5G-Enabled Tactile Internet

Abstract

The long-term ambition of the Tactile Internet is to enable a democratization of skill, and how it is being delivered globally. An integral part of this is to be able to transmit touch in perceived real-time, which is enabled by suitable robotics and haptics equipment at the edges, along with an unprecedented communications network. The fifth generation (5G) mobile communications systems will underpin this emerging Internet at the wireless edge. This paper presents the most important technology concepts which lay at the intersection of the larger Tactile Internet and the emerging 5G systems. The paper outlines the key technical requirements and architectural approaches for the Tactile Internet, pertaining to wireless access protocols, radio resource management aspects, next generation core networking capabilities, edge-cloud and edge-AI capabilities. The paper also highlights the economic impact of the Tactile Internet as well as a major shift in business models for the traditional telecommunications ecosystem.

Full text

460 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
5G-Enabled Tactile Internet
Meryem Simsek, Adnan Aijaz, Member, IEEE, Mischa Dohler, Fellow, IEEE, Joachim Sachs, and
Gerhard Fettweis, Fellow, IEEE
Abstract—The long-term ambition of the Tactile Internet is to
enable a democratization of skill, and how it is being delivered
globally. An integral part of this is to be able to transmit touch
in perceived real-time, which is enabled by suitable robotics and
haptics equipment at the edges, along with an unprecedented
communications network. The fifth generation (5G) mobile com-
munications systems will underpin this emerging Internet at the
wireless edge. This paper presents the most important technol-
ogy concepts, which lay at the intersection of the larger Tactile
Internet and the emerging 5G systems. The paper outlines the
key technical requirements and architectural approaches for the
Tactile Internet, pertaining to wireless access protocols, radio
resource management aspects, next generation core networking
capabilities, edge-cloud, and edge-AI capabilities. The paper also
highlights the economic impact of the Tactile Internet as well as a
major shift in business models for the traditional telecommunica-
tions ecosystem.
Index Terms—Tactile Internet, haptic communications,
real-time communication, edge intelligence, ultra-low latency,
ultra-high reliability, 5G, massive connectivity, OFDM.
I. INTRODUCTION
M
OBILE communications continues to play an important
role in modern economy, including consumer, health,
education, logistics, and other major industries. Mobile com-
munications networks of today have successfully connected a
vast majority of global population. After creating the Mobile
Internet, connecting billions of smart phones and laptop, the
focus of mobile communications is moving towards provid-
ing ubiquitous connectivity for machines and devices, thereby
creating the Internet-of-Things (IoT) [1].
With the technological advancements of today, stage is being
set f or the emergence of the Tactile Internet in which ultra-
reliable and ultra-responsive network connectivity will enable
it to deliver real-time control and physical tactile experiences
remotely. The Tactile Internet will provide a true paradigm
shift from content-delivery to skill-set delivery networks, and
thereby revolutionize almost every segment of the society.
As per ITU [2], the Tactile Internet will add a new dimen-
sion to human-machine interaction by delivering a low latency
Manuscript received August 7, 2015; revised October 29, 2015. Date of
publication February 11, 2016; date of current version March 15, 2016.
M. Simsek and G. Fettweis are with the Technical University Dresden,
Dresden, Germany (e-mail: meryem.simsek@tu-dresden.de).
A. Aijaz was with the Centre for Telecommunications Research, King’s
College London, London, U.K. He is now with Telecommunications Research
Laboratories, Toshiba Research Europe Ltd., Bristol, U.K.
M. Dohler is with the Centre for Telecommunications Research, King’s
College London, London, U.K.
J. Sachs is with Ericsson Research, Stockholm, Sweden.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSAC.2016.2525398
enough to build real-time interactive systems. Further, the
Tactile Internet has been described as a communication infras-
tructure combining low latency, very short transit time, high
availability and high reliability with a high level of security
[3], [4]. Associated with cloud computing proximity through
e.g. mobile edge-clouds and combined with the virtual or aug-
mented reality for sensory and haptic controls, the Tactile
Internet addresses areas with reaction times in the order of
a millisecond. Example areas are real-time gaming, industrial
automation, transportation systems, health and education.
