In 2019 IEEE World Haptics Conference (WHC), pages 1–6. IEEE, 2019.
[146] Majed Samad, Elia Gatti, Anne Hermes, Hrvoje Benko, and Cesare
Parise. Pseudo-haptic weight: Changing the perceived weight of virtual
objects by manipulating control-display ratio. In Proceedings of the 2019
CHI Conference on Human Factors in Computing Systems, CHI ’19, page
1–13, New York, NY, USA, 2019. Association for Computing Machinery.
[147] Marco Speicher, Jan Ehrlich, Vito Gentile, Donald Degraen, Salvatore
Sorce, and Antonio Kr¨
uger. Pseudo-haptic controls for mid-air finger-
based menu interaction. In Extended Abstracts of the 2019 CHI Confer-
ence on Human Factors in Computing Systems, pages 1–6, 2019.
[148] Zhaoyuan Ma, Darren Edge, Leah Findlater, and Hong Z. Tan. Haptic
keyclick feedback improves typing speed and reduces typing errors on a
flat keyboard. IEEE World Haptics Conference, WHC 2015, pages 220–
227, 2015.
[149] Mourad Bouzit, Grigore Burdea, George Popescu, and Rares Boian.
The rutgers master ii-new design force-feedback glove. IEEE/ASME
Transactions on mechatronics, 7(2):256–263, 2002.
[150] Y Nam, M Park, and R Yamane. Smart glove: hand master using
magnetorheological fluid actuators. In Proc. SPIE, volume 6794, pages
679434–1, 2007.
[151] HyunKi In, Kyu-Jin Cho, KyuRi Kim, and BumSuk Lee. Jointless
structure and under-actuation mechanism for compact hand exoskeleton.
In Rehabilitation Robotics (ICORR), 2011 IEEE International Conference
on, pages 1–6. IEEE, 2011.
[152] CyberGlove Systems. CyberTouch, 2020. http://www.
cyberglovesystems.com/cybertouch.
[153] Massimiliano Gabardi, Massimiliano Solazzi, Daniele Leonardis, and
Antonio Frisoli. A new wearable fingertip haptic interface for the ren-
dering of virtual shapes and surface features. IEEE Haptics Symposium,
HAPTICS, 2016-April:140–146, 2016.
[154] Hwan Kim, Minhwan Kim, and Woohun Lee. HapThimble : A
Wearable Haptic Device towards Usable Virtual Touch Screen. Chi ’16,
pages 3694–3705, 2016.
[155] Jay Henderson, Jeff Avery, Laurent Grisoni, and Edward Lank. Lever-
aging distal vibrotactile feedback for target acquisition. In Proceedings
of the 2019 CHI Conference on Human Factors in Computing Systems,
pages 1–11, 2019.
[156] Rajinder Sodhi, Ivan Poupyrev, Matthew Glisson, and Ali Israr. Aireal:
interactive tactile experiences in free air. ACM Transactions on Graphics
(TOG), 32(4):134, 2013.
[157] Tom Carter, Sue Ann Seah, Benjamin Long, Bruce Drinkwater, and
Sriram Subramanian. UltraHaptics : Multi-Point Mid-Air Haptic Feed-
back for Touch Surfaces. 2013.
[158] Faisal Arafsha, Longyu Zhang, Haiwei Dong, and Abdulmotaleb El
Saddik. Contactless haptic feedback: State of the art. 2015 IEEE
International Symposium on Haptic, Audio and Visual Environments and
Games, HAVE 2015 - Proceedings, 2015.
[159] C´
edric Kervegant, F´
elix Raymond, Delphine Graeff, and Julien Castet.
Touch hologram in mid-air. In ACM SIGGRAPH 2017 Emerging
Technologies, page 23. ACM, 2017.
[160] Hojin Lee, Hojun Cha, Junsuk Park, Seungmoon Choi, Hyung-Sik
Kim, and Soon-Cheol Chung. LaserStroke. Proceedings of the 29th
Annual Symposium on User Interface Software and Technology - UIST
’16 Adjunct, pages 73–74, 2016.
[161] Yoichi Ochiai, Kota Kumagai, Takayuki Hoshi, Satoshi Hasegawa,
and Yoshio Hayasaki. Cross-Field Aerial Haptics : Rendering Haptic
Feedback in Air with Light and Acoustic Fields. Chi ’16, pages 3238–
3247, 2016.
[162] Claudio Pacchierotti, Stephen Sinclair, Massimiliano Solazzi, Antonio
Frisoli, Vincent Hayward, and Domenico Prattichizzo. Wearable haptic
systems for the fingertip and the hand: taxonomy, review, and perspec-
tives. IEEE transactions on haptics, 10(4):580–600, 2017.
[163] Ju-Hwan Lee and Charles Spence. Assessing the benefits of multimodal
feedback on dual-task performance under demanding conditions. In
Proceedings of the 22nd British HCI Group Annual Conference on People
and Computers: Culture, Creativity, Interaction-Volume 1, pages 185–
192. British Computer Society, 2008.
[164] Akemi Kobayashi, Ryosuke Aoki, Norimichi Kitagawa, Toshitaka
Kimura, Youichi Takashima, and Tomohiro Yamada. Towards enhancing
force-input interaction by visual-auditory feedback as an introduction of
first use. In International Conference on Human-Computer Interaction,
pages 180–191. Springer, 2016.
[165] Andy Cockburn and Stephen Brewster. Multimodal feedback for the
acquisition of small targets. Ergonomics, 48(9):1129–1150, 2005.
[166] Nikolaos Kaklanis, Juan Gonz´
alez Calleros, Jean Vanderdonckt, and
Dimitrios Tzovaras. A haptic rendering engine of web pages for blind
users. In Proceedings of the working conference on Advanced visual
interfaces, pages 437–440, 2008.
[167] Minglu Zhu, Zhongda Sun, Zixuan Zhang, Qiongfeng Shi, Tianyiyi
He, Huicong Liu, Tao Chen, and Chengkuo Lee. Haptic-feedback smart
glove as a creative human-machine interface (hmi) for virtual/augmented
reality applications. Science Advances, 6(19):eaaz8693, 2020.
[168] Thomas Hulin, Alin Albu-Schaffer, and Gerd Hirzinger. Passivity
and stability boundaries for haptic systems with time delay. IEEE
Transactions on Control Systems Technology, 22(4):1297–1309, 2014.
[169] Yongseok Lee, Inyoung Jang, and Dongjun Lee. Enlarging just
noticeable differences of visual-proprioceptive conflict in VR using haptic
feedback. IEEE World Haptics Conference, WHC 2015, pages 19–24,
2015.
[170] Asad Tirmizi, Claudio Pacchierotti, Irfan Hussain, Gianluca Alberico,
and Domenico Prattichizzo. A perceptually-motivated deadband compres-
sion approach for cutaneous haptic feedback. IEEE Haptics Symposium,
HAPTICS, 2016-April:223–228, 2016.
[171] G¨
unter Alce, Maximilian Roszko, Henrik Edlund, Sandra Olsson,
Johan Svedberg, and Mattias Wallerg˚
ard. [poster] ar as a user interface
for the internet of things—comparing three interaction models. In 2017
IEEE International Symposium on Mixed and Augmented Reality (ISMAR-
Adjunct), pages 81–86. IEEE, 2017.
[172] Yushan Siriwardhana, Pawani Porambage, Madhusanka Liyanage, and
Mika Ylianttila. A survey on mobile augmented reality with 5g mobile
edge computing: Architectures, applications, and technical aspects. IEEE
Communications Surveys & Tutorials, 23(2):1160–1192, 2021.
[173] Gerhard Fettweis and Siavash Alamouti. 5G: Personal mobile internet
beyond what cellular did to telephony. IEEE Communications Magazine,
52(2):140–145, 2014.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 52
[174] Martin Maier, Mahfuzulhoq Chowdhury, Bhaskar Prasad Rimal, and
Dung Pham Van. The tactile internet: vision, recent progress, and open
challenges. IEEE Communications Magazine, 54(5):138–145, 2016.
[175] Adnan Aijaz, Mischa Dohler, A. Hamid Aghvami, Vasilis Friderikos,
and Magnus Frodigh. Realizing the Tactile Internet: Haptic Commu-
nications over Next Generation 5G Cellular Networks. IEEE Wireless
Communications, pages 1–8, 2016.
[176] M Simsek, A Aijaz, M Dohler, J Sachs, and G Fettweis. 5G-Enabled
Tactile Internet. IEEE Journal on Selected Areas in Communications,
34(3):460–473, 2016.
[177] Jens Pilz, Matthias Mehlhose, Thomas Wirth, Dennis Wieruch, Bernd
Holfeld, and Thomas Haustein. A Tactile Internet demonstration: 1ms
ultra low delay for wireless communications towards 5G. Proceedings -
IEEE INFOCOM, 2016-Septe(Keystone I):862–863, 2016.
[178] Eckehard Steinbach, Sandra Hirche, Marc Ernst, Fernanda Brandi,
Rahul Chaudhari, Julius Kammerl, and Iason Vittorias. Haptic commu-
nications. Proceedings of the IEEE, 100(4):937–956, 2012.
[179] Jeroen G W Wildenbeest, David A. Abbink, Cock J M Heemskerk,
Frans C T Van Der Helm, and Henri Boessenkool. The impact of haptic
feedback quality on the performance of teleoperated assembly tasks. IEEE
Transactions on Haptics, 6(2):242–252, 2013.
[180] Christoph Bachhuber and Eckehard Steinbach. Are today’s video
communication solutions ready for the tactile internet? In 2017 IEEE
Wireless Communications and Networking Conference Workshops (WC-
NCW), pages 1–6. IEEE, 2017.
[181] Lionel Sujay Vailshery. Internet of things (iot) and non-iot active device
connections worldwide from 2010 to 2025, March 2021.
[182] Joo Chan Kim, Teemu H Laine, and Christer ˚
Ahlund. Multimodal
interaction systems based on internet of things and augmented reality: A
systematic literature review. Applied Sciences, 11(4):1738, 2021.
[183] Dongsik Jo and Gerard Jounghyun Kim. Ariot: scalable augmented
reality framework for interacting with internet of things appliances
everywhere. IEEE Transactions on Consumer Electronics, 62(3):334–
340, 2016.
[184] Vincent Becker, Felix Rauchenstein, and G´
abor S¨
or¨
os. Connecting and
controlling appliances through wearable augmented reality. 2020.
[185] Stephanie Arevalo Arboleda, Franziska R¨
ucker, Tim Dierks, and Jens
Gerken. Assisting manipulation and grasping in robot teleoperation
with augmented reality visual cues. In Proceedings of the 2021 CHI
Conference on Human Factors in Computing Systems, CHI ’21, New
York, NY, USA, 2021. Association for Computing Machinery.
[186] Yongtae Park, Sangki Yun, and Kyu-Han Kim. When iot met aug-
mented reality: Visualizing the source of the wireless signal in ar view.
In Proceedings of the 17th Annual International Conference on Mobile
Systems, Applications, and Services, MobiSys ’19, page 117–129, New
York, NY, USA, 2019. Association for Computing Machinery.
[187] Carlos Bermejo Fernandez, Lik-Hang Lee, Petteri Nurmi, and Pan Hui.
Para: Privacy management and control in emerging iot ecosystems using
augmented reality. In ACM International Conference on Multimodal
Interaction, United States, 2021. Association for Computing Machinery
(ACM).
[188] Yuanzhi Cao, Zhuangying Xu, Fan Li, Wentao Zhong, Ke Huo, and
Karthik Ramani. V. ra: An in-situ visual authoring system for robot-iot
task planning with augmented reality. In Proceedings of the 2019 on
Designing Interactive Systems Conference, pages 1059–1070, 2019.
[189] Mehrnaz Sabet, Mania Orand, and David W. McDonald. Designing
telepresence drones to support synchronous, mid-air remote collaboration:
An exploratory study. In Proceedings of the 2021 CHI Conference on
Human Factors in Computing Systems, CHI ’21, New York, NY, USA,
2021. Association for Computing Machinery.
[190] Linfeng Chen, Akiyuki Ebi, Kazuki Takashima, Kazuyuki Fujita, and
Yoshifumi Kitamura. Pinpointfly: An egocentric position-pointing drone
interface using mobile ar. In SIGGRAPH Asia 2019 Emerging Technolo-
gies, SA ’19, page 34–35, New York, NY, USA, 2019. Association for
Computing Machinery.
[191] Evgeny Tsykunov, Roman Ibrahimov, Derek Vasquez, and Dzmitry
Tsetserukou. Slingdrone: Mixed reality system for pointing and in-
teraction using a single drone. In 25th ACM Symposium on Virtual
Reality Software and Technology, VRST ’19, New York, NY, USA, 2019.
Association for Computing Machinery.
[192] Andreas Riegler, Philipp Wintersberger, Andreas Riener, and Clemens
Holzmann. Augmented reality windshield displays and their potential
to enhance user experience in automated driving. i-com, 18(2):127–149,
2019.
[193] Andreas Riegler, Andreas Riener, and Clemens Holzmann. A system-
atic review of virtual reality applications for automated driving: 2009–
2020. Frontiers in Human Dynamics, page 48, 2021.
[194] Gesa Wiegand, Christian Mai, Kai Holl¨
ander, and Heinrich Hussmann.
Incarar: A design space towards 3d augmented reality applications in ve-
hicles. In Proceedings of the 11th International Conference on Automotive
User Interfaces and Interactive Vehicular Applications, AutomotiveUI
’19, page 1–13, New York, NY, USA, 2019. Association for Computing
Machinery.
[195] Mark Colley, Surong Li, and Enrico Rukzio. Increasing pedestrian
safety using external communication of autonomous vehicles for sig-
nalling hazards. In Proceedings of the 23rd International Conference on
Mobile Human-Computer Interaction, MobileHCI ’21, New York, NY,
USA, 2021. Association for Computing Machinery.
[196] Mark Colley, Svenja Krauss, Mirjam Lanzer, and Enrico Rukzio. How
should automated vehicles communicate critical situations? a comparative
analysis of visualization concepts. Proc. ACM Interact. Mob. Wearable
Ubiquitous Technol., 5(3), September 2021.
[197] Kai Holl ¨
ander, Mark Colley, Enrico Rukzio, and Andreas Butz. A
taxonomy of vulnerable road users for hci based on a systematic literature
review. In Proceedings of the 2021 CHI Conference on Human Factors
in Computing Systems, CHI ’21, New York, NY, USA, 2021. Association
for Computing Machinery.
[198] Pengyuan Zhou, Pranvera Kortoc¸i, Yui-Pan Yau, Tristan Braud, Xiujun
Wang, Benjamin Finley, Lik-Hang Lee, Sasu Tarkoma, Jussi Kangasharju,
and Pan Hui. Augmented informative cooperative perception. ArXiv,
abs/2101.05508, 2021.
[199] Sonia Mary Chacko and Vikram Kapila. Augmented reality as a
medium for human-robot collaborative tasks. In 2019 28th IEEE In-
ternational Conference on Robot and Human Interactive Communication
(RO-MAN), pages 1–8, 2019.
[200] Morteza Dianatfar, Jyrki Latokartano, and Minna Lanz. Review on
existing vr/ar solutions in human–robot collaboration. Procedia CIRP,
97:407–411, 2021. 8th CIRP Conference of Assembly Technology and
Systems.
[201] Joseph La Delfa, Mehmet Aydin Baytas¸, Emma Luke, Ben Koder, and
Florian ’Floyd’ Mueller. Designing drone chi: Unpacking the thinking
and making of somaesthetic human-drone interaction. In Proceedings
of the 2020 ACM Designing Interactive Systems Conference, DIS ’20,
page 575–586, New York, NY, USA, 2020. Association for Computing
Machinery.
[202] Joseph La Delfa, Mehmet Aydin Baytas, Rakesh Patibanda, Hazel
Ngari, Rohit Ashok Khot, and Florian ’Floyd’ Mueller. Drone chi:
Somaesthetic human-drone interaction. In Proceedings of the 2020 CHI
Conference on Human Factors in Computing Systems, CHI ’20, page
1–13, New York, NY, USA, 2020. Association for Computing Machinery.
[203] Jared A. Frank, Matthew Moorhead, and Vikram Kapila. Mobile
mixed-reality interfaces that enhance human–robot interaction in shared
spaces. Frontiers in Robotics and AI, 4:20, 2017.
[204] Antonia Meissner, Angelika Tr¨
ubswetter, Antonia S. Conti-Kufner,
and Jonas Schmidtler. Friend or foe? understanding assembly workers’
acceptance of human-robot collaboration. J. Hum.-Robot Interact., 10(1),
July 2020.
[205] Sean A. McGlynn and Wendy A. Rogers. Provisions of human-robot
friendship. In Proceedings of the Tenth Annual ACM/IEEE International
Conference on Human-Robot Interaction Extended Abstracts, HRI’15 Ex-
tended Abstracts, page 115–116, New York, NY, USA, 2015. Association
for Computing Machinery.
[206] Hyun Young Kim, Bomyeong Kim, and Jinwoo Kim. The naughty
drone: A qualitative research on drone as companion device. In Pro-
ceedings of the 10th International Conference on Ubiquitous Information
Management and Communication, IMCOM ’16, New York, NY, USA,
2016. Association for Computing Machinery.
[207] Haodan Tan, Jangwon Lee, and Gege Gao. Human-drone interaction:
Drone delivery & services for social events. In Proceedings of the
2018 ACM Conference Companion Publication on Designing Interactive
Systems, DIS ’18 Companion, page 183–187, New York, NY, USA, 2018.
Association for Computing Machinery.
[208] Ho Seok Ahn, JongSuk Choi, Hyungpil Moon, and Yoonseob Lim.
Social human-robot interaction of human-care service robots. In Com-
panion of the 2018 ACM/IEEE International Conference on Human-
Robot Interaction, HRI ’18, page 385–386, New York, NY, USA, 2018.
Association for Computing Machinery.
[209] Bethany Ann Mackey, Paul A. Bremner, and Manuel Giuliani. Im-
mersive control of a robot surrogate for users in palliative care. In
Companion of the 2020 ACM/IEEE International Conference on Human-
Robot Interaction, HRI ’20, page 585–587, New York, NY, USA, 2020.
Association for Computing Machinery.
[210] Viviane Herdel, Lee J. Yamin, Eyal Ginosar, and Jessica R. Cauchard.
Public drone: Attitude towards drone capabilities in various contexts. In

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 53
Proceedings of the 23rd International Conference on Mobile Human-
Computer Interaction, MobileHCI ’21, New York, NY, USA, 2021.
Association for Computing Machinery.
[211] Eduard Fosch-Villaronga and Adam Poulsen. Sex robots in care:
Setting the stage for a discussion on the potential use of sexual robot
technologies for persons with disabilities. In Companion of the 2021
ACM/IEEE International Conference on Human-Robot Interaction, HRI
’21 Companion, page 1–9, New York, NY, USA, 2021. Association for
Computing Machinery.