Because the Tactile Internet will be servicing really critical
aspects of society, it will need to be ultra-reliable, with a sec-
ond of outage per year, support very low latencies, and have
sufficient capacity to allow large numbers of devices to com-
municate with each other simultaneously and autonomously. It
will be able to interconnect with the traditional wired Internet,
the mobile Internet and the IoT thereby forming an Internet
of entirely new dimensions and capabilities. State-of-the-art
fourth generation (4G) mobile communications systems do not
largely f ulfil the technical requirements for the Tactile Internet.
Therefore, fifth generation (5G) mobile communications sys-
tems are expected to underpin the Tactile Internet at the wireless
edge.
5G wireless access is the wireless access solution to ful-
fill the wireless communication requirements for 2020 and
beyond [5]. At ITU, ITU-R working party 5G has the respon-
sibility for the terrestrial radio system aspects of international
mobile telecommunications (IMT) systems, which today com-
prise IMT-2000 (i.e. 3G) and IMT-Advanced (i.e. 4G) . 5G is
treated under the term IMT-2020 of which the scope is currently
being developed in a new ITU-R recommendation typically
referred to as IMT Vision. An early assessment of 5G sce-
narios and requirements has been developed in the METIS
research project [6], [7], and recently also by the telecommuni-
cations industry alliance NGMN [8]. Overall there is a common
understanding that 5G should not only support an evolution of
traditional mobile communication services, such as personal
mobile multimedia communication or personal mobile broad-
band services; 5G should in addition address novel use cases
including for example machine type communication (in several
fields like e.g. smart energy networks or smart grids, vehicular
communication and intelligent transport systems, or sensor net-
working) or novel ways of media distribution. This range of 5G
use cases pushes the requirements on 5G in several dimensions,
like latency, data rate, device and network energy efficiency,
mobility, reliability, traffic volume density, connection density,
etc. The broad range of targeted 5G capabilities will make it an
important enabler for the Tactile Internet.
0733-8716 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 461
To facilitate the expected massive increase of traffic to be
handled in a 5G system, additional spectrum has to be allo-
cated to the 5G wireless access. The spectrum range up to a
few GHz is of special importance in order to provide wide-area
coverage. However, in order to enable very high capacity and
very high data rates of multi-Gb/s spectrum above 10 GHz will
also be needed. The entire spectrum range from around 1 GHz
to the millimeter-wave range up to around 100 GHz is conse-
quently relevant for 5G. One important consideration is that at
around 2020 large deployments of LTE will be operating in the
spectrum below 6.5 GHz. It is desirable that next-generation
wireless functionalities in these bands can be deployed compat-
ibly to the deployed systems, primarily LTE, so that the largely
deployed pre-5G devices can continue to run their services.
For deployments in new spectrum, 5G wireless access can be
deployed without constraints of backwards compatibility.
Overall, a tight integration with LTE is desirable for 5G,
which is identified as an important requirement by many indus-
try players: This is mainly to enable that 5G services can be
quickly and efficiently introduced from the early 5G deploy-
ments when 5G availability is still limited [9]. In a nutshell,
5G wireless access will consist of an evolution of LTE comple-
mented with new radio technologies and architecture designs
[10].
Given these unprecedented mobile technology capabilities,
we believe that 5G will play an integral part of the Tactile
Internet connectivity ecosystem. The intersection of the Tactile
Internet and 5G is thus focus of this paper. To this end, the
paper is organized as follows. In Section II, we outline exciting
Tactile Internet applications which we envisage will be popular
once the network is operational. In Section III, we then out-
line the Tactile Internet requirements which stem directly from
the application scenarios and which resonate with many of the
5G requirements. In Sections IV–VII, we dive into technical
issues specific to the 5G mobile community, i.e. architecture,
hardware, access, radio resource management as well as radio
access, core networks and edge-cloud designs. In Section VIII,
we outline the importance and realization of edge artificial
intelligence (AI) capabilities. In Section IX, the economic
impact of the Tactile Internet is assessed. Finally, in Section X,
conclusions are drawn and future work outlined.