[212] Nina J. Rothstein, Dalton H. Connolly, Ewart J. de Visser, and Elizabeth
Phillips. Perceptions of infidelity with sex robots. In Proceedings of the
2021 ACM/IEEE International Conference on Human-Robot Interaction,
HRI ’21, page 129–139, New York, NY, USA, 2021. Association for
Computing Machinery.
[213] Giovanni Maria Troiano, Matthew Wood, and Casper Harteveld. ”and
this, kids, is how i met your mother”: Consumerist, mundane, and
uncanny futures with sex robots. In Proceedings of the 2020 CHI
Conference on Human Factors in Computing Systems, CHI ’20, page
1–17, New York, NY, USA, 2020. Association for Computing Machinery.
[214] Anna Zamansky. Dog-drone interactions: Towards an aci perspec-
tive. In Proceedings of the Third International Conference on Animal-
Computer Interaction, ACI ’16, New York, NY, USA, 2016. Association
for Computing Machinery.
[215] Jessica R. Cauchard, Jane L. E, Kevin Y. Zhai, and James A. Landay.
Drone & me: An exploration into natural human-drone interaction.
In Proceedings of the 2015 ACM International Joint Conference on
Pervasive and Ubiquitous Computing, UbiComp ’15, page 361–365, New
York, NY, USA, 2015. Association for Computing Machinery.
[216] Binh Vinh Duc Nguyen, Adalberto L. Simeone, and Andrew
Vande Moere. Exploring an architectural framework for human-building
interaction via a semi-immersive cross-reality methodology. In Pro-
ceedings of the 2021 ACM/IEEE International Conference on Human-
Robot Interaction, HRI ’21, page 252–261, New York, NY, USA, 2021.
Association for Computing Machinery.
[217] John McCarthy. What is artificial intelligence? 1998.
[218] Stuart Russell and Peter Norvig. Artificial intelligence: a modern
approach. 2002.
[219] Stephanie Dick. Artificial intelligence. 2019.
[220] Yoshua Bengio, R´
ejean Ducharme, Pascal Vincent, and Christian
Jauvin. A neural probabilistic language model. Journal of machine
learning research, 3(Feb):1137–1155, 2003.
[221] Ronan Collobert and Jason Weston. A unified architecture for natural
language processing: Deep neural networks with multitask learning. In
Proceedings of the 25th international conference on Machine learning,
pages 160–167. ACM, 2008.
[222] Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian
deep learning for computer vision? In Advances in neural information
processing systems, pages 5574–5584, 2017.
[223] Hassan Abu Alhaija, Siva Karthik Mustikovela, Lars Mescheder, An-
dreas Geiger, and Carsten Rother. Augmented reality meets deep learning
for car instance segmentation in urban scenes. In British machine vision
conference, volume 1, page 2, 2017.
[224] Shuai Zhang, Lina Yao, Aixin Sun, and Yi Tay. Deep learning based
recommender system: A survey and new perspectives. ACM Computing
Surveys (CSUR), 52(1):1–38, 2019.
[225] Jie Lu, Dianshuang Wu, Mingsong Mao, Wei Wang, and Guangquan
Zhang. Recommender system application developments: a survey. Deci-
sion Support Systems, 74:12–32, 2015.
[226] Douglas C Montgomery, Elizabeth A Peck, and G Geoffrey Vining.
Introduction to linear regression analysis. John Wiley & Sons, 2021.
[227] Thais Mayumi Oshiro, Pedro Santoro Perez, and Jos´
e Augusto
Baranauskas. How many trees in a random forest? In International
workshop on machine learning and data mining in pattern recognition,
pages 154–168. Springer, 2012.
[228] Anthony J Myles, Robert N Feudale, Yang Liu, Nathaniel A Woody,
and Steven D Brown. An introduction to decision tree modeling. Journal
of Chemometrics: A Journal of the Chemometrics Society, 18(6):275–285,
2004.
[229] Greg Hamerly and Charles Elkan. Learning the k in k-means. Advances
in neural information processing systems, 16:281–288, 2004.
[230] Svante Wold, Kim Esbensen, and Paul Geladi. Principal component
analysis. Chemometrics and intelligent laboratory systems, 2(1-3):37–52,
1987.
[231] Christopher C Paige and Michael A Saunders. Towards a generalized
singular value decomposition. SIAM Journal on Numerical Analysis,
18(3):398–405, 1981.
[232] Christopher JCH Watkins and Peter Dayan. Q-learning. Machine
learning, 8(3-4):279–292, 1992.
[233] Nathan Sprague and Dana Ballard. Multiple-goal reinforcement learn-
ing with modular sarsa (0). 2003.
[234] David Silver, Guy Lever, Nicolas Heess, Thomas Degris, Daan Wier-
stra, and Martin Riedmiller. Deterministic policy gradient algorithms. In
International conference on machine learning, pages 387–395. PMLR,
2014.
[235] Keiron O’Shea and Ryan Nash. An introduction to convolutional neural
networks. arXiv preprint arXiv:1511.08458, 2015.
[236] Wojciech Zaremba, Ilya Sutskever, and Oriol Vinyals. Recurrent neural
network regularization. arXiv preprint arXiv:1409.2329, 2014.
[237] Aidan Fuller, Zhong Fan, Charles Day, and Chris Barlow. Digital
twin: Enabling technologies, challenges and open research. IEEE access,
8:108952–108971, 2020.
[238] Mohamed Habib Farhat, Xavier Chiementin, Fakher Chaari, Fabrice
Bolaers, and Mohamed Haddar. Digital twin-driven machine learning:
ball bearings fault severity classification. Measurement Science and
Technology, 32(4):044006, 2021.
[239] Giulio Paolo Agnusdei, Valerio Elia, and Maria Grazia Gnoni. A
classification proposal of digital twin applications in the safety domain.
Computers & Industrial Engineering, page 107137, 2021.
[240] Farzin Piltan and Jong-Myon Kim. Bearing anomaly recognition using
an intelligent digital twin integrated with machine learning. Applied
Sciences, 11(10):4602, 2021.
[241] Gao Yiping, Li Xinyu, and Liang Gao. A deep lifelong learning method
for digital twin-driven defect recognition with novel classes. Journal of
Computing and Information Science in Engineering, 21(3):031004, 2021.
[242] Eric J Tuegel, Anthony R Ingraffea, Thomas G Eason, and S Michael
Spottswood. Reengineering aircraft structural life prediction using a
digital twin. International Journal of Aerospace Engineering, 2011, 2011.
[243] Dmitry Kostenko, Nikita Kudryashov, Michael Maystrishin, Vadim
Onufriev, Vyacheslav Potekhin, and Alexey Vasiliev. Digital twin ap-
plications: Diagnostics, optimisation and prediction. Annals of DAAAM
& Proceedings, 29, 2018.
[244] Torbjørn Moi, Andrej Cibicik, and Terje Rølv˚
ag. Digital twin based
condition monitoring of a knuckle boom crane: An experimental study.
Engineering Failure Analysis, 112:104517, 2020.
[245] Wladimir Hofmann and Fredrik Branding. Implementation of an
iot-and cloud-based digital twin for real-time decision support in port
operations. IFAC-PapersOnLine, 52(13):2104–2109, 2019.
[246] Jay Lee, Moslem Azamfar, Jaskaran Singh, and Shahin Siahpour.
Integration of digital twin and deep learning in cyber-physical systems:
towards smart manufacturing. IET Collaborative Intelligent Manufactur-
ing, 2(1):34–36, 2020.
[247] Heikki Laaki, Yoan Miche, and Kari Tammi. Prototyping a digital twin
for real time remote control over mobile networks: Application of remote
surgery. IEEE Access, 7:20325–20336, 2019.
[248] Ying Liu, Lin Zhang, Yuan Yang, Longfei Zhou, Lei Ren, Fei Wang,
Rong Liu, Zhibo Pang, and M Jamal Deen. A novel cloud-based
framework for the elderly healthcare services using digital twin. IEEE
Access, 7:49088–49101, 2019.
[249] Gary White, Anna Zink, Lara Codec ´
a, and Siobh´
an Clarke. A digital
twin smart city for citizen feedback. Cities, 110:103064, 2021.
[250] Li Wan, Timea Nochta, and JM Schooling. Developing a city-level
digital twin–propositions and a case study. In International Conference
on Smart Infrastructure and Construction 2019 (ICSIC) Driving data-
informed decision-making, pages 187–194. ICE Publishing, 2019.
[251] Ziran Wang, Xishun Liao, Xuanpeng Zhao, Kyungtae Han, Prashant
Tiwari, Matthew J Barth, and Guoyuan Wu. A digital twin paradigm:
Vehicle-to-cloud based advanced driver assistance systems. In 2020
IEEE 91st Vehicular Technology Conference (VTC2020-Spring), pages
1–6. IEEE, 2020.
[252] Timo Ruohom¨
aki, Enni Airaksinen, Petteri Huuska, Outi Kes¨
aniemi,
Mikko Martikka, and Jarmo Suomisto. Smart city platform enabling
digital twin. In 2018 International Conference on Intelligent Systems
(IS), pages 155–161. IEEE, 2018.
[253] Qinglin Qi and Fei Tao. Digital twin and big data towards smart
manufacturing and industry 4.0: 360 degree comparison. Ieee Access,
6:3585–3593, 2018.
[254] Qingfei Min, Yangguang Lu, Zhiyong Liu, Chao Su, and Bo Wang.
Machine learning based digital twin framework for production optimiza-
tion in petrochemical industry. International Journal of Information
Management, 49:502–519, 2019.
[255] Bhuman Soni and Philip Hingston. Bots trained to play like a human
are more fun. In 2008 IEEE International Joint Conference on Neural

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 54
Networks (IEEE World Congress on Computational Intelligence), pages
363–369. IEEE, 2008.
[256] Rob Gallagher. No sex please, we are finite state machines: On the
melancholy sexlessness of the video game. Games and Culture, 7(6):399–
418, 2012.
[257] Damian Isla. Building a better battle. In Game developers conference,
san francisco, volume 32, 2008.
[258] Xianwen Zhu. Behavior tree design of intelligent behavior of non-
player character (npc) based on unity3d. Journal of Intelligent & Fuzzy
Systems, 37(5):6071–6079, 2019.
[259] Marek Kopel and Tomasz Hajas. Implementing ai for non-player
characters in 3d video games. In Asian Conference on Intelligent
Information and Database Systems, pages 610–619. Springer, 2018.
[260] Ramiro A Agis, Sebastian Gottifredi, and Alejandro J Garc´
ıa. An event-
driven behavior trees extension to facilitate non-player multi-agent coor-
dination in video games. Expert Systems with Applications, 155:113457,
2020.
[261] Pedro Melendez. Controlling non-player characters using support
vector machines. In Proceedings of the 2009 Conference on Future Play
on@ GDC Canada, pages 33–34, 2009.
[262] Hiram Ponce and Ricardo Padilla. A hierarchical reinforcement
learning based artificial intelligence for non-player characters in video
games. In Mexican International Conference on Artificial Intelligence,
pages 172–183. Springer, 2014.
[263] Kristi´
an Kovalsk`
y and George Palamas. Neuroevolution vs reinforce-
ment learning for training non player characters in games: The case of a
self driving car. In International Conference on Intelligent Technologies
for Interactive Entertainment, pages 191–206. Springer, 2020.
[264] Hao Wang, Yang Gao, and Xingguo Chen. Rl-dot: A reinforcement
learning npc team for playing domination games. IEEE Transactions on
Computational intelligence and AI in Games, 2(1):17–26, 2009.
[265] Frank G Glavin and Michael G Madden. Learning to shoot in first
person shooter games by stabilizing actions and clustering rewards for
reinforcement learning. In 2015 IEEE Conference on Computational
Intelligence and Games (CIG), pages 344–351. IEEE, 2015.
[266] Frank G Glavin and Michael G Madden. Skilled experience catalogue:
A skill-balancing mechanism for non-player characters using reinforce-
ment learning. In 2018 IEEE Conference on Computational Intelligence
and Games (CIG), pages 1–8. IEEE, 2018.
[267] Fei-Yue Wang, Jun Jason Zhang, Xinhu Zheng, Xiao Wang, Yong
Yuan, Xiaoxiao Dai, Jie Zhang, and Liuqing Yang. Where does alphago
go: From church-turing thesis to alphago thesis and beyond. IEEE/CAA
Journal of Automatica Sinica, 3(2):113–120, 2016.
[268] Alanah Davis, John D Murphy, Dawn Owens, Deepak Khazanchi, and
Ilze Zigurs. Avatars, people, and virtual worlds: Foundations for research
in metaverses. Journal of the Association for Information Systems,
10(2):90, 2009.
[269] Anton Nijholt. Humans as avatars in smart and playable cities. In 2017
International Conference on Cyberworlds (CW), pages 190–193. IEEE,
2017.
[270] Panayiotis Koutsabasis, Spyros Vosinakis, Katerina Malisova, and
Nikos Paparounas. On the value of virtual worlds for collaborative design.
Design Studies, 33(4):357–390, 2012.
[271] Xin Yi, Ekta Walia, and Paul Babyn. Generative adversarial network
in medical imaging: A review. Medical image analysis, 58:101552, 2019.
[272] Yanghua Jin, Jiakai Zhang, Minjun Li, Yingtao Tian, Huachun Zhu,
and Zhihao Fang. Towards the automatic anime characters creation with
generative adversarial networks. arXiv preprint arXiv:1708.05509, 2017.
[273] Hongyu Li and Tianqi Han. Towards diverse anime face generation:
Active label completion and style feature network. In Eurographics (Short
Papers), pages 65–68, 2019.
[274] Koichi Hamada, Kentaro Tachibana, Tianqi Li, Hiroto Honda, and
Yusuke Uchida. Full-body high-resolution anime generation with progres-
sive structure-conditional generative adversarial networks. In Proceedings
of the European Conference on Computer Vision (ECCV) Workshops,
pages 0–0, 2018.
[275] Menglei Chai, Tianjia Shao, Hongzhi Wu, Yanlin Weng, and Kun Zhou.
Autohair: Fully automatic hair modeling from a single image. ACM
Transactions on Graphics, 35(4), 2016.
[276] Takayuki Niki and Takashi Komuro. Semi-automatic creation of an
anime-like 3d face model from a single illustration. In 2019 International
Conference on Cyberworlds (CW), pages 53–56. IEEE, 2019.
[277] Tianyang Shi, Yi Yuan, Changjie Fan, Zhengxia Zou, Zhenwei Shi, and
Yong Liu. Face-to-parameter translation for game character auto-creation.
In Proceedings of the IEEE/CVF International Conference on Computer
Vision, pages 161–170, 2019.
[278] Igor Chal ´
as, Petra Urbanov´
a, Vojtˇ
ech Juˇ
r´
ık, Zuzana Ferkov´
a, Marie
Jandov´
a, Jiˇ
r´
ı Sochor, and Barbora Kozl´
ıkov´
a. Generating various com-
posite human faces from real 3d facial images. The Visual Computer,
33(4):443–458, 2017.
[279] R Herbrich. Drivatars and forza motorsport. http://www.vagamelabs.
com/drivatars-trade-and- forza-motorsport.htm, 2010.
[280] Jorge Mu ˜
noz, German Gutierrez, and Araceli Sanchis. A human-like
torcs controller for the simulated car racing championship. In Proceedings
of the 2010 IEEE Conference on Computational Intelligence and Games,
pages 473–480. IEEE, 2010.
[281] Benjamin Geisler. An Empirical Study of Machine Learning Algorithms
Applied to Modeling Player Behavior in a” First Person Shooter” Video
Game. PhD thesis, Citeseer, 2002.
[282] Matheus RF Mendonc¸a, Heder S Bernardino, and Raul F Neto. Sim-
ulating human behavior in fighting games using reinforcement learning
and artificial neural networks. In 2015 14th Brazilian symposium on
computer games and digital entertainment (SBGames), pages 152–159.
IEEE, 2015.
[283] Dianlei Xu, Yong Li, Xinlei Chen, Jianbo Li, Pan Hui, Sheng Chen,
and Jon Crowcroft. A survey of opportunistic offloading. IEEE
Communications Surveys & Tutorials, 20(3):2198–2236, 2018.
[284] Chris Berg, Sinclair Davidson, and Jason Potts. Blockchain technology
as economic infrastructure: Revisiting the electronic markets hypothesis.
Frontiers in Blockchain, 2:22, 2019.
[285] Wei Cai, Zehua Wang, Jason B Ernst, Zhen Hong, Chen Feng,
and Victor CM Leung. Decentralized applications: The blockchain-
empowered software system. IEEE Access, 6:53019–53033, 2018.
[286] Michael Nofer, Peter Gomber, Oliver Hinz, and Dirk Schiereck.
Blockchain. Business & Information Systems Engineering, 59(3):183–
187, 2017.
[287] Roy Lai and David LEE Kuo Chuen. Blockchain–from public to
private. In Handbook of Blockchain, Digital Finance, and Inclusion,
Volume 2, pages 145–177. Elsevier, 2018.
[288] Ethan Buchman. Tendermint: Byzantine fault tolerance in the age of
blockchains. PhD thesis, 2016.
[289] Aggelos Kiayias, Alexander Russell, Bernardo David, and Roman
Oliynykov. Ouroboros: A provably secure proof-of-stake blockchain
protocol. In Annual International Cryptology Conference, pages 357–
388. Springer, 2017.
[290] Xinxin Fan and Qi Chai. Roll-dpos: a randomized delegated proof of
stake scheme for scalable blockchain-based internet of things systems.
In Proceedings of the 15th EAI International Conference on Mobile and
Ubiquitous Systems: Computing, Networking and Services, pages 482–
484, 2018.
[291] Diego Ongaro and John Ousterhout. In search of an understandable
consensus algorithm. In 2014 {USENIX}Annual Technical Conference
({USENIX} {ATC}14), pages 305–319, 2014.
[292] Satoshi Nakamoto. Bitcoin: A peer-to-peer electronic cash system.
Decentralized Business Review, page 21260, 2008.
[293] Jaysing Bhosale and Sushil Mavale. Volatility of select crypto-
currencies: A comparison of bitcoin, ethereum and litecoin. Annu. Res.
J. SCMS, Pune, 6, 2018.
[294] David Schwartz, Noah Youngs, Arthur Britto, et al. The ripple protocol
consensus algorithm. Ripple Labs Inc White Paper, 5(8):151, 2014.
[295] Gavin Wood et al. Ethereum: A secure decentralised generalised
transaction ledger. Ethereum project yellow paper, 151(2014):1–32, 2014.
[296] Shi-Feng Sun, Man Ho Au, Joseph K Liu, and Tsz Hon Yuen. Ringct
2.0: A compact accumulator-based (linkable ring signature) protocol for
blockchain cryptocurrency monero. In European Symposium on Research
in Computer Security, pages 456–474. Springer, 2017.
[297] J¨
urgen Sturm, Nikolas Engelhard, Felix Endres, Wolfram Burgard, and
Daniel Cremers. A benchmark for the evaluation of rgb-d slam systems.