II. A
PPLICATIONS AND SERVICES
The Tactile Internet will enhance the way of communication
and lead to more realistic social interaction in various envi-
ronments. Current wireless local area network (WLAN) and
cellular systems do not yield anything close to achieving an
end-to-end latency of 1ms which is crucial for Tactile Internet
applications as shown in Section III.A. It therefore is diffi-
cult to comprehend a complete list of possible Tactile Internet
applications which can emerge. In this section, some main
examples are provided to show the ground-breaking potential
of the Tactile Internet.
A. Automation in Industry
Industrial automation together with machine type communi-
cation is one of the applications discussed within the framework
of 5G systems. Within such applications, various control pro-
cesses exist and require different end-to-end latency, data rate,
reliability and security [11], [12]. The sensitivity of rapidly
moving devices’ control circuits is, for example, significantly
below 1 ms per sensor [12]–[14]. Hence, the automation in
industry is a key application field in the Tactile Internet. Today,
control processes are realized by fast wired connection, e.g.
the industrial Ethernet. In the future, these wired systems are
aimed to be fully or partially replaced by wireless systems in
order to enable high flexibility in production, i.e. the industrial
revolution [15]. This requires a guaranteed reliability and mini-
mum end-to-end latency and can be enabled by Tactile Internet
solutions.
B. Autonomous Driving
Fully automated driving and platooning of vehicles is dis-
cussed as a new step in mobility within the context of 5G.
A considerable and sustainable reduction of road accidents
and traffic jams can be realized by autonomous driving, i.e.
vehicle-to-vehicle or vehicle-to-infrastructure communication
and coordination. The time needed for collision avoidance in
today’s applications for vehicle safety is below 10 ms. If a bi-
directional data exchange for automatic driving manoeuvres is
considered, a latency in the order of a millisecond will likely be
needed. This can technically be realized by the Tactile Internet
and its 1 ms end-to-end latency.
Fully autonomous driving is expected to change the traffic
behavior entirely. Especially small distances between auto-
mated vehicles, in particular in platoons, potentially safety-
critical situations need to be detected earlier than with
human drivers. This requires ultra-high reliable and pro-
active/predictive behavior in future wireless communication
systems.
C. Robotics
In recent years, the technical potential of robotics has
increased in various fields. Its demonstrated potential comes
with increased complexity, and various challenges, so that
autonomous robotics will find their application only in a lim-
ited range of rather specific areas, e.g. autonomous driving, in
the near future. Remotely controlled robots with real-time, syn-
chronous and visual-haptic feedback, however, seem to be a
promising alternative to autonomous robots. Controlling robots
must happen at latency reaction times that are fast enough for
the robot and its object. If a real-time controlling and commu-
nication is not guaranteed, robots will move in fitful manner
which may lead to an oscillatory behavior. For many robotics
scenarios in manufacturing this has led to a maximum latency
target of a communication link of 100 µs, and round-trip reac-
tion times of 1 ms, the target as discussed for the Tactile
Internet.
D. Healthcare
Tele-diagnosis, tele-surgery and tele-rehabilitation are just
some of the many potential applications of the Tactile Internet
in healthcare. Using advanced tele-diagnostic tools, medical

462 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
expertise could be available anywhere and anytime regard-
less of the physician’s location [16]. Hereby, a tele-robot at
the patient’s location will be controlled by the physician, so
that not only audio and/or visual information but also haptic
feedback is provided. The same technical principle is applied
to tele-surgery applications. In tele-rehabilitation techniques
can be used for patients to remotely steer and control his/her
motions. In all Tactile Internet based healthcare technologies,
high fidelity and extreme precision is fundamental to enable the
deployment of tele-medical technology.
E. Virtual and Augmented Reality
Existing virtual and augmented reality applications can sig-
nificantly benefit from the availability of the Tactile Internet.
The virtual reality is a shared, haptic virtual environment in
which s everal users are physically coupled via a simulation
tool to jointly/collaboratively perform tasks by perceiving the
objects not only audio-visually but also via the touch sense.
In augmented reality, on the other hand, the combination of
real and computer generated content is visualized in the user’s
field of view. The major goal of future augmented reality
applications, compared to today’s static information augmen-
tation, is the visualization of dynamic content and up-to-date
information.
Haptic feedback in virtual reality is a prerequisite for high-
fidelity interaction. Especially, the perception of objects in
virtual reality via the sense of touch leads to various appli-
cations relying on high level of precision. This precision can
only be realized if the latency between the users and the virtual
reality is a few milliseconds.
The augmentation of additional information into a user’s
field of view enables the development of many assistance sys-
tems, e.g. maintenance, driver-assistance systems, education.
With the Tactile Internet the content in augmented reality can be
moved from static to dynamic. This enables a real-time virtual
extension of a user’s field of view, so that possible dangerous
events can be identified and avoided.
F. Further Tactile Internet Applications
Additional Tactile Internet applications are serious gam-
ing, education, individualized manufacturing, and unmanned
autonomous systems. Serious games are real-world simulations
designed for the purpose of solving a problem. The end-to-
end delay in the interaction between players and games is a
key factor influencing the quality of players’ experience and
the game’s usability, since the delay influences directly the
perceived realism of the game.
Individualized manufacturing unlike the mass production in
today’s assembly line based production processes will enable
the manufacturing of good in production islands. Hereby,
mobile robots will deliver assembly parts on demand. This
requires a wireless real-time tactile communication network
among the mobile robots.
Unmanned autonomous or remotely controlled systems are
increasingly used in a large number of contexts to sup-
port humans in dangerous and difficult-to-reach environments,
remotely controlled by humans, or for tasks that are too tedious
or repetitive for humans. The remote control of an unmanned
aircraft, for example, can be realized with high precision and
without any reaction delay with a reduced end-to-end latency
as a Tactile Internet application.
III. T
ACTILE INTERNET REQUIREMENTS
The Tactile Internet, wherein humans will wirelessly control
real and virtual objects, will not be realized without overcom-
ing the enormous system design challenges. Some of the most
stringent design challenges for the Tactile Internet have been
recently presented in [17].
Human beings’ interaction with their environment is cru-
cial. Our perceptual processes limit the speed of our interaction
with our environment. We experience interaction with a techni-
cal system as intuitive and natural only if the feedback of the
system is adapted to our human reaction time. Consequently,
the requirements for technical systems enabling real-time inter-
actions depend on the participating human senses. Hereby,
reaction times of about 100 ms, 10 ms, and 1ms is required for
auditory, visual, and manual interaction, respectively. Realizing
these reaction times, all human senses can, in principle, interact
with machines. Hence, human beings should not only be able
to see and hear things far away, but also touch and feel them.
Transmitting accurately the equivalent of human touch via data
networks is the vision to close the data cycle [18], which is
aimed to be realized by the Tactile Internet.
In the following, we highlight the key technical requirements
for realizing the Tactile Internet.
A. Ultra-Responsive Connectivity
The Tactile Internet requires ultra-responsive network con-
nectivity i.e., end-to-end latency on the order of 1 ms [19],
[20]. For real-time transmission as otherwise the tactile users
will experience cyber-sickness, which occurs primarily as a
result of conflicts between visual, vestibular, and propriocep-
tive sensory systems [21]. Thus, if eyes perceive a movement
which is slightly delayed compared to what is perceived by the
vestibular system while the remainder of the human being’s
body remains static, this delay leads to cyber-sickness. This
is especially important for technical systems with tactile and
haptic interaction or for mission critical communications, e.g.
machine-type communication which enable r eal-time control
and automation of dynamic processes in industrial automation,
manufacturing, traffic management, etc.
The end-to-end latency (round-trip delay) in technical sys-
tems includes the time spent in the transmission of the informa-
tion from a sensor (or human in case of haptic interaction) via
the communication infrastructure to a control server; the pro-
cessing of the information and the eventual retransmission via
the communication infrastructure back to the actuator (human).
Considering an end-to-end latency of 1 ms the latency budget
for wireless transmission is even lower than 1 ms (see Fig. 3).