In 2012 IEEE/RSJ international conference on intelligent robots and
systems, pages 573–580. IEEE, 2012.
[298] Cesar Cadena, Luca Carlone, Henry Carrillo, Yasir Latif, Davide
Scaramuzza, Jos´
e Neira, Ian Reid, and John J Leonard. Past, present,
and future of simultaneous localization and mapping: Toward the robust-
perception age. IEEE Transactions on robotics, 32(6):1309–1332, 2016.
[299] Safa Ouerghi, Nicolas Ragot, R´
emi Boutteau, and Xavier Savatier.
Comparative study of a commercial tracking camera and orb-slam2 for
person localization. In 15th International Conference on Computer Vision
Theory and Applications, pages 357–364. SCITEPRESS-Science and
Technology Publications, 2020.
[300] Raul Mur-Artal and Juan D Tard´
os. Orb-slam2: An open-source slam
system for monocular, stereo, and rgb-d cameras. IEEE transactions on
robotics, 33(5):1255–1262, 2017.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 55
[301] Fanyu Zeng, Wenchao Zeng, and Yan Gan. Orb-slam2 with 6dof
motion. In 2018 IEEE 3rd International Conference on Image, Vision
and Computing (ICIVC), pages 556–559. IEEE, 2018.
[302] David G Lowe. Distinctive image features from scale-invariant key-
points. International journal of computer vision, 60(2):91–110, 2004.
[303] Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. Orb:
An efficient alternative to sift or surf. In 2011 International conference
on computer vision, pages 2564–2571. Ieee, 2011.
[304] Stefan Milz, Georg Arbeiter, Christian Witt, Bassam Abdallah, and
Senthil Yogamani. Visual slam for automated driving: Exploring the
applications of deep learning. In Proceedings of the IEEE Conference
on Computer Vision and Pattern Recognition Workshops, pages 247–257,
2018.
[305] Gerhard Reitmayr, Tobias Langlotz, Daniel Wagner, Alessandro Mul-
loni, Gerhard Schall, Dieter Schmalstieg, and Qi Pan. Simultaneous
localization and mapping for augmented reality. In 2010 International
Symposium on Ubiquitous Virtual Reality, pages 5–8. IEEE, 2010.
[306] Esha Nerurkar, Simon Lynen, and Sheng Zhao. System and method
for concurrent odometry and mapping, October 13 2020. US Patent
10,802,147.
[307] Joydeep Biswas and Manuela Veloso. Depth camera based indoor
mobile robot localization and navigation. In 2012 IEEE International
Conference on Robotics and Automation, pages 1697–1702. IEEE, 2012.
[308] Mrinal K Paul, Kejian Wu, Joel A Hesch, Esha D Nerurkar, and Ster-
gios I Roumeliotis. A comparative analysis of tightly-coupled monocular,
binocular, and stereo vins. In 2017 IEEE International Conference on
Robotics and Automation (ICRA), pages 165–172. IEEE, 2017.
[309] Johannes L Sch ¨
onberger, Marc Pollefeys, Andreas Geiger, and Torsten
Sattler. Semantic visual localization. In Proceedings of the IEEE
conference on computer vision and pattern recognition, pages 6896–6906,
2018.
[310] Ricardo R Barioni, Lucas Figueiredo, Kelvin Cunha, and Veronica
Teichrieb. Human pose tracking from rgb inputs. In 2018 20th Symposium
on Virtual and Augmented Reality (SVR), pages 176–182. IEEE, 2018.
[311] Hideaki Uchiyama and Eric Marchand. Object detection and pose
tracking for augmented reality: Recent approaches. In 18th Korea-Japan
Joint Workshop on Frontiers of Computer Vision (FCV), 2012.
[312] Armelle Bauer, Debanga Raj Neog, Ali-Hamadi Dicko, Dinesh K Pai,
Franc¸ois Faure, Olivier Palombi, and Jocelyne Troccaz. Anatomical aug-
mented reality with 3d commodity tracking and image-space alignment.
Computers & Graphics, 69:140–153, 2017.
[313] Thies Pfeiffer and Patrick Renner. Eyesee3d: a low-cost approach
for analyzing mobile 3d eye tracking data using computer vision and
augmented reality technology. In Proceedings of the Symposium on Eye
Tracking Research and Applications, pages 195–202, 2014.
[314] Sebastian Kapp, Michael Barz, Sergey Mukhametov, Daniel Sonntag,
and Jochen Kuhn. Arett: Augmented reality eye tracking toolkit for head
mounted displays. Sensors, 21(6):2234, 2021.
[315] Ruohan Zhang, Calen Walshe, Zhuode Liu, Lin Guan, Karl Muller, Jake
Whritner, Luxin Zhang, Mary Hayhoe, and Dana Ballard. Atari-head:
Atari human eye-tracking and demonstration dataset. In Proceedings of
the AAAI conference on artificial intelligence, volume 34, pages 6811–
6820, 2020.
[316] Kyle Krafka, Aditya Khosla, Petr Kellnhofer, Harini Kannan, Suchen-
dra Bhandarkar, Wojciech Matusik, and Antonio Torralba. Eye tracking
for everyone. In Proceedings of the IEEE conference on computer vision
and pattern recognition, pages 2176–2184, 2016.
[317] Mykhaylo Andriluka, Umar Iqbal, Eldar Insafutdinov, Leonid
Pishchulin, Anton Milan, Juergen Gall, and Bernt Schiele. Posetrack:
A benchmark for human pose estimation and tracking. In Proceedings of
the IEEE conference on computer vision and pattern recognition, pages
5167–5176, 2018.
[318] Valentin Bazarevsky, Ivan Grishchenko, Karthik Raveendran, Tyler
Zhu, Fan Zhang, and Matthias Grundmann. Blazepose: On-device real-
time body pose tracking. CVPR Workshop, 2020.
[319] Ling Shao, Jungong Han, Dong Xu, and Jamie Shotton. Computer
vision for rgb-d sensors: Kinect and its applications [special issue intro.].
IEEE transactions on cybernetics, 43(5):1314–1317, 2013.
[320] Juan C N ´
u˜
nez, Ra´
ul Cabido, Antonio S Montemayor, and Juan J
Pantrigo. Real-time human body tracking based on data fusion from
multiple rgb-d sensors. Multimedia Tools and Applications, 76(3):4249–
4271, 2017.
[321] Lin Wang and Kuk-Jin Yoon. Coaug-mr: An mr-based interactive office
workstation design system via augmented multi-person collaboration.
arXiv preprint arXiv:1907.03107, 2019.
[322] Qi Dang, Jianqin Yin, Bin Wang, and Wenqing Zheng. Deep learning
based 2d human pose estimation: A survey. Tsinghua Science and
Technology, 24(6):663–676, 2019.
[323] Zhe Cao, Gines Hidalgo, Tomas Simon, Shih-En Wei, and Yaser
Sheikh. Openpose: realtime multi-person 2d pose estimation using part
affinity fields. IEEE transactions on pattern analysis and machine
intelligence, 43(1):172–186, 2019.
[324] Hao-Shu Fang, Shuqin Xie, Yu-Wing Tai, and Cewu Lu. Rmpe:
Regional multi-person pose estimation. In Proceedings of the IEEE
international conference on computer vision, pages 2334–2343, 2017.
[325] Dushyant Mehta, Srinath Sridhar, Oleksandr Sotnychenko, Helge
Rhodin, Mohammad Shafiei, Hans-Peter Seidel, Weipeng Xu, Dan Casas,
and Christian Theobalt. Vnect: Real-time 3d human pose estimation with
a single rgb camera. ACM Transactions on Graphics (TOG), 36(4):1–14,
2017.
[326] Jinbao Wang, Shujie Tan, Xiantong Zhen, Shuo Xu, Feng Zheng,
Zhenyu He, and Ling Shao. Deep 3d human pose estimation: A review.
Computer Vision and Image Understanding, page 103225, 2021.
[327] Fang Hu, Peng He, Songlin Xu, Yin Li, and Cheng Zhang. Fingertrak:
Continuous 3d hand pose tracking by deep learning hand silhouettes
captured by miniature thermal cameras on wrist. Proceedings of the ACM
on Interactive, Mobile, Wearable and Ubiquitous Technologies, 4(2):1–24,
2020.
[328] Xin-Yu Huang, Meng-Shiun Tsai, and Ching-Chun Huang. 3d virtual-
reality interaction system. In 2019 IEEE International Conference on
Consumer Electronics-Taiwan (ICCE-TW), pages 1–2. IEEE, 2019.
[329] Erika D’Antonio, Juri Taborri, Eduardo Palermo, Stefano Rossi, and
Fabrizio Patan`
e. A markerless system for gait analysis based on openpose
library. In 2020 IEEE International Instrumentation and Measurement
Technology Conference (I2MTC), pages 1–6. IEEE, 2020.
[330] Roman Bajireanu, Joao AR Pereira, Ricardo JM Veiga, Joao DP Sardo,
Pedro JS Cardoso, Roberto Lam, and Joao MF Rodrigues. Mobile human
shape superimposition: an initial approach using openpose. International
Journal of Computers, 4, 2019.
[331] Cristina Nuzzi, Stefano Ghidini, Roberto Pagani, Simone Pasinetti,
Gabriele Coffetti, and Giovanna Sansoni. Hands-free: a robot augmented
reality teleoperation system. In 2020 17th International Conference on
Ubiquitous Robots (UR), pages 617–624. IEEE, 2020.
[332] Xuanyu Wang, Yang Wang, Yan Shi, Weizhan Zhang, and Qinghua
Zheng. Avatarmeeting: An augmented reality remote interaction system
with personalized avatars. In Proceedings of the 28th ACM International
Conference on Multimedia, pages 4533–4535, 2020.
[333] Youn-ji Shin, Hyun-ju Lee, Jun-hee Kim, Da-young Kwon, Seon-
ae Lee, Yun-jin Choo, Ji-hye Park, Ja-hyun Jung, Hyoung-suk Lee,
and Joon-ho Kim. Non-face-to-face online home training application
study using deep learning-based image processing technique and standard
exercise program. The Journal of the Convergence on Culture Technology,
7(3):577–582, 2021.
[334] Ce Zheng, Wenhan Wu, Taojiannan Yang, Sijie Zhu, Chen Chen,
Ruixu Liu, Ju Shen, Nasser Kehtarnavaz, and Mubarak Shah. Deep
learning-based human pose estimation: A survey. arXiv preprint
arXiv:2012.13392, 2020.
[335] Luiz Jos ´
e Schirmer Silva, Djalma L´
ucio Soares da Silva, Alberto Bar-
bosa Raposo, Luiz Velho, and H´
elio Cˆ
ortes Vieira Lopes. Tensorpose:
Real-time pose estimation for interactive applications. Computers &
Graphics, 85:1–14, 2019.
[336] Katarzyna Czesak, Raul Mohedano, Pablo Carballeira, Julian Cabrera,
and Narciso Garcia. Fusion of pose and head tracking data for immersive
mixed-reality application development. In 2016 3DTV-Conference: The
True Vision-Capture, Transmission and Display of 3D Video (3DTV-
CON), pages 1–4. IEEE, 2016.
[337] Eric Marchand, Hideaki Uchiyama, and Fabien Spindler. Pose esti-
mation for augmented reality: a hands-on survey. IEEE transactions on
visualization and computer graphics, 22(12):2633–2651, 2015.
[338] Yongzhi Su, Jason Rambach, Nareg Minaskan, Paul Lesur, Alain
Pagani, and Didier Stricker. Deep multi-state object pose estimation for
augmented reality assembly. In 2019 IEEE International Symposium on
Mixed and Augmented Reality Adjunct (ISMAR-Adjunct), pages 222–227.
IEEE, 2019.
[339] Pooja Nagpal and Piyush Prasad. Pose estimation and 3d model overlay
in real time for applications in augmented reality. In Intelligent Systems,
pages 201–208. Springer, 2021.
[340] Norman Murray, Dave Roberts, Anthony Steed, Paul Sharkey, Paul
Dickerson, and John Rae. An assessment of eye-gaze potential within
immersive virtual environments. ACM Transactions on Multimedia Com-
puting, Communications, and Applications (TOMM), 3(4):1–17, 2007.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 56
[341] Adrian Haffegee and Russell Barrow. Eye tracking and gaze based
interaction within immersive virtual environments. In International
Conference on Computational Science, pages 729–736. Springer, 2009.
[342] Vildan Tanriverdi and Robert JK Jacob. Interacting with eye movements
in virtual environments. In Proceedings of the SIGCHI conference on
Human Factors in Computing Systems, pages 265–272, 2000.
[343] Viviane Clay, Peter K¨
onig, and Sabine Koenig. Eye tracking in virtual
reality. Journal of Eye Movement Research, 12(1), 2019.
[344] Sylvia Peißl, Christopher D Wickens, and Rithi Baruah. Eye-tracking
measures in aviation: A selective literature review. The International
Journal of Aerospace Psychology, 28(3-4):98–112, 2018.
[345] Lin Wang and Kuk-Jin Yoon. Psat-gan: Efficient adversarial attacks
against holistic scene understanding. IEEE Transactions on Image
Processing, 2021.
[346] Sercan Turkmen. Scene understanding through semantic image seg-
mentation in augmented reality. 2019.
[347] Xiang Li, Yuan Tian, Fuyao Zhang, Shuxue Quan, and Yi Xu. Object
detection in the context of mobile augmented reality. In 2020 IEEE
International Symposium on Mixed and Augmented Reality (ISMAR),
pages 156–163. IEEE, 2020.
[348] Gaurav Chaurasia, Arthur Nieuwoudt, Alexandru-Eugen Ichim,
Richard Szeliski, and Alexander Sorkine-Hornung. Passthrough+ real-
time stereoscopic view synthesis for mobile mixed reality. Proceedings
of the ACM on Computer Graphics and Interactive Techniques, 3(1):1–17,
2020.
[349] Matthias Schr ¨
oder and Helge Ritter. Deep learning for action recogni-
tion in augmented reality assistance systems. In ACM SIGGRAPH 2017
Posters, pages 1–2. 2017.
[350] Lin Wang, Yujeong Chae, Sung-Hoon Yoon, Tae-Kyun Kim, and
Kuk-Jin Yoon. Evdistill: Asynchronous events to end-task learning via
bidirectional reconstruction-guided cross-modal knowledge distillation.
In Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition, pages 608–619, 2021.
[351] Lin Wang, Yujeong Chae, and Kuk-Jin Yoon. Dual transfer learning for
event-based end-task prediction via pluggable event to image translation.
ICCV, 2021.
[352] Leonardo Tanzi, Pietro Piazzolla, Francesco Porpiglia, and Enrico
Vezzetti. Real-time deep learning semantic segmentation during intra-
operative surgery for 3d augmented reality assistance. International
Journal of Computer Assisted Radiology and Surgery, pages 1–11, 2021.
[353] Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff,
and Hartwig Adam. Encoder-decoder with atrous separable convolution
for semantic image segmentation. In Proceedings of the European
conference on computer vision (ECCV), pages 801–818, 2018.
[354] Tae-young Ko and Seung-ho Lee. Novel method of semantic segmen-
tation applicable to augmented reality. Sensors, 20(6):1737, 2020.
[355] Luyang Liu, Hongyu Li, and Marco Gruteser. Edge assisted real-
time object detection for mobile augmented reality. In The 25th Annual
International Conference on Mobile Computing and Networking, pages
1–16, 2019.
[356] William S Noble. What is a support vector machine? Nature
biotechnology, 24(12):1565–1567, 2006.
[357] Jenny Lin, Xingwen Guo, Jingyu Shao, Chenfanfu Jiang, Yixin Zhu,
and Song-Chun Zhu. A virtual reality platform for dynamic human-
scene interaction. In SIGGRAPH ASIA 2016 virtual reality meets physical
reality: Modelling and simulating virtual humans and environments, pages
1–4. 2016.
[358] Peer Sch ¨
utt, Max Schwarz, and Sven Behnke. Semantic interaction in
augmented reality environments for microsoft hololens. In 2019 European
Conference on Mobile Robots (ECMR), pages 1–6. IEEE, 2019.
[359] Huanle Zhang, Bo Han, Cheuk Yiu Ip, and Prasant Mohapatra.
Slimmer: Accelerating 3d semantic segmentation for mobile augmented
reality. In 2020 IEEE 17th International Conference on Mobile Ad Hoc
and Sensor Systems (MASS), pages 603–612. IEEE, 2020.
[360] Daiki Kido, Tomohiro Fukuda, and Nobuyoshi Yabuki. Assessing fu-
ture landscapes using enhanced mixed reality with semantic segmentation
by deep learning. Advanced Engineering Informatics, 48:101281, 2021.
[361] Menandro Roxas, Tomoki Hori, Taiki Fukiage, Yasuhide Okamoto,
and Takeshi Oishi. Occlusion handling using semantic segmentation and
visibility-based rendering for mixed reality. In Proceedings of the 24th
ACM Symposium on Virtual Reality Software and Technology, pages 1–8,
2018.
[362] Hengshuang Zhao, Xiaojuan Qi, Xiaoyong Shen, Jianping Shi, and
Jiaya Jia. Icnet for real-time semantic segmentation on high-resolution
images. In Proceedings of the European conference on computer vision
(ECCV), pages 405–420, 2018.
[363] Mennatullah Siam, Mostafa Gamal, Moemen Abdel-Razek, Senthil
Yogamani, and Martin Jagersand. Rtseg: Real-time semantic segmentation
comparative study. In 2018 25th IEEE International Conference on Image
Processing (ICIP), pages 1603–1607. IEEE, 2018.
[364] Sachin Mehta, Mohammad Rastegari, Anat Caspi, Linda Shapiro,
and Hannaneh Hajishirzi. Espnet: Efficient spatial pyramid of dilated
convolutions for semantic segmentation. In Proceedings of the european
conference on computer vision (ECCV), pages 552–568, 2018.
[365] Yifan Liu, Ke Chen, Chris Liu, Zengchang Qin, Zhenbo Luo, and
Jingdong Wang. Structured knowledge distillation for semantic segmen-
tation. In Proceedings of the IEEE/CVF Conference on Computer Vision
and Pattern Recognition, pages 2604–2613, 2019.
[366] Lin Wang and Kuk-Jin Yoon. Knowledge distillation and student-
teacher learning for visual intelligence: A review and new outlooks. IEEE
Transactions on Pattern Analysis and Machine Intelligence, 2021.
[367] Daiki Kido, Tomohiro Fukuda, and Nobuyoshi Yabuki. Mobile mixed
reality for environmental design using real-time semantic segmentation
and video communication-dynamic occlusion handling and green view
index estimation. 2020.
[368] Andrija Gajic, Ester Gonzalez-Sosa, Diego Gonzalez-Morin, Marcos
Escudero-Vinolo, and Alvaro Villegas. Egocentric human segmentation
for mixed reality. arXiv preprint arXiv:2005.12074, 2020.
[369] Long Chen, Wen Tang, Nigel W John, Tao Ruan Wan, and Jian J
Zhang. Context-aware mixed reality: A learning-based framework for
semantic-level interaction. In Computer Graphics Forum, volume 39,
pages 484–496. Wiley Online Library, 2020.
[370] Youssef Hbali, Lahoucine Ballihi, Mohammed Sadgal, and Abdelaziz
El Fazziki. Face detection for augmented reality application using
boosting-based techniques. Int. J. Interact. Multim. Artif. Intell., 4(2):22–
28, 2016.
[371] Nahuel A Mangiarua, Jorge S Ierache, and Mar´
ıa J Ab´
asolo. Scalable
integration of image and face based augmented reality. In International
Conference on Augmented Reality, Virtual Reality and Computer Graph-
ics, pages 232–242. Springer, 2020.
[372] Xueshi Lu, Difeng Yu, Hai-Ning Liang, Wenge Xu, Yuzheng Chen,
Xiang Li, and Khalad Hasan. Exploration of hands-free text entry
techniques for virtual reality. In 2020 IEEE International Symposium
on Mixed and Augmented Reality (ISMAR), pages 344–349. IEEE, 2020.
[373] Tanja Koji´
c, Danish Ali, Robert Greinacher, Sebastian M¨
oller, and
Jan-Niklas Voigt-Antons. User experience of reading in virtual real-
ity—finding values for text distance, size and contrast. In 2020 Twelfth
International Conference on Quality of Multimedia Experience (QoMEX),
pages 1–6. IEEE, 2020.
[374] Amin Golnari, Hossein Khosravi, and Saeid Sanei. Deepfacear: deep
face recognition and displaying personal information via augmented
reality. In 2020 International Conference on Machine Vision and Image
Processing (MVIP), pages 1–7. IEEE, 2020.
[375] Bernardo Marques, Paulo Dias, Jo˜
ao Alves, and Beatriz Sousa Santos.
Adaptive augmented reality user interfaces using face recognition for
smart home control. In International Conference on Human Systems
Engineering and Design: Future Trends and Applications, pages 15–19.
Springer, 2019.
[376] Jan Svensson and Jonatan Atles. Object detection in augmented reality.
Master’s theses in mathematical sciences, 2018.
[377] Alessandro Acquisti, Ralph Gross, and Frederic D Stutzman. Face
recognition and privacy in the age of augmented reality. Journal of
Privacy and Confidentiality, 6(2):1, 2014.
[378] Alessandro Acquisti, Ralph Gross, and Fred Stutzman. Faces of
facebook: Privacy in the age of augmented reality. BlackHat USA, 2:1–20,
2011.
[379] Ellysse Dick. How to address privacy questions raised by the expansion
of augmented reality in public spaces. Technical report, Information
Technology and Innovation Foundation, 2020.
[380] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-
cnn: Towards real-time object detection with region proposal networks.
Advances in neural information processing systems, 28:91–99, 2015.
[381] Joseph Redmon and Ali Farhadi. Yolov3: An incremental improvement.
arXiv preprint arXiv:1804.02767, 2018.
[382] Alexey Bochkovskiy, Chien-Yao Wang, and Hong-Yuan Mark Liao.
Yolov4: Optimal speed and accuracy of object detection. arXiv preprint
arXiv:2004.10934, 2020.
[383] Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott
Reed, Cheng-Yang Fu, and Alexander C Berg. Ssd: Single shot multibox
detector. In European conference on computer vision, pages 21–37.
Springer, 2016.
[384] Sagar Mahurkar. Integrating yolo object detection with augmented
reality for ios apps. In 2018 9th IEEE Annual Ubiquitous Computing,

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 57
Electronics & Mobile Communication Conference (UEMCON), pages
585–589. IEEE, 2018.
[385] Martin Simony, Stefan Milzy, Karl Amendey, and Horst-Michael Gross.
Complex-yolo: An euler-region-proposal for real-time 3d object detection
on point clouds. In Proceedings of the European Conference on Computer
Vision (ECCV) Workshops, pages 0–0, 2018.
[386] Haythem Bahri, David Krˇ
cmaˇ
r´
ık, and Jan Koˇ
c´
ı. Accurate object
detection system on hololens using yolo algorithm. In 2019 International
Conference on Control, Artificial Intelligence, Robotics & Optimization
(ICCAIRO), pages 219–224. IEEE, 2019.
[387] Fatima El Jamiy and Ronald Marsh. Distance estimation in virtual
reality and augmented reality: A survey. In 2019 IEEE International
Conference on Electro Information Technology (EIT), pages 063–068.
IEEE, 2019.
[388] Daniel Scharstein and Richard Szeliski. A taxonomy and evaluation of
dense two-frame stereo correspondence algorithms. International journal
of computer vision, 47(1):7–42, 2002.
[389] Po Kong Lai, Shuang Xie, Jochen Lang, and Robert Lagani`
ere. Real-
time panoramic depth maps from omni-directional stereo images for 6
dof videos in virtual reality. In 2019 IEEE Conference on Virtual Reality
and 3D User Interfaces (VR), pages 405–412. IEEE, 2019.
[390] Ling Li, Xiaojian Li, Shanlin Yang, Shuai Ding, Alireza Jolfaei, and
Xi Zheng. Unsupervised-learning-based continuous depth and motion es-
timation with monocular endoscopy for virtual reality minimally invasive
surgery. IEEE Transactions on Industrial Informatics, 17(6):3920–3928,
2020.
[391] Donald R Lampton, Daniel P McDonald, Michael Singer, and James P
Bliss. Distance estimation in virtual environments. In Proceedings of the
human factors and ergonomics society annual meeting, volume 39, pages
1268–1272. SAGE Publications Sage CA: Los Angeles, CA, 1995.
[392] Jack M Loomis, Joshua M Knapp, et al. Visual perception of
egocentric distance in real and virtual environments. Virtual and adaptive
environments, 11:21–46, 2003.
[393] Peter Willemsen, Mark B Colton, Sarah H Creem-Regehr, and
William B Thompson. The effects of head-mounted display mechanics
on distance judgments in virtual environments. In Proceedings of the 1st
Symposium on Applied Perception in Graphics and Visualization, pages
35–38, 2004.
[394] Kristina Prokopetc and Romain Dupont. Towards dense 3d reconstruc-
tion for mixed reality in healthcare: Classical multi-view stereo vs deep
learning. In Proceedings of the IEEE/CVF International Conference on
Computer Vision Workshops, pages 0–0, 2019.
[395] Alberto Bad´
ıas, David Gonz´
alez, Ic´
ıar Alfaro, Francisco Chinesta, and
El´
ıas Cueto. Real-time interaction of virtual and physical objects in
mixed reality applications. International Journal for Numerical Methods
in Engineering, 121(17):3849–3868, 2020.
[396] Jiamin Ping, Bruce H Thomas, James Baumeister, Jie Guo, Dongdong
Weng, and Yue Liu. Effects of shading model and opacity on depth
perception in optical see-through augmented reality. Journal of the Society
for Information Display, 28(11):892–904, 2020.
[397] Masayuki Kanbara, Takashi Okuma, Haruo Takemura, and Naokazu
Yokoya. A stereoscopic video see-through augmented reality system
based on real-time vision-based registration. In Proceedings IEEE Virtual
Reality 2000 (Cat. No. 00CB37048), pages 255–262. IEEE, 2000.
[398] Jan Fischer and Dirk Bartz. Handling photographic imperfections and
aliasing in augmented reality. 2006.
[399] Na Li and Yao Liu. Applying vertexshuffle toward 360-degree
video super-resolution on focused-icosahedral-mesh. arXiv preprint
arXiv:2106.11253, 2021.
[400] Yi Zhu, Xinyu Li, Chunhui Liu, Mohammadreza Zolfaghari, Yuanjun
Xiong, Chongruo Wu, Zhi Zhang, Joseph Tighe, R Manmatha, and Mu Li.
A comprehensive study of deep video action recognition. arXiv preprint
arXiv:2012.06567, 2020.
[401] Cezary Sieluzycki, Patryk Kaczmarczyk, Janusz Sobecki, Kazimierz
Witkowski, Jarosław Ma´
sli´
nski, and Wojciech Cie´
sli´
nski. Microsoft
kinect as a tool to support training in professional sports: augmented
reality application to tachi-waza techniques in judo. In 2016 Third Eu-
ropean Network Intelligence Conference (ENIC), pages 153–158. IEEE,
2016.
[402] Dongjin Huang, Chao Wang, Youdong Ding, and Wen Tang. Virtual
throwing action recognition based on time series data mining in an
augmented reality system. In 2010 International Conference on Audio,
Language and Image Processing, pages 955–959. IEEE, 2010.
[403] Cen Rao and Mubarak Shah. View-invariance in action recognition. In
Proceedings of the 2001 IEEE Computer Society Conference on Computer
Vision and Pattern Recognition. CVPR 2001, volume 2, pages II–II. IEEE,
2001.
[404] Daeho Lee and SeungGwan Lee. Vision-based finger action recognition
by angle detection and contour analysis. ETRI journal, 33(3):415–422,
2011.
[405] Jiaqi Dong, Zisheng Tang, and Qunfei Zhao. Gesture recognition in
augmented reality assisted assembly training. In Journal of Physics:
Conference Series, volume 1176, page 032030. IOP Publishing, 2019.
[406] Seungeun Chung, Jiyoun Lim, Kyoung Ju Noh, Gague Kim, and
Hyuntae Jeong. Sensor data acquisition and multimodal sensor fusion
for human activity recognition using deep learning. Sensors, 19(7):1716,
2019.
[407] Javier Mar´
ın-Morales, Carmen Llinares, Jaime Guixeres, and Mariano
Alca˜
niz. Emotion recognition in immersive virtual reality: From statistics
to affective computing. Sensors, 20(18):5163, 2020.
[408] Young D Kwon, Jagmohan Chauhan, Abhishek Kumar, Pan Hui, and
Cecilia Mascolo. Exploring system performance of continual learning for
mobile and embedded sensing applications. In ACM/IEEE Symposium on
Edge Computing. Association for Computing Machinery (ACM), 2021.
[409] Lin Wang, Tae-Kyun Kim, and Kuk-Jin Yoon. Joint framework for
single image reconstruction and super-resolution with an event camera.
IEEE Transactions on Pattern Analysis & Machine Intelligence, (01):1–1,
2021.
[410] Lin Wang, Tae-Kyun Kim, and Kuk-Jin Yoon. Eventsr: From asyn-
chronous events to image reconstruction, restoration, and super-resolution
via end-to-end adversarial learning. In Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition, pages 8315–
8325, 2020.
[411] Lin Wang and Kuk-Jin Yoon. Semi-supervised student-teacher learning
for single image super-resolution. Pattern Recognition, 121:108206, 2022.
[412] Lin Wang, Yo-Sung Ho, Kuk-Jin Yoon, et al. Event-based high dynamic
range image and very high frame rate video generation using conditional
generative adversarial networks. In Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition, pages 10081–
10090, 2019.
[413] Xiaojuan Xu and Jin Zhu. Artistic color virtual reality implementation
based on similarity image restoration. Complexity, 2021, 2021.
[414] XL Zhao and YG Wang. Optimization and simulation of image
restoration in virtual reality. Computer Simulation, 34(4):440–443, 2017.
[415] Chengquan Qiao, Wenwen Zhang, Decai Gong, and Yuxuan Gong.
In situ virtual restoration of artifacts by imaging technology. Heritage
Science, 8(1):1–13, 2020.
[416] Shohei Mori, Sei Ikeda, and Hideo Saito. A survey of diminished real-
ity: Techniques for visually concealing, eliminating, and seeing through
real objects. IPSJ Transactions on Computer Vision and Applications,
9(1):1–14, 2017.
[417] Marek ˇ
Zuˇ
zi, Jan ˇ
Cejka, Fabio Bruno, Dimitrios Skarlatos, and Fotis
Liarokapis. Impact of dehazing on underwater marker detection for
augmented reality. Frontiers in Robotics and AI, 5:92, 2018.
[418] Bunyo Okumura, Masayuki Kanbara, and Naokazu Yokoya. Aug-
mented reality based on estimation of defocusing and motion blurring
from captured images. In 2006 IEEE/ACM International Symposium on
Mixed and Augmented Reality, pages 219–225. IEEE, 2006.
[419] Bunyo Okumura, Masayuki Kanbara, and Naokazu Yokoya. Image
composition based on blur estimation from captured image for augmented
reality. In Proc. of IEEE Virtual Reality, pages 128–134, 2006.
[420] Paolo Clini, Emanuele Frontoni, Ramona Quattrini, and Roberto
Pierdicca. Augmented reality experience: From high-resolution acqui-
sition to real time augmented contents. Advances in Multimedia, 2014,
2014.
[421] Dejan Graboviˇ
cki´
c, Pablo Benitez, Juan C Mi˜
nano, Pablo Zamora,
Marina Buljan, Bharathwaj Narasimhan, Milena I Nikolic, Jesus Lopez,
Jorge Gorospe, Eduardo Sanchez, et al. Super-resolution optics for
virtual reality. In Digital Optical Technologies 2017, volume 10335, page
103350G. International Society for Optics and Photonics, 2017.
[422] Bharathwaj Appan Narasimhan. Ultra-compact pancake optics based
on thineyes super-resolution technology for virtual reality headsets. In
Digital Optics for Immersive Displays, volume 10676, page 106761G.
International Society for Optics and Photonics, 2018.
[423] Chia-Hui Feng, Yu-Hsiu Hung, Chao-Kuang Yang, Liang-Chi Chen,
Wen-Cheng Hsu, and Shih-Hao Lin. Applying holo360 video and
image super-resolution generative adversarial networks to virtual reality
immersion. In International Conference on Human-Computer Interaction,
pages 569–584. Springer, 2020.
[424] Alisa Korinevskaya and Ilya Makarov. Fast depth map super-resolution
using deep neural network. In 2018 IEEE international symposium on
mixed and augmented reality adjunct (ISMAR-Adjunct), pages 117–122.
IEEE, 2018.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 58
[425] Vida Fakour Sevom, E. Guldogan, and J. K¨
am¨
ar¨
ainen. 360 panorama
super-resolution using deep convolutional networks. In VISIGRAPP,
2018.
[426] Dehua Song, Yunhe Wang, Hanting Chen, Chang Xu, Chunjing Xu, and
DaCheng Tao. Addersr: Towards energy efficient image super-resolution.
In Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition, pages 15648–15657, 2021.
[427] Cloud ar/vr whitepaper, April 2019.
[428] Jens Grubert, Tobias Langlotz, Stefanie Zollmann, and Holger Re-
genbrecht. Towards pervasive augmented reality: Context-awareness in
augmented reality. IEEE Transactions on Visualization and Computer
Graphics, 23(6):1706–1724, 2017.
[429] Katerina Mania, Bernard D. Adelstein, Stephen R. Ellis, and Michael I.
Hill. Perceptual sensitivity to head tracking latency in virtual environ-
ments with varying degrees of scene complexity. In Proceedings of the 1st
Symposium on Applied Perception in Graphics and Visualization, APGV
’04, page 39–47, New York, NY, USA, 2004. Association for Computing
Machinery.
[430] Richard L Holloway. Registration error analysis for augmented reality.
Presence: Teleoperators & Virtual Environments, 6(4):413–432, 1997.
[431] Henry Fuchs, Mark A Livingston, Ramesh Raskar, Kurtis Keller,
Jessica R Crawford, Paul Rademacher, Samuel H Drake, Anthony A
Meyer, et al. Augmented reality visualization for laparoscopic surgery. In
International Conference on Medical Image Computing and Computer-
Assisted Intervention, pages 934–943. Springer, 1998.
[432] Luc Soler, St´
ephane Nicolau, J´
erˆ
ome Schmid, Christophe Koehl,
Jacques Marescaux, Xavier Pennec, and Nicholas Ayache. Virtual reality
and augmented reality in digestive surgery. In Third IEEE and ACM
International Symposium on Mixed and Augmented Reality, pages 278–
279. IEEE, 2004.
[433] Phattanapon Rhienmora, Kugamoorthy Gajananan, Peter Haddawy,
Matthew N Dailey, and Siriwan Suebnukarn. Augmented reality haptics
system for dental surgical skills training. In Proceedings of the 17th
ACM Symposium on Virtual Reality Software and Technology, pages 97–
98, 2010.
[434] Lu Li and Ji Zhou. Virtual reality technology based developmental de-
signs of multiplayer-interaction-supporting exhibits of science museums:
taking the exhibit of” virtual experience on an aircraft carrier” in china
science and technology museum as an example. In Proceedings of the
15th ACM SIGGRAPH Conference on Virtual-Reality Continuum and Its
Applications in Industry-Volume 1, pages 409–412, 2016.
[435] Tristan Braud, ZHOU Pengyuan, Jussi Kangasharju, and HUI Pan.
Multipath computation offloading for mobile augmented reality. In 2020
IEEE International Conference on Pervasive Computing and Communi-
cations (PerCom), pages 1–10. IEEE, 2020.
[436] Abid Yaqoob and Gabriel-Miro Muntean. A combined field-of-view
prediction-assisted viewport adaptive delivery scheme for 360° videos.
IEEE Transactions on Broadcasting, 67(3):746–760, 2021.
[437] Abbas Mehrabi, Matti Siekkinen, Teemu K¨
am¨
ar¨
ainen, and Antti Yl¨
a-
J¨
a¨
aski. Multi-tier cloudvr: Leveraging edge computing in remote rendered
virtual reality. ACM Transactions on Multimedia Computing, Communi-
cations, and Applications (TOMM), 17(2):1–24, 2021.
[438] Ang Li, Xiaowei Yang, Srikanth Kandula, and Ming Zhang. Cloudcmp:
Comparing public cloud providers. In Proceedings of the 10th ACM
SIGCOMM Conference on Internet Measurement, IMC ’10, page 1–14,
New York, NY, USA, 2010. Association for Computing Machinery.
[439] Mahadev Satyanarayanan, Paramvir Bahl, Ramon Caceres, and Nigel
Davies. The case for vm-based cloudlets in mobile computing. IEEE
Pervasive Computing, 8(4):14–23, 2009.
[440] Pengyuan Zhou, Tristan Braud, Aleksandr Zavodovski, Zhi Liu, Xianfu
Chen, Pan Hui, and Jussi Kangasharju. Edge-facilitated augmented
vision in vehicle-to-everything networks. IEEE Transactions on Vehicular
Technology, 69(10):12187–12201, 2020.
[441] T. Braud, F. H. Bijarbooneh, D. Chatzopoulos, and P. Hui. Future
networking challenges: The case of mobile augmented reality. In 2017
IEEE 37th International Conference on Distributed Computing Systems
(ICDCS). pages 1796-1807, June 2017.
[442] Jeffrey Dean and Luiz Andr´
e Barroso. The tail at scale. Communica-
tions of the ACM, 56:74–80, 2013.