B. Ultra-Reliable Connectivity
The phenomenal success of cellular networks has been based
on providing ubiquitous and reliable wide area coverage for

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 463
voice and text communications. With 4G and the success of
mobile computing devices the industry is targeting to bring
the same reliability and ubiquity of access to mobile inter-
net applications such as web browsing and audio and video
streaming. The Long Term Evolution (LTE) today already is
providing effective data rates of around 50 Mb/s. However, con-
sidering technology and market forces in 10 years, we must
be able to address cellular speeds of 10 Gb/s or more and
introduce new applications [5], [6], [22]. While high data rates
will be a key feature of 5G networks, another key challenge is
to be able to provide carrier grade access reliability. Beyond
single digit end-to-end latency, ultra-reliable network connec-
tivity is an important requirement for the Tactile Internet.
Reliability refers here to the probability to guarantee a required
function/performance under stated conditions for a given time
interval [23]. The specific reliability requirements differ for
various types of services and applications.
Demands for the highest possible reliability are associated
with requirements for real-time response. This becomes clear,
with applications addressed in Section II requiring a reliable
reception of rapidly transmitted data. A failure rate even below
10
−7
might be necessary in some 5G applications [24], [25].
This corresponds to merely 3.17 seconds of outage per year.
Wireless systems of today are built around the perception that
a link with 3% outage is a good link. However, when two links
with uncorrelated channels are combined, 3% outage per link
generates a combined outage of approximately 10
−3
.Andfive
uncorrelated links can already achieve an outage of less than
10
−7
[4]!
Hence, the simultaneous connection to multiple links (multi-
connectivity) might be a potential solution for achieving the
required hard-bound (i.e. not average!) reliability for tactile
applications [26]. The achieved reliability will also have a
positive impact onto delay since less re-transmissions will be
needed.
C. Security and Privacy
Safety and Privacy are also the key requirements for the
Tactile Internet. With stringent latency constraints, security
must be embedded in the physical transmission and ideally
be of low computational overhead. Novel coding techniques
need to be developed for tactile applications that allow only
the legitimate receivers to process a secure message. Absolute
security will, hereby, be achieved if an illegitimate receiver can-
not decode the date even with infinite computational power.
This rises a challenge, especially in massive connectivity appli-
cations. Identification of legitimate receivers requires novel
reliable and low-delay methods. One such method could be
the usage of hardware specific attributes such as biometric
fingerprints.
D. Tactile Data
The Tactile Internet must handle the tactile information in the
same way as the conventional audio/visual information. Hence,
tactile encoding mechanisms are needed which facilitate trans-
mission of tactile information over packet-switched networks.
Besides, there must be provisioning of audio/visual sensory
feedback due to the highly multi-dimensional nature of human
tactile perception [27].
E. Edge Intelligence
The Tactile Internet must overcome the fundamental limi-
tation due to finite speed of light. Without this, the range of
tactile services and applications would be limited to 100km
(assuming most is through fiber). To overcome this, the Tactile
Internet must support a hybrid composition of machine and
human actuation mixing real tactile actuation with intelligence-
based predictive actuation. Such predictive actuation should
be in close proximity of the tactile edge. Therefore, the edge
of the network (mobile edge cloud) must be equipped with
intelligence to facilitate predictive caching as well as inter-
polation/extrapolation of human actions. This necessitates the
development of novel artificial intelligence techniques for edge
cloud architectures.
IV. A
RCHITECTURE AND TECHNOLOGY COMPONENTS
Unlike the conventional Internet which provides the medium
for audio and visual transport, the Tactile Internet will pro-
vide the medium for transporting t ouch and actuation in real-
time i.e., ability of haptic and non-haptic control through the
Internet. Unlike auditory and visual senses, the sense of touch
occurs bilaterally i.e., it is sensed by imposing a motion on
an environment and feeling the environment by a distortion or
reaction force [28]. The key distinction between haptic and non-
haptic control is that in case of the former, there is actually a
haptic feedback (kinesthetic or vibro-tactile) from the system,
in addition to audio/visual feedback, thereby closing a global
control loop; whereas in case of the latter, the feedback can
only be audio/visual and, hence, there is no notion of a control
loop. It should be noted that the haptic control is inherent to a
majority of tactile applications.
As shown in Fig. 1, the end-to-end architecture for the
Tactile Internet can be split into three distinct domains: a master
domain, a network domain, and a controlled domain.