[443] Lorenzo Corneo, Maximilian Eder, Nitinder Mohan, Aleksandr Za-
vodovski, Suzan Bayhan, Walter Wong, Per Gunningberg, Jussi Kan-
gasharju, and J¨
org Ott. Surrounded by the clouds: A comprehensive
cloud reachability study. In Proceedings of the Web Conference 2021,
pages 295–304, 2021.
[444] Khang Dang The, Mohan Nitinder, Corneo Lorenzo, Zavodovski
Aleksandr, Ott J¨
org, and Jussi Kangasharju. Cloudy with a chance of
short rtts: Analyzing cloud connectivity in the internet. In ACM Internet
Measurements Conference. ACM, 2021.
[445] Lik-Hang Lee, Abhishek Kumar, Susanna Pirttikangas, and Timo Ojala.
When augmented reality meets edge ai: A vision of collective urban
interfaces. In INTERDISCIPLINARY URBAN AI: DIS Workshop, 2020.
[446] Shu Shi, Varun Gupta, Michael Hwang, and Rittwik Jana. Mobile vr
on edge cloud: a latency-driven design. In Proceedings of the 10th ACM
Multimedia Systems Conference, pages 222–231, 2019.
[447] Zhuo Chen, Wenlu Hu, Junjue Wang, Siyan Zhao, Brandon Amos,
Guanhang Wu, Kiryong Ha, Khalid Elgazzar, Padmanabhan Pillai,
Roberta Klatzky, Daniel Siewiorek, and Mahadev Satyanarayanan. An
empirical study of latency in an emerging class of edge computing
applications for wearable cognitive assistance. In Proceedings of the
Second ACM/IEEE Symposium on Edge Computing, SEC ’17, New York,
NY, USA, 2017. Association for Computing Machinery.
[448] Kiryong Ha, Zhuo Chen, Wenlu Hu, Wolfgang Richter, Padmanabhan
Pillai, and Mahadev Satyanarayanan. Towards wearable cognitive assis-
tance. In Proceedings of the 12th annual international conference on
Mobile systems, applications, and services, pages 68–81, 2014.
[449] Yun Chao Hu, Milan Patel, Dario Sabella, Nurit Sprecher, and Valerie
Young. Mobile edge computing—a key technology towards 5g. ETSI
white paper, 11(11):1–16, 2015.
[450] Wenxiao Zhang, Sikun Lin, Farshid Bijarbooneh, Hao-Fei Cheng,
Tristan Braud, Pengyuan Zhou, Lik-Hang Lee, and Pan Hui. Edgexar: A
6-dof camera multi-target interactionframework for mar with user-friendly
latencycompensation using edge computing. In Proceedings of the ACM
on HCI (Engineering Interactive Computing Systems), 2022.
[451] Wenxiao Zhang, Bo Han, and Pan Hui. Jaguar: Low latency mobile
augmented reality with flexible tracking. In Proceedings of the 26th ACM
international conference on Multimedia, pages 355–363, 2018.
[452] Pengyuan Zhou, Wenxiao Zhang, Tristan Braud, Pan Hui, and Jussi
Kangasharju. Arve: Augmented reality applications in vehicle to edge
networks. In Proceedings of the 2018 Workshop on Mobile Edge
Communications, pages 25–30, 2018.
[453] Peng Lin, Qingyang Song, Dan Wang, Richard Yu, Lei Guo, and Victor
Leung. Resource management for pervasive edge computing-assisted
wireless vr streaming in industrial internet of things. IEEE Transactions
on Industrial Informatics, 2021.
[454] Sabyasachi Gupta, Jacob Chakareski, and Petar Popovski. Millimeter
wave meets edge computing for mobile vr with high-fidelity 8k scalable
360° video. In 2019 IEEE 21st International Workshop on Multimedia
Signal Processing (MMSP), pages 1–6. IEEE, 2019.
[455] Mohammed S Elbamby, Cristina Perfecto, Mehdi Bennis, and Klaus
Doppler. Edge computing meets millimeter-wave enabled vr: Paving the
way to cutting the cord. In 2018 IEEE Wireless Communications and
Networking Conference (WCNC), pages 1–6. IEEE, 2018.
[456] Apple. View 360° video in a vr headset in motion, June 2021.
[457] Qualcomm. Oculus quest 2: How snapdragon xr2 powers the next
generation of vr, October 2020.
[458] Facebook. Introducing oculus air link, a wireless way to play pc vr
games on oculus quest 2, plus infinite office updates, support for 120 hz
on quest 2, and more., April 2021.
[459] Ilija Had ˇ
zi´
c, Yoshihisa Abe, and Hans C Woithe. Edge computing
in the epc: A reality check. In Proceedings of the Second ACM/IEEE
Symposium on Edge Computing, pages 1–10, 2017.
[460] Nitinder Mohan, Aleksandr Zavodovski, Pengyuan Zhou, and Jussi
Kangasharju. Anveshak: Placing edge servers in the wild. In Proceedings
of the 2018 Workshop on Mobile Edge Communications, pages 7–12,
2018.
[461] Pengyuan Zhou, Benjamin Finley, Xuebing Li, Sasu Tarkoma, Jussi
Kangasharju, Mostafa Ammar, and Pan Hui. 5g mec computation handoff
for mobile augmented reality. arXiv preprint arXiv:2101.00256, 2021.
[462] Pei Ren, Xiuquan Qiao, Yakun Huang, Ling Liu, Calton Pu, Schahram
Dustdar, and Jun-Liang Chen. Edge ar x5: An edge-assisted multi-user
collaborative framework for mobile web augmented reality in 5g and
beyond. IEEE Transactions on Cloud Computing, 2020.
[463] Mike Jia and Weifa Liang. Delay-sensitive multiplayer augmented
reality game planning in mobile edge computing. In Proceedings of
the 21st ACM International Conference on Modeling, Analysis and
Simulation of Wireless and Mobile Systems, pages 147–154, 2018.
[464] Hsin-Yuan Chen, Ruey-Tzer Hsu, Ying-Chiao Chen, Wei-Chen Hsu,
and Polly Huang. Ar game traffic characterization: a case of pok´
emon go
in a flash crowd event. In Proceedings of the 19th Annual International
Conference on Mobile Systems, Applications, and Services, pages 493–
494, 2021.
[465] Jianmei Dai, Zhilong Zhang, Shiwen Mao, and Danpu Liu. A view
synthesis-based 360° vr caching system over mec-enabled c-ran. IEEE

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 59
Transactions on Circuits and Systems for Video Technology, 30(10):3843–
3855, 2019.
[466] Zhuojia Gu, Hancheng Lu, Peilin Hong, and Yongdong Zhang. Re-
liability enhancement for vr delivery in mobile-edge empowered dual-
connectivity sub-6 ghz and mmwave hetnets. IEEE Transactions on
Wireless Communications, 2021.
[467] Yanwei Liu, Jinxia Liu, Antonios Argyriou, and Song Ci. Mec-assisted
panoramic vr video streaming over millimeter wave mobile networks.
IEEE Transactions on Multimedia, 21(5):1302–1316, 2018.
[468] Yahoo. Real-world metaverse ’twinworld’ selected as 5g telco edge
cloud testbed for 3 global mobile carriers, August 2021.
[469] Niantic. Niantic planet-scale ar alliance accelerates social ar future in
codename: Urban legends, March 2021.
[470] Ronald Leenes. Privacy in the metaverse. In IFIP International Summer
School on the Future of Identity in the Information Society, pages 95–112.
Springer, 2007.
[471] Ben Falchuk, Shoshana Loeb, and Ralph Neff. The social metaverse:
Battle for privacy. IEEE Technology and Society Magazine, 37(2):52–61,
2018.
[472] Fatima Alqubaisi, Ahmad Samer Wazan, Liza Ahmad, and David W
Chadwick. Should we rush to implement password-less single factor
fido2 based authentication? In 2020 12th Annual Undergraduate Research
Conference on Applied Computing (URC), pages 1–6. IEEE, 2020.
[473] Morey J Haber. Passwordless authentication. In Privileged Attack
Vectors, pages 87–98. Springer, 2020.
[474] Juliet Lodge. Nameless and faceless: The role of biometrics in realising
quantum (in) security and (un) accountability. In Security and Privacy in
Biometrics, pages 311–337. Springer, 2013.
[475] Ghislaine Boddington. The internet of bodies—alive, connected and
collective: the virtual physical future of our bodies and our senses. Ai &
Society, pages 1–17, 2021.
[476] Nalini Ratha, Jonathan Connell, Ruud M Bolle, and Sharat Chikkerur.
Cancelable biometrics: A case study in fingerprints. In 18th International
Conference on Pattern Recognition (ICPR’06), volume 4, pages 370–373.
IEEE, 2006.
[477] Osama Ouda, Norimichi Tsumura, and Toshiya Nakaguchi. Bioen-
coding: A reliable tokenless cancelable biometrics scheme for protect-
ing iriscodes. IEICE TRANSACTIONS on Information and Systems,
93(7):1878–1888, 2010.
[478] Mark D Ryan. Cloud computing privacy concerns on our doorstep.
Communications of the ACM, 54(1):36–38, 2011.
[479] Alfredo Cuzzocrea. Privacy and security of big data: current challenges
and future research perspectives. In Proceedings of the first international
workshop on privacy and secuirty of big data, pages 45–47, 2014.
[480] Yuhong Liu, Yan Lindsay Sun, Jungwoo Ryoo, Syed Rizvi, and
Athanasios V Vasilakos. A survey of security and privacy challenges in
cloud computing: solutions and future directions. Journal of Computing
Science and Engineering, 9(3):119–133, 2015.
[481] Muhammad Baqer Mollah, Md Abul Kalam Azad, and Athanasios
Vasilakos. Security and privacy challenges in mobile cloud computing:
Survey and way ahead. Journal of Network and Computer Applications,
84:38–54, 2017.
[482] Keith Bonawitz, Hubert Eichner, Wolfgang Grieskamp, Dzmitry Huba,
Alex Ingerman, Vladimir Ivanov, Chloe Kiddon, Jakub Koneˇ
cn`
y, Stefano
Mazzocchi, H Brendan McMahan, et al. Towards federated learning at
scale: System design. arXiv preprint arXiv:1902.01046, 2019.
[483] Jiale Zhang, Bing Chen, Yanchao Zhao, Xiang Cheng, and Feng Hu.
Data security and privacy-preserving in edge computing paradigm: Survey
and open issues. IEEE access, 6:18209–18237, 2018.
[484] Yahoo. How big is the metaverse?, July 2015.
[485] Pushkara Ravindra, Aakash Khochare, Siva Prakash Reddy, Sarthak
Sharma, Prateeksha Varshney, and Yogesh Simmhan. Echo: An adaptive
orchestration platform for hybrid dataflows across cloud and edge. In
International Conference on Service-Oriented Computing, pages 395–
410. Springer, 2017.
[486] Lorenzo Carnevale, Antonio Celesti, Antonino Galletta, Schahram
Dustdar, and Massimo Villari. From the cloud to edge and iot: a
smart orchestration architecture for enabling osmotic computing. In 2018
32nd International Conference on Advanced Information Networking and
Applications Workshops (WAINA), pages 419–424. IEEE, 2018.
[487] Yulei Wu. Cloud-edge orchestration for the internet-of-things: Archi-
tecture and ai-powered data processing. IEEE Internet of Things Journal,
2020.
[488] Shikhar Suryavansh, Chandan Bothra, Mung Chiang, Chunyi Peng,
and Saurabh Bagchi. Tango of edge and cloud execution for reliability.
In Proceedings of the 4th Workshop on Middleware for Edge Clouds &
Cloudlets, pages 10–15, 2019.
[489] Ayman Younis, Brian Qiu, and Dario Pompili. Latency-aware hybrid
edge cloud framework for mobile augmented reality applications. In 2020
17th Annual IEEE International Conference on Sensing, Communication,
and Networking (SECON), pages 1–9. IEEE, 2020.
[490] Wuyang Zhang, Jiachen Chen, Yanyong Zhang, and Dipankar Ray-
chaudhuri. Towards efficient edge cloud augmentation for virtual reality
mmogs. In Proceedings of the Second ACM/IEEE Symposium on Edge
Computing, pages 1–14, 2017.
[491] Jingbo Zhao, Robert S Allison, Margarita Vinnikov, and Sion Jennings.
Estimating the motion-to-photon latency in head mounted displays. In
2017 IEEE Virtual Reality (VR), pages 313–314. IEEE, 2017.
[492] NGMN Alliance. 5g white paper. Next generation mobile networks,
white paper, 1, 2015.
[493] Sidi Lu, Yongtao Yao, and Weisong Shi. Collaborative learning on the
edges: A case study on connected vehicles. In 2nd {USENIX}Workshop
on Hot Topics in Edge Computing (HotEdge 19), 2019.
[494] Ulrich Lampe, Qiong Wu, Sheip Dargutev, Ronny Hans, Andr´
e Miede,
and Ralf Steinmetz. Assessing latency in cloud gaming. In International
Conference on Cloud Computing and Services Science, pages 52–68.
Springer, 2013.
[495] Zenja Ivkovic, Ian Stavness, Carl Gutwin, and Steven Sutcliffe. Quan-
tifying and mitigating the negative effects of local latencies on aiming in
3d shooter games. In Proceedings of the 33rd Annual ACM Conference
on Human Factors in Computing Systems, CHI ’15, page 135–144, New
York, NY, USA, 2015. Association for Computing Machinery.
[496] Peter Lincoln, Alex Blate, Montek Singh, Turner Whitted, Andrei
State, Anselmo Lastra, and Henry Fuchs. From motion to photons in 80
microseconds: Towards minimal latency for virtual and augmented reality.
IEEE transactions on visualization and computer graphics, 22(4):1367–
1376, 2016.
[497] BJ Challacombe, LR Kavoussi, and P Dasgupta. Trans-oceanic teler-
obotic surgery. BJU international (Papier), 92(7):678–680, 2003.
[498] Teemu K¨
am¨
ar¨
ainen, Matti Siekkinen, Antti Yl¨
a-J¨
a¨
aski, Wenxiao
Zhang, and Pan Hui. A measurement study on achieving imperceptible
latency in mobile cloud gaming. In Proceedings of the 8th ACM on
Multimedia Systems Conference, MMSys’17, page 88–99, New York, NY,
USA, 2017. Association for Computing Machinery.
[499] time(7) Linux User’s Manual.
[500] Joel Hestness, Stephen W Keckler, and David A Wood. Gpu computing
pipeline inefficiencies and optimization opportunities in heterogeneous
cpu-gpu processors. In 2015 IEEE International Symposium on Workload
Characterization, pages 87–97. IEEE, 2015.
[501] Dongzhu Xu, Anfu Zhou, Xinyu Zhang, Guixian Wang, Xi Liu,
Congkai An, Yiming Shi, Liang Liu, and Huadong Ma. Understanding
operational 5g: A first measurement study on its coverage, performance
and energy consumption. In Proceedings of the Annual conference of the
ACM Special Interest Group on Data Communication on the applications,
technologies, architectures, and protocols for computer communication,
pages 479–494, 2020.
[502] 3GPP. Study on scenarios and requirements for next generation access
technologies. Technical report, 2018.
[503] Lorenzo Corneo, Maximilian Eder, Nitinder Mohan, Aleksandr Za-
vodovski, Suzan Bayhan, Walter Wong, Per Gunningberg, Jussi Kan-
gasharju, and J¨
org Ott. Surrounded by the clouds: A comprehensive
cloud reachability study. In Proceedings of the Web Conference 2021,
WWW ’21, page 295–304, New York, NY, USA, 2021. Association for
Computing Machinery.
[504] Lorenzo Corneo, Nitinder Mohan, Aleksandr Zavodovski, Walter
Wong, Christian Rohner, Per Gunningberg, and Jussi Kangasharju. (how
much) can edge computing change network latency? In 2021 IFIP
Networking Conference (IFIP Networking), pages 1–9, 2021.
[505] Mostafa Ammar, Ellen Zegura, and Yimeng Zhao. A vision for zero-
hop networking (zen). In 2017 IEEE 37th International Conference on
Distributed Computing Systems (ICDCS), pages 1765–1770. IEEE, 2017.
[506] Khaled Diab and Mohamed Hefeeda. Joint content distribution and
traffic engineering of adaptive videos in telco-cdns. In IEEE INFOCOM
2019 - IEEE Conference on Computer Communications, pages 1342–
1350, 2019.
[507] Mehdi Bennis, M ´
erouane Debbah, and H. Vincent Poor. Ultrareliable
and low-latency wireless communication: Tail, risk, and scale. Proceed-
ings of the IEEE, 106(10):1834–1853, 2018.
[508] Afif Osseiran, Jose F Monserrat, and Patrick Marsch. 5G mobile and
wireless communications technology. Cambridge University Press, 2016.
[509] Nurul Huda Mahmood, Stefan B ¨
ocker, Andrea Munari, Federico
Clazzer, Ingrid Moerman, Konstantin Mikhaylov, Onel Lopez, Ok-Sun
Park, Eric Mercier, Hannes Bartz, et al. White paper on critical

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 60
and massive machine type communication towards 6g. arXiv preprint
arXiv:2004.14146, 2020.
[510] Tristan Braud, Dimitris Chatzopoulos, and Pan Hui. Machine type
communications in 6g. In 6G Mobile Wireless Networks, pages 207–231.
Springer, 2021.
[511] NGMN Alliance. Description of network slicing concept. NGMN 5G
P, 1(1), 2016.
[512] Marko H ¨
oyhty¨
a, Kalle L¨
ahetkangas, Jani Suomalainen, Mika Hoppari,
Kaisa Kujanp¨
a¨
a, Kien Trung Ngo, Tero Kippola, Marjo Heikkil¨
a, Harri
Posti, Jari M¨
aki, Tapio Savunen, Ari Hulkkonen, and Heikki Kokkinen.
Critical communications over mobile operators’ networks: 5g use cases
enabled by licensed spectrum sharing, network slicing and qos control.
IEEE Access, 6:73572–73582, 2018.
[513] Claudia Campolo, Antonella Molinaro, Antonio Iera, and Francesco
Menichella. 5g network slicing for vehicle-to-everything services. IEEE
Wireless Communications, 24(6):38–45, 2017.
[514] Tarik Taleb, Ibrahim Afolabi, Konstantinos Samdanis, and Faqir Zarrar
Yousaf. On multi-domain network slicing orchestration architecture and
federated resource control. IEEE Network, 33(5):242–252, 2019.
[515] Kei Sakaguchi, Thomas Haustein, Sergio Barbarossa, Emilio Calvanese
Strinati, Antonio Clemente, Giuseppe Destino, Aarno P¨
arssinen, Ilgyu
Kim, Heesang Chung, Junhyeong Kim, et al. Where, when, and how
mmwave is used in 5g and beyond. IEICE Transactions on Electronics,
100(10):790–808, 2017.