A. Master Domain
The master domain usually consists of a human (opera-
tor) and a human system interface (HSI). The HSI is actually a
haptic device (master r obot), which converts the human input to
tactile input through various tactile coding techniques. The hap-
tic device allows a user to touch, feel, and manipulate objects in
real and virtual environments, and primarily controls the oper-
ation of the controlled domain as discussed later. It should be
noted that in some applications, multiple operators can collab-
oratively control the operation of a single controlled domain.
The master domain also has the provisioning for auditory and
visual feedbacks. In addition to being important requirement
for non-haptic control, the auditory and visual feedbacks play
a critical role in increasing the perceptual performance as the
human brain naturally integrates different sensory modalities
[29].
State-of-the-art haptic devices, available from vendors like
Geomagic and Sensable are usually designed in the form of a

464 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
Fig. 1. Functional representation of the Tactile Internet architecture.
linkage-based system which consists of a robotic arm attached
to a stylus. The robotic arm tracks the position of the stylus
and is capable of exerting a force on its tip. To truly realize
the vision of the Tactile Internet, further developments on hap-
tic devices are needed; particularly in increasing the degrees
of freedom (DoF) to meet the demands of envisioned applica-
tions and embedding the network interface for direct or indirect
communication with the cellular network.
B. Controlled Domain
The controlled domain consists of a teleoperator (controlled
robot) and is directly controlled by the master domain through
various command signals. The teleoperator interacts with vari-
ous objects in the remote environment. Typically, no apriori
knowledge exists about the environment. Through command
and feedback signals, energy is exchanged between the master
and controlled domains thereby closing a global control loop.
C. Network Domain
The network domain provides the medium for bilateral com-
munication between the master and controlled domains, and
therefore kinesthetically couples the human to the remote envi-
ronment. Ideally, the operator is completely immersed into
the remote environment. The Tactile Internet requires ultra-
reliable and ultra-responsive network connectivity that would
enable typical reliabilities and latencies for real-time haptic
interaction. The underlying 5G-driven communication archi-
tecture, composed of the Radio Access Network (RAN) and
Core Network CN), is expected to meet the key requirements in
realizing the vision of the Tactile Internet.
To this end, the important functions of the 5G RAN in the
Tactile Internet ecosystem are as follows: i) efficient support
of various Radio Access Technologies (RATs) such as tradi-
tional cellular, emerging millimeter-wave, massive MIMO, full-
duplex, etc. ii) Tactile QoE/QoS aware scheduling and radio
resource management for tactile applications in co-existence
of other vertical applications such as machine-to-machine,
vehicle-to-vehicle, smart grids, etc. iii) efficient packet delivery
through reliable radio protocols and physical (PHY) layer, and
iv) optimal resolution of air-interface conflicts through novel
medium access control (MAC) techniques. The key function-
alities of the 5G Core Network (CN) relevant to the Tactile
Internet are as follows: i) dynamic application-aware QoS
provisioning, ii) edge-cloud access, and iii) security.
Although a number of research efforts are focusing on 5G
systems, there is no unanimous agreement on a 5G network
architecture yet. However, both the academic and industrial
communities have a general consensus that 5G networks must
be designed in a flexible manner such that one network, based
on a common physical infrastructure, is efficiently shared
among different vertical applications. Such sharing will be pos-
sible through greater degree of abstraction of 5G networks
wherein different network slices would be allocated to differ-
ent vertical application sectors. A network slice is defined as
a connectivity service based on various customizable software-
defined functions that govern geographical coverage area, avail-
ability, robustness, capacity, and security [30]. Such slicing
approach provides more of a network on-demand functionality.
The recent trends of network function virtualization
(NFV) (providing abstraction [31]) and software defined net-
working (SDN) (providing flexibility [32]) are critical in shap-
ing such an envisioned architecture. NFV provides the sepa-
ration of network functions from the hardware infrastructure;
the network function can be managed as a software module
that can be deployed in any standard cloud computing infras-
tructure. On the other hand, SDN provides an architectural
framework wherein control and data planes are decoupled,
and enables direct programmability of network control through
software-based controllers [33].
Whilst SDN/NFV was initially proposed for the Internet
infrastructure, the approach is now being considered/used
closer to the edge in the cellular CN and even Cloud-RAN. That
opens the interesting possibility to provide an end-to-end flex-
ible/abstracted architecture based on radio-aware SDN/NFV
slicing in the networking domain and network-aware Radio
Resource Management (RRM) & scheduler approaches in the
wireless domain. Through such a coupling, it is possible to

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 465
Fig. 2. A logical approach to the 5G network architecture based on a common
physical infrastructure.
design one network in a flexible manner offering different
end-to-end network slices to different vertical applications.
For example, the logical architectural approach, illustrated in
Fig. 2, builds on a common programmable physical infras-
tructure and an NFV-enabled network cloud that provides the
required protocol stack functionalities. The SDN controller pro-
vides the functionality of programming the core and radio
access network. The software implementation of network func-
tions, termed as Virtualized Network Functions (VNFs) soft-
ware, is deployed on the underlying infrastructure. The VNF
management and orchestration framework is used to monitor,
manage, and troubleshoot VNFs software. Such an architectural
approach enables flexible and dynamic slicing of end-to-end
network and service resources, which is particularly attractive
to cater for the requirements of tactile as well as other vertical
applications.
Said architectural approach can only be realized through
advanced hardware, PHY, network and radio protocols, under-
pinned by cloud and content-centric designs. These are now
discussed in more details in subsequent sections.
V. H
ARDWARE AND PHY DESIGN
The presented applications and requirements of the Tactile
Internet can only be realized by novel hardware and Physical
layer design. In this section, we present potential solutions for
the ultra-low latency requirements of the Tactile Internet.
A. Revolution in Hardware
As highlighted in Section III.A, for achieving a round-trip
latency of 1ms, the communication delay due to the speed of
light needs to be considered as well. Within 1 ms light trav-
els (assuming mostly fiber) some 200 km. This means that the
maximum distance for a steering and control server (see Fig. 3
to be placed from the point of tactile interaction by the users
is 100 km away. This distance assumes no processing nor net-
work congestion delays anywhere along the communication.
Considering a more realistic scenario by taking the additional
signal processing, protocol handling, and switching and net-
work delays into account, this requires the control/steering
server, to be in the range of a few kilometers from the tactile
Fig. 3. Examplary latency objectives of Tactile Internet systems.
point of interaction. Hence, control and steering servers need to
be built as close to the base station (access point) as possible.
The concept of putting servers at the edge of the mobile
radio access network has been coined as the (mobile) edge
cloud and the possibly best way to enable this proximity is to
combine servers into the same box as base stations and access
points. As a solution to this, the vision of a highly adaptive
and energy-efficient computing (HAEC) box has been estab-
lished [34], [35]. The HAEC box is a novel concept on how
computing design can be built by utilizing optical and wireless
chip-to-chip communication to achieve significant increase in
performance compared to today’s servers. Hence, the HAEC
box is a very performant, energy-adaptive computing platform
which can replace a base station and a control/steering server.
The integration of wireless and optical transceiver function-
ality is integrated into one 3D-stacked chip, where individual
dies are thinned and vertically stacked on each other. This
allows very compact and powerful systems, with the additional
benefit of mixed technologies, e.g. a digital processor could be
realized in a certain process (e.g. complementary metal-oxide
semiconductor (CMOS)), while the analog front-end could be
realized in another process, allowing higher transmit power.
Such a technology offers short connections between front-end.
With the availability of 3D chip stack integration, antenna
designs will not be limited to chip based materials, i.e. semicon-
ductors and insulator materials. Different materials can be used
in one chip stack. This enables, for example, antenna designers
to consider materials with low losses at high frequencies and
low permittivity for broadband antenna design.
The main target of the HAEC box is to allow enhanced
run-time adaptivity together with adaptive high performance
computing in an energy efficient way. Hereby, energy efficiency
is achieved by providing direct links, which reduce the amount
of switches between nodes, and by possibly completely turn-
ing off links depending on the required data rates. In addition,
HAEC box like servers must have real-time operating systems
which can guarantee an extremely small response time, so that
they can cope with the latency requirements of Tactile Internet
applications.