[516] Kjell Brunnstr ¨
om, Sergio Ariel Beker, Katrien De Moor, Ann Dooms,
Sebastian Egger, Marie-Neige Garcia, Tobias Hossfeld, Satu Jumisko-
Pyykk¨
o, Christian Keimel, Mohamed-Chaker Larabi, et al. Qualinet white
paper on definitions of quality of experience. 2013.
[517] Eirini Liotou, Dimitris Tsolkas, Nikos Passas, and Lazaros Merakos.
Quality of experience management in mobile cellular networks: key issues
and design challenges. IEEE Communications Magazine, 53(7):145–153,
2015.
[518] Mukundan Venkataraman and Mainak Chatterjee. Inferring video qoe
in real time. IEEE Network, 25(1):4–13, 2011.
[519] Yanjiao Chen, Kaishun Wu, and Qian Zhang. From qos to qoe: A
tutorial on video quality assessment. IEEE Communications Surveys &
Tutorials, 17(2):1126–1165, 2014.
[520] Sabina Barakovi´
c, Jasmina Barakovi´
c, and Himzo Bajri´
c. Qoe dimen-
sions and qoe measurement of ngn services. In Proceedings of the 18th
Telecommunications Forum, TELFOR 2010, 2010.
[521] Tristan Braud, Farshid Hassani Bijarbooneh, Dimitris Chatzopoulos,
and Pan Hui. Future networking challenges: The case of mobile
augmented reality. In 2017 IEEE 37th International Conference on
Distributed Computing Systems (ICDCS), pages 1796–1807. IEEE, 2017.
[522] Mohammad A. Hoque, Ashwin Rao, Abhishek Kumar, Mostafa Am-
mar, Pan Hui, and Sasu Tarkoma. Sensing multimedia contexts on mobile
devices. In Proceedings of the 30th ACM Workshop on Network and
Operating Systems Support for Digital Audio and Video, NOSSDAV ’20,
page 40–46, New York, NY, USA, 2020. Association for Computing
Machinery.
[523] Yanyuan Qin, Shuai Hao, Krishna R. Pattipati, Feng Qian, Subhabrata
Sen, Bing Wang, and Chaoqun Yue. Quality-aware strategies for opti-
mizing abr video streaming qoe and reducing data usage. In Proceedings
of the 10th ACM Multimedia Systems Conference, MMSys ’19, page
189–200, New York, NY, USA, 2019. Association for Computing Ma-
chinery.
[524] Lingyan Zhang, Shangguang Wang, and Rong N. Chang. Qcss: A
qoe-aware control plane for adaptive streaming service over mobile edge
computing infrastructures. In 2018 IEEE International Conference on
Web Services (ICWS), pages 139–146, 2018.
[525] Maroua Ben Attia, Kim-Khoa Nguyen, and Mohamed Cheriet. Dy-
namic qoe/qos-aware queuing for heterogeneous traffic in smart home.
IEEE Access, 7:58990–59001, 2019.
[526] Lukas Sevcik, Miroslav Voznak, and Jaroslav Frnda. Qoe prediction
model for multimedia services in ip network applying queuing policy. In
International Symposium on Performance Evaluation of Computer and
Telecommunication Systems (SPECTS 2014), pages 593–598, 2014.
[527] Eirini Liotou, Konstantinos Samdanis, Emmanouil Pateromichelakis,
Nikos Passas, and Lazaros Merakos. Qoe-sdn app: A rate-guided qoe-
aware sdn-app for http adaptive video streaming. IEEE Journal on
Selected Areas in Communications, 36(3):598–615, 2018.
[528] Faqir Zarrar Yousaf, Marco Gramaglia, Vasilis Friderikos, Borislava
Gajic, Dirk Von Hugo, Bessem Sayadi, Vincenzo Sciancalepore, and
Marcos Rates Crippa. Network slicing with flexible mobility and qos/qoe
support for 5g networks. In 2017 IEEE International Conference on
Communications Workshops (ICC Workshops), pages 1195–1201. IEEE,
2017.
[529] Marilynn P Wylie-Green and Tommy Svensson. Throughput, capacity,
handover and latency performance in a 3gpp lte fdd field trial. In 2010
IEEE Global Telecommunications Conference GLOBECOM 2010, pages
1–6. IEEE, 2010.
[530] Johanna Heinonen, Pekka Korja, Tapio Partti, Hannu Flinck, and Petteri
P¨
oyh¨
onen. Mobility management enhancements for 5g low latency
services. In 2016 IEEE International Conference on Communications
Workshops (ICC), pages 68–73. IEEE, 2016.
[531] M¨
uge Erel- ¨
Ozc¸evik and Berk Canberk. Road to 5g reduced-latency: A
software defined handover model for embb services. IEEE Transactions
on Vehicular Technology, 68(8):8133–8144, 2019.
[532] Tristan Braud, Teemu K¨
am¨
ar¨
ainen, Matti Siekkinen, and Pan Hui.
Multi-carrier measurement study of mobile network latency: The tale
of hong kong and helsinki. In 2019 15th International Conference on
Mobile Ad-Hoc and Sensor Networks (MSN), pages 1–6, 2019.
[533] Gregory J Pottie. Wireless sensor networks. In 1998 Information
Theory Workshop (Cat. No. 98EX131), pages 139–140. IEEE, 1998.
[534] Enrico Natalizio and Valeria Loscr´
ı. Controlled mobility in mobile
sensor networks: advantages, issues and challenges. Telecommunication
Systems, 52(4):2411–2418, 2013.
[535] Sukhchandan Randhawa and Sushma Jain. Data aggregation in wireless
sensor networks: Previous research, current status and future directions.
Wireless Personal Communications, 97(3):3355–3425, 2017.
[536] Nancy Miller and Peter Steenkiste. Collecting network status infor-
mation for network-aware applications. In Proceedings IEEE INFOCOM
2000. Conference on Computer Communications. Nineteenth Annual Joint
Conference of the IEEE Computer and Communications Societies (Cat.
No. 00CH37064), volume 2, pages 641–650. IEEE, 2000.
[537] J¨
urg Bolliger and Thomas Gross. A framework based approach to
the development of network aware applications. IEEE transactions on
Software Engineering, 24(5):376–390, 1998.
[538] Jinwei Cao, K.M. McNeill, Dongsong Zhang, and J.F. Nunamaker. An
overview of network-aware applications for mobile multimedia delivery.
In 37th Annual Hawaii International Conference on System Sciences,
2004. Proceedings of the, pages 10 pp.–, 2004.
[539] Jose Santos, Tim Wauters, Bruno Volckaert, and Filip De Turck.
Towards network-aware resource provisioning in kubernetes for fog com-
puting applications. In 2019 IEEE Conference on Network Softwarization
(NetSoft), pages 351–359. IEEE, 2019.
[540] Su Wang, Yichen Ruan, Yuwei Tu, Satyavrat Wagle, Christopher G
Brinton, and Carlee Joe-Wong. Network-aware optimization of distributed
learning for fog computing. IEEE/ACM Transactions on Networking,
2021.
[541] Fan Jiang, Claris Castillo, and Stan Ahalt. Cachalot: A network-aware,
cooperative cache network for geo-distributed, data-intensive applications.
In NOMS 2018-2018 IEEE/IFIP Network Operations and Management
Symposium, pages 1–9. IEEE, 2018.
[542] Jingxuan Zhang, Luis Contreras, Kai Gao, Francisco Cano, Patri-
cia Cano, Anais Escribano, and Y Richard Yang. Sextant: Enabling
automated network-aware application optimization in carrier networks.
In 2021 IFIP/IEEE International Symposium on Integrated Network
Management (IM), pages 586–593. IEEE, 2021.
[543] Chunshan Xiong, Yunfei Zhang, Richard Yang, Gang Li, Yixue Lei,
and Yunbo Han. MoWIE for Network Aware Application. Internet-Draft
draft-huang-alto-mowie-for-network-aware-app-03, Internet Engineering
Task Force, July 2021. Work in Progress.
[544] Xipeng Zhu, Ruiming Zheng, Dacheng Yang, Huichun Liu, and Jilei
Hou. Radio-aware tcp optimization in mobile network. In 2017 IEEE
Wireless Communications and Networking Conference (WCNC), pages
1–5. IEEE, 2017.
[545] Eman Ramadan, Arvind Narayanan, Udhaya Kumar Dayalan, Ros-
tand AK Fezeu, Feng Qian, and Zhi-Li Zhang. Case for 5g-aware
video streaming applications. In Proceedings of the 1st Workshop on
5G Measurements, Modeling, and Use Cases, pages 27–34, 2021.
[546] Anya Kolesnichenko, Joshua McVeigh-Schultz, and Katherine Isbis-
ter. Understanding emerging design practices for avatar systems in the
commercial social vr ecology. In Proceedings of the 2019 on Designing
Interactive Systems Conference, DIS ’19, page 241–252, New York, NY,
USA, 2019. Association for Computing Machinery.
[547] Klaus Fuchs, Daniel Meusburger, Mirella Haldimann, and Alexander
Ilic. Nutritionavatar: Designing a future-self avatar for promotion of
balanced, low-sodium diet intention: Framework design and user study.
In Proceedings of the 13th Biannual Conference of the Italian SIGCHI
Chapter: Designing the next Interaction, CHItaly ’19, New York, NY,
USA, 2019. Association for Computing Machinery.
[548] Konstantinos Tsiakas, Deborah Cnossen, Tim H.C. Muyrers,
Danique R.C. Stappers, Romain H.A. Toebosch, and Emilia Barakova.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 61
Futureme: Negotiating learning goals with your future learning-self avatar.
In The 14th PErvasive Technologies Related to Assistive Environments
Conference, PETRA 2021, page 262–263, New York, NY, USA, 2021.
Association for Computing Machinery.
[549] Cherie Lacey and Catherine Caudwell. Cuteness as a ’dark pattern’
in home robots. In Proceedings of the 14th ACM/IEEE International
Conference on Human-Robot Interaction, HRI ’19, page 374–381. IEEE
Press, 2019.
[550] Ana Paiva, Iolanda Leite, Hana Boukricha, and Ipke Wachsmuth.
Empathy in virtual agents and robots: A survey. ACM Trans. Interact.
Intell. Syst., 7(3), September 2017.
[551] Kazuaki Takeuchi, Yoichi Yamazaki, and Kentaro Yoshifuji. Avatar
work: Telework for disabled people unable to go outside by using avatar
robots. In Companion of the 2020 ACM/IEEE International Conference
on Human-Robot Interaction, HRI ’20, page 53–60, New York, NY, USA,
2020. Association for Computing Machinery.
[552] Marc Erich Latoschik, Daniel Roth, Dominik Gall, Jascha Achenbach,
Thomas Waltemate, and Mario Botsch. The effect of avatar realism
in immersive social virtual realities. In Proceedings of the 23rd ACM
Symposium on Virtual Reality Software and Technology, VRST ’17, New
York, NY, USA, 2017. Association for Computing Machinery.
[553] Martin Kocur, Sarah Graf, and Valentin Schwind. The impact of
missing fingers in virtual reality. In 26th ACM Symposium on Virtual
Reality Software and Technology, VRST ’20, New York, NY, USA, 2020.
Association for Computing Machinery.
[554] Gordon Brown and Michael Prilla. The effects of consultant avatar size
and dynamics on customer trust in online consultations. In Proceedings
of the Conference on Mensch Und Computer, MuC ’20, page 239–249,
New York, NY, USA, 2020. Association for Computing Machinery.
[555] Guo Freeman and Divine Maloney. Body, avatar, and me: The
presentation and perception of self in social virtual reality. Proc. ACM
Hum.-Comput. Interact., 4(CSCW3), January 2021.
[556] Rabindra Ratan and B ´
eatrice S. Hasler. Playing well with virtual
classmates: Relating avatar design to group satisfaction. In Proceedings
of the 17th ACM Conference on Computer Supported Cooperative Work
&; Social Computing, CSCW ’14, page 564–573, New York, NY, USA,
2014. Association for Computing Machinery.
[557] Xiaozhou Wei, Lijun Yin, Zhiwei Zhu, and Qiang Ji. Avatar-
mediated face tracking and lip reading for human computer interaction.
In Proceedings of the 12th Annual ACM International Conference on
Multimedia, MULTIMEDIA ’04, page 500–503, New York, NY, USA,
2004. Association for Computing Machinery.
[558] Dooley Murphy. Building a hybrid virtual agent for testing user
empathy and arousal in response to avatar (micro-)expressions. In
Proceedings of the 23rd ACM Symposium on Virtual Reality Software
and Technology, VRST ’17, New York, NY, USA, 2017. Association for
Computing Machinery.
[559] Heike Brock, Shigeaki Nishina, and Kazuhiro Nakadai. To animate or
anime-te? investigating sign avatar comprehensibility. In Proceedings of
the 18th International Conference on Intelligent Virtual Agents, IVA ’18,
page 331–332, New York, NY, USA, 2018. Association for Computing
Machinery.
[560] Marc Erich Latoschik, Daniel Roth, Dominik Gall, Jascha Achenbach,
Thomas Waltemate, and Mario Botsch. The effect of avatar realism
in immersive social virtual realities. In Proceedings of the 23rd ACM
Symposium on Virtual Reality Software and Technology, VRST ’17, New
York, NY, USA, 2017. Association for Computing Machinery.
[561] Dominic Kao and D. Fox Harrell. Exploring the impact of avatar color
on game experience in educational games. In Proceedings of the 2016
CHI Conference Extended Abstracts on Human Factors in Computing
Systems, CHI EA ’16, page 1896–1905, New York, NY, USA, 2016.
Association for Computing Machinery.
[562] Jean-Luc Lugrin, Ivan Polyschev, Daniel Roth, and Marc Erich
Latoschik. Avatar anthropomorphism and acrophobia. In Proceedings of
the 22nd ACM Conference on Virtual Reality Software and Technology,
VRST ’16, page 315–316, New York, NY, USA, 2016. Association for
Computing Machinery.
[563] Chang Yun, Zhigang Deng, and Merrill Hiscock. Can local avatars
satisfy a global audience? a case study of high-fidelity 3d facial avatar
animation in subject identification and emotion perception by us and
international groups. Comput. Entertain., 7(2), June 2009.
[564] Florian Mathis, Kami Vaniea, and Mohamed Khamis. Observing virtual
avatars: The impact of avatars’ fidelity on identifying interactions. In
Academic Mindtrek 2021, Mindtrek 2021, page 154–164, New York, NY,
USA, 2021. Association for Computing Machinery.
[565] Yutaka Ishii, Tomio Watanabe, and Yoshihiro Sejima. Development of
an embodied avatar system using avatar-shadow’s color expressions with
an interaction-activated communication model. In Proceedings of the
Fourth International Conference on Human Agent Interaction, HAI ’16,
page 337–340, New York, NY, USA, 2016. Association for Computing
Machinery.
[566] Changyeol Choi, Joohee Jun, Jiwoong Heo, and Kwanguk (Kenny)
Kim. Effects of virtual-avatar motion-synchrony levels on full-body
interaction. In Proceedings of the 34th ACM/SIGAPP Symposium on
Applied Computing, SAC ’19, page 701–708, New York, NY, USA, 2019.
Association for Computing Machinery.
[567] Juyoung Lee, Myungho Lee, Gerard Jounghyun Kim, and Jae-In
Hwang. Effects of synchronized leg motion in walk-in-place utilizing
deep neural networks for enhanced body ownership and sense of presence
in vr. In 26th ACM Symposium on Virtual Reality Software and
Technology, VRST ’20, New York, NY, USA, 2020. Association for
Computing Machinery.
[568] Anne Thaler, Anna C. Wellerdiek, Markus Leyrer, Ekaterina Volkova-
Volkmar, Nikolaus F. Troje, and Betty J. Mohler. The role of avatar
fidelity and sex on self-motion recognition. In Proceedings of the 15th
ACM Symposium on Applied Perception, SAP ’18, New York, NY, USA,
2018. Association for Computing Machinery.
[569] Robert J. Moore, E. Cabell Hankinson Gathman, Nicolas Ducheneaut,
and Eric Nickell. Coordinating Joint Activity in Avatar-Mediated Inter-
action, page 21–30. Association for Computing Machinery, New York,
NY, USA, 2007.
[570] Myoung Ju Won, Sangin Park, SungTeac Hwang, and Mincheol
Whang. Development of realistic digital expression of human avatars
through pupillary responses based on heart rate. In Proceedings of the
33rd Annual ACM Conference Extended Abstracts on Human Factors in
Computing Systems, CHI EA ’15, page 287–290, New York, NY, USA,
2015. Association for Computing Machinery.
[571] Mark L. Knapp. Nonverbal communication in human interaction. 1972.
[572] Anna Samira Praetorius, Lara Krautmacher, Gabriela Tullius, and
Crist´
obal Curio. User-avatar relationships in various contexts: Does
context influence a users’ perception and choice of an avatar? In Mensch
Und Computer 2021, MuC ’21, page 275–280, New York, NY, USA,
2021. Association for Computing Machinery.
[573] N. Yee and J. Bailenson. The proteus effect: The effect of trans-
formed self-representation on behavior. Human Communication Research,
33:271–290, 2007.
[574] Anna Samira Praetorius and Daniel G ¨
orlich. How Avatars Influence
User Behavior: A Review on the Proteus Effect in Virtual Environments
and Video Games. Association for Computing Machinery, New York,
NY, USA, 2020.
[575] Yifang Li, Nishant Vishwamitra, Bart P. Knijnenburg, Hongxin Hu,
and Kelly Caine. Effectiveness and users’ experience of obfuscation as
a privacy-enhancing technology for sharing photos. Proc. ACM Hum.-
Comput. Interact., 1(CSCW), December 2017.
[576] Divine Maloney. Mitigating negative effects of immersive virtual
avatars on racial bias. In Proceedings of the 2018 Annual Symposium
on Computer-Human Interaction in Play Companion Extended Abstracts,
CHI PLAY ’18 Extended Abstracts, page 39–43, New York, NY, USA,
2018. Association for Computing Machinery.
[577] Erica L. Neely. No player is ideal: Why video game designers cannot
ethically ignore players’ real-world identities. SIGCAS Comput. Soc.,
47(3):98–111, September 2017.
[578] Kangsoo Kim, Gerd Bruder, and Greg Welch. Exploring the effects
of observed physicality conflicts on real-virtual human interaction in
augmented reality. In Proceedings of the 23rd ACM Symposium on Virtual
Reality Software and Technology, VRST ’17, New York, NY, USA, 2017.
Association for Computing Machinery.
[579] Arjun Nagendran, Remo Pillat, Charles Hughes, and Greg Welch.
Continuum of virtual-human space: Towards improved interaction strate-
gies for physical-virtual avatars. In Proceedings of the 11th ACM
SIGGRAPH International Conference on Virtual-Reality Continuum and
Its Applications in Industry, VRCAI ’12, page 135–142, New York, NY,
USA, 2012. Association for Computing Machinery.
[580] Lijuan Zhang and Steve Oney. Flowmatic: An immersive authoring tool
for creating interactive scenes in virtual reality. Proceedings of the 33rd
Annual ACM Symposium on User Interface Software and Technology,
2020.
[581] R. Horst and R. D ¨
orner. Virtual reality forge: Pattern-oriented authoring
of virtual reality nuggets. 25th ACM Symposium on Virtual Reality
Software and Technology, 2019.
[582] Larry Cutler, Amy Tucker, R. Schiewe, Justin Fischer, Nathaniel
Dirksen, and Eric Darnell. Authoring interactive vr narratives on baba
yaga and bonfire. Special Interest Group on Computer Graphics and
Interactive Techniques Conference Talks, 2020.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 62
[583] Arnaud Prouzeau, Yuchen Wang, Barrett Ens, Wesley Willett, and
T. Dwyer. Corsican twin: Authoring in situ augmented reality visuali-
sations in virtual reality. Proceedings of the International Conference on
Advanced Visual Interfaces, 2020.
[584] Danilo Gasques, Janet G. Johnson, Tommy Sharkey, and Nadir Weibel.
Pintar: Sketching spatial experiences in augmented reality. Companion
Publication of the 2019 on Designing Interactive Systems Conference
2019 Companion, 2019.
[585] Henning Pohl, Tor-Salve Dalsgaard, Vesa Krasniqi, and Kasper Horn-
bæk. Body layars: A toolkit for body-based augmented reality. 26th ACM
Symposium on Virtual Reality Software and Technology, 2020.
[586] Maximilian Speicher, Katy Lewis, and Michael Nebeling. Designers,
the stage is yours! medium-fidelity prototyping of augmented & virtual
reality interfaces with 360theater. Proceedings of the ACM on Human-
Computer Interaction, 5:1 – 25, 2021.
[587] Germ´
an Leiva, Cuong Nguyen, R. Kazi, and Paul Asente. Pronto:
Rapid augmented reality video prototyping using sketches and enaction.
Proceedings of the 2020 CHI Conference on Human Factors in Comput-
ing Systems, 2020.
[588] Michael Nebeling, Janet Nebeling, Ao Yu, and Rob Rumble. Protoar:
Rapid physical-digital prototyping of mobile augmented reality applica-
tions. Proceedings of the 2018 CHI Conference on Human Factors in
Computing Systems, 2018.
[589] Subramanian Chidambaram, Hank Huang, Fengming He, Xun Qian,
Ana M. Villanueva, Thomas Redick, W. Stuerzlinger, and K. Ramani.
Processar: An augmented reality-based tool to create in-situ procedural
2d/3d ar instructions. Designing Interactive Systems Conference 2021,
2021.
[590] Leon M ¨
uller, Ken Pfeuffer, Jan Gugenheimer, Bastian Pfleging, Sarah
Prange, and Florian Alt. Spatialproto: Exploring real-world motion
captures for rapid prototyping of interactive mixed reality. Proceedings
of the 2021 CHI Conference on Human Factors in Computing Systems,
2021.
[591] Michael Nebeling and Katy Madier. 360proto: Making interactive
virtual reality & augmented reality prototypes from paper. Proceedings
of the 2019 CHI Conference on Human Factors in Computing Systems,
2019.
[592] Gabriel Freitas, M. Pinho, M. Silveira, and F. Maurer. A systematic
review of rapid prototyping tools for augmented reality. 2020 22nd
Symposium on Virtual and Augmented Reality (SVR), pages 199–209,
2020.
[593] Narges Ashtari, Andrea Bunt, J. McGrenere, Michael Nebeling, and
Parmit K. Chilana. Creating augmented and virtual reality applications:
Current practices, challenges, and opportunities. Proceedings of the 2020
CHI Conference on Human Factors in Computing Systems, 2020.
[594] Veronika Krauß, A. Boden, Leif Oppermann, and Ren´
e Reiners.
Current practices, challenges, and design implications for collaborative
ar/vr application development. Proceedings of the 2021 CHI Conference
on Human Factors in Computing Systems, 2021.
[595] Maximilian Speicher, Brian D. Hall, Ao Yu, Bowen Zhang, Haihua
Zhang, Janet Nebeling, and Michael Nebeling. Xd-ar: Challenges and
opportunities in cross-device augmented reality application development.
Proc. ACM Hum.-Comput. Interact., 2(EICS), June 2018.
[596] Michael Nebeling, Shwetha Rajaram, Liwei Wu, Yifei Cheng, and
Jaylin Herskovitz. Xrstudio: A virtual production and live streaming
system for immersive instructional experiences. In Proceedings of the
2021 CHI Conference on Human Factors in Computing Systems, CHI
’21, New York, NY, USA, 2021. Association for Computing Machinery.
[597] Peng Wang, Xiaoliang Bai, Mark Billinghurst, Shusheng Zhang, Xi-
angyu Zhang, Shuxia Wang, Weiping He, Yuxiang Yan, and Hongyu Ji.
Ar/mr remote collaboration on physical tasks: A review. Robotics and
Computer-Integrated Manufacturing, 72:102071, 2021.
[598] Zhenyi He, Ruofei Du, and K. Perlin. Collabovr: A reconfigurable
framework for creative collaboration in virtual reality. 2020 IEEE
International Symposium on Mixed and Augmented Reality (ISMAR),
pages 542–554, 2020.
[599] Chiwon Lee, Hyunjong Joo, and Soojin Jun. Social vr as the new
normal? understanding user interactions for the business arena. In
Extended Abstracts of the 2021 CHI Conference on Human Factors in
Computing Systems, CHI EA ’21, New York, NY, USA, 2021. Association
for Computing Machinery.
[600] Daniel Sarkady, Larissa Neuburger, and R. Egger. Virtual reality as a
travel substitution tool during covid-19. Information and Communication
Technologies in Tourism 2021, pages 452 – 463, 2020.
[601] Rainer Winkler, Sebastian Hobert, Antti Salovaara, Matthias S¨
ollner,
and Jan Marco Leimeister. Sara, the lecturer: Improving learning in online
education with a scaffolding-based conversational agent. In Proceedings
of the 2020 CHI Conference on Human Factors in Computing Systems,
CHI ’20, page 1–14, New York, NY, USA, 2020. Association for
Computing Machinery.
[602] Philip Weber, Thomas Ludwig, Sabrina Brodesser, and Laura
Gr¨
onewald. “it’s a kind of art!”: Understanding food influencers as
influential content creators. In Proceedings of the 2021 CHI Conference
on Human Factors in Computing Systems, CHI ’21, New York, NY, USA,
2021. Association for Computing Machinery.
[603] Abdelberi Chaabane, Terence Chen, Mathieu Cunche, Emiliano
De Cristofaro, Arik Friedman, and Mohamed Ali Kaafar. Censorship
in the wild: Analyzing internet filtering in syria. In Proceedings of
the 2014 Conference on Internet Measurement Conference, IMC ’14,
page 285–298, New York, NY, USA, 2014. Association for Computing
Machinery.
[604] Jong Chun Park and Jedidiah R. Crandall. Empirical study of
a national-scale distributed intrusion detection system: Backbone-level
filtering of html responses in china. In Proceedings of the 2010 IEEE
30th International Conference on Distributed Computing Systems, ICDCS
’10, page 315–326, USA, 2010. IEEE Computer Society.
[605] Simurgh Aryan, Homa Aryan, and J. Alex Halderman. Internet
censorship in iran: A first look. In 3rd USENIX Workshop on Free and
Open Communications on the Internet (FOCI 13), Washington, D.C.,
August 2013. USENIX Association.
[606] Ehsan ul Haq, Tristan Braud, Young D. Kwon, and Pan Hui. Enemy at
the gate: Evolution of twitter user’s polarization during national crisis.
In 2020 IEEE/ACM International Conference on Advances in Social
Networks Analysis and Mining (ASONAM), pages 212–216, 2020.
[607] Zubair Nabi. The anatomy of web censorship in pakistan. In 3rd
USENIX Workshop on Free and Open Communications on the Internet
(FOCI 13), Washington, D.C., August 2013. USENIX Association.
[608] Jakub Dalek, Bennett Haselton, Helmi Noman, Adam Senft, Masashi
Crete-Nishihata, Phillipa Gill, and Ronald J. Deibert. A method for
identifying and confirming the use of url filtering products for censorship.
In Proceedings of the 2013 Conference on Internet Measurement Confer-
ence, IMC ’13, page 23–30, New York, NY, USA, 2013. Association for
Computing Machinery.
[609] Ram Sundara Raman, Prerana Shenoy, Katharina Kohls, and Roya
Ensafi. Censored planet: An internet-wide, longitudinal censorship
observatory. In Proceedings of the 2020 ACM SIGSAC Conference on
Computer and Communications Security, CCS ’20, page 49–66, New
York, NY, USA, 2020. Association for Computing Machinery.
[610] Divine Maloney, Guo Freeman, and Donghee Yvette Wohn. ”talking
without a voice”: Understanding non-verbal communication in social
virtual reality. Proc. ACM Hum.-Comput. Interact., 4(CSCW2), October
2020.
[611] Peter L. Stanchev, Desislava Paneva-Marinova, and Alexander Iliev.
Enhanced user experience and behavioral patterns for digital cultural
ecosystems. In Proceedings of the 9th International Conference on
Management of Digital EcoSystems, MEDES ’17, page 287–292, New
York, NY, USA, 2017. Association for Computing Machinery.
[612] Bokyung Lee, Gyeol Han, Jundong Park, and Daniel Saakes. Consumer
to Creator: How Households Buy Furniture to Inform Design and Fabri-
cation Interfaces, page 484–496. Association for Computing Machinery,
New York, NY, USA, 2017.
[613] Josh Urban Davis, Fraser Anderson, Merten Stroetzel, Tovi Grossman,
and George Fitzmaurice. Designing co-creative ai for virtual environ-
ments. In Creativity and Cognition, C&C ’21, New York, NY, USA,
2021. Association for Computing Machinery.
[614] David A. Shamma and Daragh Bryne. An introduction to arts and
digital culture inside multimedia. In Proceedings of the 23rd ACM
International Conference on Multimedia, MM ’15, page 1329–1330, New
York, NY, USA, 2015. Association for Computing Machinery.
[615] Samer Abdallah, Emmanouil Benetos, Nicolas Gold, Steven Harg-
reaves, Tillman Weyde, and Daniel Wolff. The digital music lab: A big
data infrastructure for digital musicology. J. Comput. Cult. Herit., 10(1),
January 2017.
[616] Mauro Dragoni, Sara Tonelli, and Giovanni Moretti. A knowledge
management architecture for digital cultural heritage. J. Comput. Cult.
Herit., 10(3), July 2017.
[617] Nicola Barbuti. From digital cultural heritage to digital culture:
Evolution in digital humanities. In Proceedings of the 1st International
Conference on Digital Tools & Uses Congress, DTUC ’18, New York,
NY, USA, 2018. Association for Computing Machinery.
[618] Chuan-en Lin, Ta Ying Cheng, and Xiaojuan Ma. Architect: Building
interactive virtual experiences from physical affordances by bringing
human-in-the-loop. In Proceedings of the 2020 CHI Conference on

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 63
Human Factors in Computing Systems, CHI ’20, page 1–13, New York,
NY, USA, 2020. Association for Computing Machinery.
[619] Hannu Kukka, Johanna Ylipulli, Jorge Goncalves, Timo Ojala, Matias
Kukka, and Mirja Syrj¨
al¨
a. Creator-centric study of digital art exhibitions
on interactive public displays. In Proceedings of the 16th International
Conference on Mobile and Ubiquitous Multimedia, MUM ’17, page
37–48, New York, NY, USA, 2017. Association for Computing Machin-
ery.
[620] Elizabeth F. Churchill and Sara Bly. Culture vultures: Considering
culture and communication in virtual environments. SIGGROUP Bull.,
21(1):6–11, April 2000.
[621] Osku Torro, Henri Jalo, and Henri Pirkkalainen. Six reasons why
virtual reality is a game-changing computing and communication platform
for organizations. Commun. ACM, 64(10):48–55, September 2021.
[622] Cameron Harwick. Cryptocurrency and the problem of intermediation.
The Independent Review, 20(4):569–588, 2016.
[623] Henry Dunning Macleod. The elements of political economy. Longman,
Brown, Green, Lonqmaus and Roberts, 1858.
[624] Ferdinando M Ametrano. Hayek money: The cryptocurrency price
stability solution. Available at SSRN 2425270, 2016.
[625] Richard K Lyons and Ganesh Viswanath-Natraj. What keeps stable-
coins stable? Technical report, National Bureau of Economic Research,
2020.
[626] Min Kyung Lee, Anuraag Jain, Hea Jin Cha, Shashank Ojha, and
Daniel Kusbit. Procedural justice in algorithmic fairness: Leveraging
transparency and outcome control for fair algorithmic mediation. Proc.
ACM Hum.-Comput. Interact., 3(CSCW), November 2019.
[627] Allison Woodruff, Sarah E. Fox, Steven Rousso-Schindler, and Jeffrey
Warshaw. A qualitative exploration of perceptions of algorithmic fairness.
In Proceedings of the 2018 CHI Conference on Human Factors in
Computing Systems, CHI ’18, page 1–14, New York, NY, USA, 2018.
Association for Computing Machinery.
[628] Alex Zarifis, Leonidas Efthymiou, Xusen Cheng, and Salomi
Demetriou. Consumer trust in digital currency enabled transactions. In
International Conference on Business Information Systems, pages 241–
254. Springer, 2014.
[629] Immaculate Dadiso Motsi-Omoijiade. Financial intermediation in
cryptocurrency markets–regulation, gaps and bridges. In Handbook of
Blockchain, Digital Finance, and Inclusion, Volume 1, pages 207–223.
Elsevier, 2018.
[630] Ludwig Christian Schaupp and Mackenzie Festa. Cryptocurrency
adoption and the road to regulation. In Proceedings of the 19th Annual
International Conference on Digital Government Research: Governance
in the Data Age, pages 1–9, 2018.
[631] Larry D. Wall. Fractional reserve cryptocurrency banks, Apr 2019.
[632] Sean Foley, Jonathan R Karlsen, and T¯
alis J Putnin¸ ˇ
s. Sex, drugs, and
bitcoin: How much illegal activity is financed through cryptocurrencies?
The Review of Financial Studies, 32(5):1798–1853, 2019.
[633] Ioannis N Kessides. Market concentration, contestability, and sunk
costs. The Review of Economics and Statistics, pages 614–622, 1990.
[634] Avinash Dixit and Nicholas Stern. Oligopoly and welfare: A unified
presentation with applications to trade and development. European
Economic Review, 19(1):123–143, 1982.
[635] Pier Giuseppe Sessa, Neil Walton, and Maryam Kamgarpour. Exploring
the vickrey-clarke-groves mechanism for electricity markets. IFAC-
PapersOnLine, 50(1):189–194, 2017.
[636] Paul R. Milgrom. Putting auction theory to work. 2004.
[637] Thorstein Veblen and C Wright Mills. The theory of the leisure class.
Routledge, 2017.
[638] George A Akerlof. The market for “lemons”: Quality uncertainty and
the market mechanism. In Uncertainty in economics, pages 235–251.
Elsevier, 1978.
[639] Benjamin Klein, Andres V Lerner, and Kevin M Murphy. The
economics of copyright” fair use” in a networked world. American
Economic Review, 92(2):205–208, 2002.
[640] Zheng Wang, Dongying Lu, Dong Zhang, Meijun Sun, and Yan Zhou.
Fake modern chinese painting identification based on spectral–spatial
feature fusion on hyperspectral image. Multidimensional Systems and
Signal Processing, 27(4):1031–1044, 2016.
[641] Ahmed Elgammal, Yan Kang, and Milko Den Leeuw. Picasso, matisse,
or a fake? automated analysis of drawings at the stroke level for attribution
and authentication. In Thirty-second AAAI conference on artificial
intelligence, 2018.
[642] Jiang Wang, Yang Song, Thomas Leung, Chuck Rosenberg, Jingbin
Wang, James Philbin, Bo Chen, and Ying Wu. Learning fine-grained
image similarity with deep ranking. In Proceedings of the IEEE
conference on computer vision and pattern recognition, pages 1386–1393,
2014.
[643] Sean Bell and Kavita Bala. Learning visual similarity for product
design with convolutional neural networks. ACM transactions on graphics
(TOG), 34(4):1–10, 2015.
[644] Paarijaat Aditya, Rijurekha Sen, Peter Druschel, Seong Joon Oh,
Rodrigo Benenson, Mario Fritz, Bernt Schiele, Bobby Bhattacharjee, and
Tong Tong Wu. I-pic: A platform for privacy-compliant image capture.
In Proceedings of the 14th annual international conference on mobile
systems, applications, and services, pages 235–248, 2016.
[645] Jiayu Shu, Rui Zheng, and Pan Hui. Cardea: Context-aware visual
privacy protection for photo taking and sharing. In Proceedings of the
9th ACM Multimedia Systems Conference, pages 304–315, 2018.
[646] Alessandro Acquisti, Curtis Taylor, and Liad Wagman. The economics
of privacy. Journal of economic Literature, 54(2):442–92, 2016.
[647] Ranjan Pal, Jon Crowcroft, Abhishek Kumar, Pan Hui, Hamed Haddadi,
Swades De, Irene Ng, Sasu Tarkoma, and Richard Mortier. Privacy
markets in the Apps and IoT age. Technical Report UCAM-CL-TR-925,
University of Cambridge, Computer Laboratory, September 2018.
[648] Ranjan Pal, Jon Crowcroft, Yixuan Wang, Yong Li, Swades De, Sasu
Tarkoma, Mingyan Liu, Bodhibrata Nag, Abhishek Kumar, and Pan Hui.
Preference-based privacy markets. IEEE Access, 8:146006–146026, 2020.
[649] Soumya Sen, Carlee Joe-Wong, Sangtae Ha, and Mung Chiang. A
survey of smart data pricing: Past proposals, current plans, and future
trends. ACM Comput. Surv., 46(2), November 2013.
[650] Laura Schelenz. Diversity-aware recommendations for social justice?
exploring user diversity and fairness in recommender systems. In Adjunct
Proceedings of the 29th ACM Conference on User Modeling, Adaptation
and Personalization, UMAP ’21, page 404–410, New York, NY, USA,
2021. Association for Computing Machinery.
[651] Clyde W. Holsapple and Jiming Wu. User acceptance of virtual worlds:
The hedonic framework. SIGMIS Database, 38(4):86–89, October 2007.
[652] Abhisek Dash, Anurag Shandilya, Arindam Biswas, Kripabandhu
Ghosh, Saptarshi Ghosh, and Abhijnan Chakraborty. Summarizing
user-generated textual content: Motivation and methods for fairness in
algorithmic summaries. Proc. ACM Hum.-Comput. Interact., 3(CSCW),
November 2019.
[653] Ruotong Wang, F. Maxwell Harper, and Haiyi Zhu. Factors influencing
perceived fairness in algorithmic decision-making: Algorithm outcomes,
development procedures, and individual differences. In Proceedings of
the 2020 CHI Conference on Human Factors in Computing Systems, CHI
’20, page 1–14, New York, NY, USA, 2020. Association for Computing
Machinery.
[654] Joice Yulinda Luke and Lidya Wati Evelina. Exploring indonesian
young females online social networks (osns) addictions: A case study
of mass communication female undergraduate students. In Proceedings
of the 3rd International Conference on Communication and Information
Processing, ICCIP ’17, page 400–404, New York, NY, USA, 2017.
Association for Computing Machinery.
[655] Xiang Ding, Jing Xu, Guanling Chen, and Chenren Xu. Beyond
smartphone overuse: Identifying addictive mobile apps. In Proceedings
of the 2016 CHI Conference Extended Abstracts on Human Factors in
Computing Systems, CHI EA ’16, page 2821–2828, New York, NY, USA,
2016. Association for Computing Machinery.
[656] Simone Lanette, Phoebe K. Chua, Gillian Hayes, and Melissa Maz-
manian. How much is ’too much’? the role of a smartphone addiction
narrative in individuals’ experience of use. Proc. ACM Hum.-Comput.
Interact., 2(CSCW), November 2018.
[657] Amala V. Rajan, N. Nassiri, Vishwesh Akre, Rejitha Ravikumar, Amal
Nabeel, Maryam Buti, and Fatima Salah. Virtual reality gaming addiction.
2018 Fifth HCT Information Technology Trends (ITT), pages 358–363,
2018.
[658] Ramazan Ertel, O. Karakas¸, and Yusuf Do˘
gru. A qualitative research
on the supportive components of pokemon go addiction. AJIT-e: Online
Academic Journal of Information Technology, 8:271–289, 2017.
[659] Eui Jun Jeong, Dan J. Kim, and Dong Min Lee. Game addiction from
psychosocial health perspective. In Proceedings of the 17th International
Conference on Electronic Commerce 2015, ICEC ’15, New York, NY,
USA, 2015. Association for Computing Machinery.
[660] Robert Tyminski. Addiction to cyberspace: virtual reality gives analysts
pause for the modern psyche. International Journal of Jungian Studies,
10:91–102, 2018.
[661] Camino L ´
opez Garc´
ıa, Mar´
ıa Cruz S´
anchez G´
omez, and Ana Garc´
ıa-
Valc´
arcel Mu˜
noz Repiso. Scales for measuring internet addiction in covid-
19 times: Is the time variable still a key factor in measuring this addiction?
In Eighth International Conference on Technological Ecosystems for

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 64
Enhancing Multiculturality, TEEM’20, page 600–604, New York, NY,
USA, 2020. Association for Computing Machinery.
[662] Tomoyuki Segawa, Thomas Baudry, Alexis Bourla, Jean-Victor Blanc,
Charles Siegfried Peretti, St´
ephane Mouchabac, and Florian Ferreri.
Virtual reality (vr) in assessment and treatment of addictive disorders:
A systematic review. Frontiers in Neuroscience, 13, 2019.
[663] Ashley Colley, Jacob Thebault-Spieker, Allen Yilun Lin, Donald De-
graen, Benjamin Fischman, Jonna H¨
akkil¨
a, Kate Kuehl, Valentina Nisi,
Nuno Jardim Nunes, Nina Wenig, Dirk Wenig, Brent Hecht, and Johannes
Sch¨
oning. The geography of pok´
emon go: Beneficial and problematic
effects on places and movement. In Proceedings of the 2017 CHI
Conference on Human Factors in Computing Systems, CHI ’17, page
1179–1192, New York, NY, USA, 2017. Association for Computing
Machinery.
[664] Xin Tong, Ankit Gupta, Henry Lo, Amber Choo, Diane Gromala, and
Christopher D. Shaw. Chasing lovely monsters in the wild, exploring
players’ motivation and play patterns of pok´
emon go: Go, gone or go
away? In Companion of the 2017 ACM Conference on Computer Sup-
ported Cooperative Work and Social Computing, CSCW ’17 Companion,
page 327–330, New York, NY, USA, 2017. Association for Computing
Machinery.
[665] Russell Belk. Extended self and the digital world. Current Opinion in
Psychology, 10:50–54, 2016. Consumer behavior.
[666] Richard Lewis and Molly Taylor-Poleskey. Hidden town in 3d:
Teaching and reinterpreting slavery virtually at a living history museum.
J. Comput. Cult. Herit., 14(2), May 2021.
[667] Kelsey Virginia Dufresne and Bryce Stout. Anchorhold Afference:
Virtual Reality, Radical Compassion, and Embodied Positionality. As-
sociation for Computing Machinery, New York, NY, USA, 2021.
[668] Despoina Chatzakou, Ilias Leontiadis, Jeremy Blackburn, Emiliano De
Cristofaro, Gianluca Stringhini, Athena Vakali, and Nicolas Kourtellis.
Detecting cyberbullying and cyberaggression in social media. ACM Trans.
Web, 13(3), October 2019.
[669] Ruidong Yan, Yi Li, Deying Li, Yongcai Wang, Yuqing Zhu, and Weili
Wu. A stochastic algorithm based on reverse sampling technique to fight
against the cyberbullying. ACM Trans. Knowl. Discov. Data, 15(4), March
2021.
[670] Vivek K. Singh and Connor Hofenbitzer. Fairness across network
positions in cyberbullying detection algorithms. In Proceedings of
the 2019 IEEE/ACM International Conference on Advances in Social
Networks Analysis and Mining, ASONAM ’19, page 557–559, New York,
NY, USA, 2019. Association for Computing Machinery.
[671] Zahra Ashktorab and Jessica Vitak. Designing cyberbullying mitigation
and prevention solutions through participatory design with teenagers.
In Proceedings of the 2016 CHI Conference on Human Factors in
Computing Systems, CHI ’16, page 3895–3905, New York, NY, USA,
2016. Association for Computing Machinery.
[672] Zahra Ashktorab, Eben Haber, Jennifer Golbeck, and Jessica Vitak.
Beyond cyberbullying: Self-disclosure, harm and social support on askfm.
In Proceedings of the 2017 ACM on Web Science Conference, WebSci
’17, page 3–12, New York, NY, USA, 2017. Association for Computing
Machinery.
[673] Haewoon Kwak, Jeremy Blackburn, and Seungyeop Han. Exploring
cyberbullying and other toxic behavior in team competition online games.
In Proceedings of the 33rd Annual ACM Conference on Human Factors
in Computing Systems, CHI ’15, page 3739–3748, New York, NY, USA,
2015. Association for Computing Machinery.
[674] Clyde W. Holsapple and Jiming Wu. User acceptance of virtual worlds:
The hedonic framework. SIGMIS Database, 38(4):86–89, October 2007.
[675] K. J. Fietkiewicz, E. Lins, K. S. Baran, and W. G. Stock. Inter-
generational comparison of social media use: Investigating the online
behavior of different generational cohorts. In 2016 49th Hawaii Inter-
national Conference on System Sciences (HICSS), pages 3829–3838, Los
Alamitos, CA, USA, jan 2016. IEEE Computer Society.
[676] Marshall Van Alstyne. Why not immortality? Commun. ACM,
56(11):29–31, November 2013.
[677] Wenjun Hou, Huijie Han, Liang Hong, and Wei Yin. Chci: A crowd-
sourcing human-computer interaction framework for cultural heritage
knowledge. In Proceedings of the ACM/IEEE Joint Conference on Digital
Libraries in 2020, JCDL ’20, page 551–552, New York, NY, USA, 2020.
Association for Computing Machinery.
[678] C´
edric Denis-R´
emis, Olivier Codou, and Jean-Fabrice Lebraty. Rela-
tion of green it and affective attitude within the technology acceptance
model : The cases of france and china. Management & Avenir, 39:371–
385, 2010.
[679] Sun Zhe, T.N. Wong, and L.H. Lee. Using data envelopment analysis
for supplier evaluation with environmental considerations. In 2013 IEEE
International Systems Conference (SysCon), pages 20–24, 2013.
[680] L.H. LEE, T.N. WONG, and Z. SUN. An agent-based framework for
partner selection with sustainability considerations. IFAC Proceedings
Volumes, 46(9):168–173, 2013. 7th IFAC Conference on Manufacturing
Modelling, Management, and Control.
[681] Mel Slater, Cristina Gonzalez-Liencres, Patrick Haggard, Charlotte
Vinkers, Rebecca Gregory-Clarke, Steve Jelley, Zillah Watson, Graham
Breen, Raz Schwartz, William Steptoe, Dalila Szostak, Shivashankar
Halan, Deborah Fox, and Jeremy Silver. The ethics of realism in virtual
and augmented reality. In Frontiers in Virtual Reality, 2020.
[682] Abraham Hani Mhaidli and Florian Schaub. Identifying manipulative
advertising techniques in xr through scenario construction. In Proceedings
of the 2021 CHI Conference on Human Factors in Computing Systems,
CHI ’21, New York, NY, USA, 2021. Association for Computing Ma-
chinery.
[683] Ghaith Bader Al-Suwaidi and Mohamed Jamal Zemerly. Locating
friends and family using mobile phones with global positioning system
(gps). In 2009 IEEE/ACS International Conference on Computer Systems
and Applications, pages 555–558. IEEE, 2009.
[684] Alessandro Acquisti, Laura Brandimarte, and George Loewenstein.
Privacy and human behavior in the age of information. Science,
347(6221):509–514, 2015.
[685] Adil Rasheed, Omer San, and Trond Kvamsdal. Digital twin: Values,
challenges and enablers from a modeling perspective. Ieee Access,
8:21980–22012, 2020.
[686] Ana Reyna, Cristian Mart´
ın, Jaime Chen, Enrique Soler, and Manuel
D´
ıaz. On blockchain and its integration with iot. challenges and
opportunities. Future generation computer systems, 88:173–190, 2018.
[687] A Sghaier Omar and Otman Basir. Capability-based non-fungible
tokens approach for a decentralized aaa framework in iot. In Blockchain
Cybersecurity, Trust and Privacy, pages 7–31. Springer, 2020.
[688] Lanxiang Chen, Wai-Kong Lee, Chin-Chen Chang, Kim-Kwang Ray-
mond Choo, and Nan Zhang. Blockchain based searchable encryption
for electronic health record sharing. Future generation computer systems,
95:420–429, 2019.
[689] Haihan Duan, Jiaye Li, Sizheng Fan, Zhonghao Lin, Xiao Wu, and Wei
Cai. Metaverse for social good: A university campus prototype. arXiv
preprint arXiv:2108.08985, 2021.
[690] Business Standard. Users shun WhatsApp to
join Telegram, Signal amid rising data concerns,
2021. https://www.business-standard.com/article/companies/
users-shun- whatsapp-to- join-telegram-signal-amid-rising\protect\
penalty-\@M- data-concerns- 121010900271 1.html.
[691] Davide Salanitri, Glyn Lawson, and Brian Waterfield. The relationship
between presence and trust in virtual reality. In Proceedings of the
European Conference on Cognitive Ergonomics, ECCE ’16, New York,
NY, USA, 2016. Association for Computing Machinery.
[692] Chun-Cheng Chang, Rebecca A Grier, Jason Maynard, John Shutko,
Mike Blommer, Radhakrishnan Swaminathan, and Reates Curry. Using a
situational awareness display to improve rider trust and comfort with an
av taxi. In Proceedings of the Human Factors and Ergonomics Society
Annual Meeting, volume 63, pages 2083–2087. SAGE Publications Sage
CA: Los Angeles, CA, 2019.
[693] Abhishek Kumar, Tristan Braud, Young D. Kwon, and Pan Hui.
Aquilis: Using contextual integrity for privacy protection on mobile
devices. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., 4(4),
December 2020.
[694] Anthony Cuthbertson. Google admits giving hundreds of firms to your
gmail inbox. The Independent, 2018.
[695] The IEEE Global Initiative on Ethics of Autonomous
and Intelligent Systems. Extended Reality in A/IS, 2021.
https://standards.ieee.org/content/dam/ieee-standards/standards/web/
documents/other/ead/EAD1e extended reality.pdf.
[696] Brianna Dym and Casey Fiesler. Social norm vulnerability and its
consequences for privacy and safety in an online community. Proc. ACM
Hum.-Comput. Interact., 4(CSCW2), October 2020.
[697] Paritosh Bahirat, Yangyang He, Abhilash Menon, and Bart Knijnen-
burg. A data-driven approach to developing iot privacy-setting interfaces.
In 23rd International Conference on Intelligent User Interfaces, IUI ’18,
page 165–176, New York, NY, USA, 2018. Association for Computing
Machinery.
[698] Abhishek Kumar, Tristan Braud, Lik-Hang Lee, and Pan Hui. Theo-
phany: Multimodal speech augmentation in instantaneous privacy chan-
nels. In Proceedings of the 29th ACM International Conference on

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 65
Multimedia (MM ’21), October 20–24, 2021, Virtual Event, China.
Association for Computing Machinery (ACM), 2021.
[699] Koki Nagano, Jaewoo Seo, Kyle San, Aaron Hong, Mclean Goldwhite,
Jun Xing, Stuti Rastogi, Jiale Kuang, Aviral Agarwal, Hanwei Kung, et al.
Deep learning-based photoreal avatars for online virtual worlds in ios. In
ACM SIGGRAPH 2018 Real-Time Live!, pages 1–1. 2018.
[700] Abhishek Kumar, Tristan Braud, Sasu Tarkoma, and Pan Hui. Trust-
worthy ai in the age of pervasive computing and big data. In 2020 IEEE
International Conference on Pervasive Computing and Communications
Workshops (PerCom Workshops), pages 1–6, 2020.
[701] Abhishek Kumar, Benjamin Finley, Tristan Braud, Sasu Tarkoma, and
Pan Hui. Sketching an ai marketplace: Tech, economic, and regulatory
aspects. IEEE Access, 9:13761–13774, 2021.
[702] Abhishek Kumar, Purva Grover, Arpan Kumar Kar, and Ashis K.
Pani. It consulting: A systematic literature review. In Arpan Kumar
Kar, P. Vigneswara Ilavarasan, M.P. Gupta, Yogesh K. Dwivedi, Matti
M¨
antym¨
aki, Marijn Janssen, Antonis Simintiras, and Salah Al-Sharhan,
editors, Digital Nations – Smart Cities, Innovation, and Sustainability,
pages 474–484, Cham, 2017. Springer International Publishing.
[703] Richard Cloete, Chris Norval, and Jatinder Singh. A call for auditable
virtual, augmented and mixed reality. In 26th ACM Symposium on Virtual
Reality Software and Technology, pages 1–6, 2020.
[704] Daisuke Wakabayashi. Self-Driving Uber Car Kills Pedestrian in
Arizona, Where Robots Roam, 2018. https://www.nytimes.com/2018/03/
19/technology/uber-driverless-fatality.html.
[705] Ahmad Yousef Alhilal, Tristan Braud, and Pan Hui. A roadmap toward
a unified space communication architecture. IEEE Access, 9:99633–
99650, 2021.
Lik-Hang Lee received the Ph.D. degree from
SyMLab, Hong Kong University of Science and
Technology, and the bachelor’s and M.Phil. degrees
from the University of Hong Kong. He is an assis-
tant professor (tenure-track) with Korea Advanced
Institute of Science and Technology (KAIST), South
Korea. He is also the Director of the Augmented
Reality and Media Laboratory, KAIST. He has built
and designed various human-centric computing spe-
cialising in augmented and virtual realities (AR/VR).
He is also a founder of an AR startup company,
promoting AR-driven education and serving over 100 Hong Kong and Macau
schools.
Tristan Braud is an assistant professor at the Hong
Kong University of Science and Technology, within
the Systems and Media Laboratory (SymLab). He
received a Masters of Engineering degree from both
Grenoble INP, Grenoble, France and Politecnico di
Torino, Turin, Italy, and a PhD degree from Uni-
versit Grenoble-Alpes, Grenoble, France. His major
research interests include pervasive and mobile com-
puting, cloud and edge computing, human centered
system designs and augmented reality with a specific
focus on human-centred systems. With his research,
Dr. Braud aims at bridging the gap between designing novel systems, and the
human factor inherent to every new technology.
Pengyuan Zhou received his PhD from the Uni-
versity of Helsinki. He was a Europe Union Marie-
Curie ITN Early Stage Researcher from 2015 to
2018. He is currently a research associate professor
at the School of Cyberspace Science and Tech-
nology, University of Science and Technology of
China (USTC). He is also a faculty member of
the Data Space Lab, USTC. His research focuses
on distributed networking AI systems, mixed reality
development, and vehicular networks.
Lin Wang is a Postdoc researcher in Visual In-
telligence Lab., Dept. of Mechanical Engineering,
Korea Advanced Institute of Science and Tech-
nology (KAIST). His research interests include
neuromorphic camera-based vision, low-level vi-
sion (especially image super-solution, HDR imag-
ing, and image restoration), deep learning (espe-
cially adversarial learning, transfer learning, semi-
/self-unsupervised learning) and computer vision-
supported AR/MR for intelligent systems.
Dianlei Xu is a joint doctoral student in the Depart-
ment of Computer Science, Helsinki, Finland and
Beijing National Research Center for Information
Science and Technology (BNRist), Department of
Electronic Engineering, Tsinghua University, Bei-
jing, China. His research interests include edge/fog
computing and edge intelligence.
Zijun Lin is an undergraduate student at Univer-
sity College London (UCL) and a research intern
at the Augmented Reality and Media Lab at Ko-
rea Advanced Institute of Science and Technology
(KAIST). His research interests are the applica-
tion of behavioural economics in human-computer
interaction and the broader role of economics in
computer science.
Abhishek Kumar is a PhD student at the Systems
and Media Lab in the Department of Computer
Science at the University of Helsinki. He has a
MSc in Industrial and Systems Engineering, and
a BSc in Computer Science and Engineering. His
research focus is primarily in areas of Multimedia
Computing, Multimodal Computing and Interaction
with special focus on privacy.

JOURNAL OF L
A
T
E
X CLASS FILES, VOL. 14, NO. 8, SEPTEMBER 2021 66
Carlos Bermejo Fernandez received his Ph.D. from
the Hong Kong University of Science and Technol-
ogy (HKUST). His research interests include human-
computer interaction, privacy, and augmented reality.
He is currently a Postdoc researcher at the SyMLab
in the Department of Computer Science at HKUST.
Pan Hui received the Ph.D. degree from Computer
Laboratory, University of Cambridge, and the bache-
lor and M.Phil. degrees from the University of Hong
Kong. He is a Professor of Computational Media
and Arts and Director of the HKUST-DT System
and Media Laboratory at the Hong Kong University
of Science and Technology, and the Nokia Chair
of Data Science at the University of Helsinki. He
has published around 400 research papers and with
over 21,000 citations. He has 32 granted and filed
European and U.S. patents in the areas of augmented
reality, data science, and mobile computing. He is an ACM Distinguished
Scientist, a Member of the Academy of Europe, a Fellow of the IEEE, and
an International Fellow of the Royal Academy of Engineering.