In addition, the HAEC box is expected to have 10
4
times
the computing performance per unit volume of today’s servers
[34]. These 3D chip stacks will not only contain processors but
also memory yielding an extreme powerful server. Equipped
with a very performant energy-adaptive computing platform
and being in proximity, i.e. at the network edge, a system of
unprecedented local compute power can be realized. This will
enable many new Tactile Internet applications.

466 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
B. PHY Design for Ultra-Low Latency
In order to obtain a 1 ms end-to-end latency for the Tactile
Internet, it is important to understand the chain between sensors
and actuators. Fig. 3 shows an exemplary of latency objec-
tives of a mobile-wireless communication system for the Tactile
Internet. The sensor measures, pre-processes and provides its
data to the embedded system controlling the air interface. The
air interface then passes the data through all protocol lay-
ers to the physical (PHY) layer. The same happens at the
receiving side, for example a base-station with a connected
‘(mobile) edge-cloud’, with the data provided to a control
server.
Most of today’s broadband communication systems are
based on Orthogonal Frequency Division Multiplexing
(OFDM) mainly due to its robustness against multi-path
channels [36], [37]. However, to achieve an end-to-end latency
of 1 ms, the physical transmission must have very small
packets which requires a one-way PHY layer transmission of
100 µs as shown in Fig. 3. Since the packet error correction
encoding at the transmitter and the error correction decoding
and detection at the receiver limit the packet size to less than
the target latency, a packet must be smaller than 100 µs packet
duration. In current Long Term Evolution (LTE) cellular
systems, however, the sub-carrier spacing is 15 kHz and the
duration of one OFDM symbol is on the order of 70 µs.
This numerology together with the reference symbols design,
channel estimation, and channel coding requires significant
revision of the cellular PHY for the Tactile Internet, which
might become reality with the 5G system.
One way to overcome these limitations for achieving 1 ms
delay in OFDM based systems is to change the OFDM
numerology, i.e. symbol duration, sub-carrier spacing etc., and
enable high levels of diversity, and fast channel estimation
together with fast channel decoding (e.g. with convolutional
codes). In [24], [38], it has been shown that an OFDM based
system with changed OFDM numerology can achieve reliable
transmission with 1 ms delay.
In general, in OFDM, a cyclic prefix (CP) is added to each
symbol to avoid inter-symbol interference. The low latency
required for Tactile Internet applications, however, demands for
short bursts of data, meaning that OFDM signals with one CP
per symbol may present a prohibitive low spectral efficiency.
Additionally, OFDM square pulse shaping leads to high out-of-
band (OOB) emission which poses a challenge for opportunistic
[37] and dynamic spectrum access [39], [40]. These challenges
motivate an investigation into alternative waveforms to OFDM
for the next generation networks. Hence, alternative multicar-
rier schemes are currently being evaluated as candidates for
the PHY layer of the fifth generation of mobile communication
systems.
For suitably l ow latency i n Tactile Internet applications high
efficiency must be achieved with short burst transmissions. An
efficient way to achieve this is to filter a group of subcarriers to
reduce the OOB emission. In this case, since the bandwidth of
the filter covers several subcarriers, its impulse response can
be short, so that high spectral efficiency is reached in short
burst transmissions. These multiplexing schemes mainly do not
consider any CP, so that they are more sensitive to small time
misalignment.
Various modulation schemes are discussed for 5G, e.g. Filter
Bank Multicarrier (FBMC), Universal Filtered Multicarrier
(UFMC), Bi-orthogonal OFDM (BFDM), etc. [41]–[43].)
Another modulation scheme discussed for 5G, is the
Generalized Frequency Division Multiplexing (GFDM), one
promising solution for the 5G PHY layer [44], [45]. GFDM
is a flexible multicarrier modulation scheme allowing to cover
CP OFDM as a special case. In addition, GFDM is based on
the modulation of independent blocks consisting of K sub-
carriers carrying M subsymbols. The subcarriers are filtered
by circularly shifting a prototype pulse in time and frequency
domain. This makes GFDM a waveform capable to achieve
low OOB emission, which is a major feature for 5G net-
works. For low-latency real-time applications, the signal length
must be reduced [46]. Because GFDM is confined in a block
structure of M × K samples, it is possible to design the time-
frequency structure in a way that the time constraints of Tactile
Internet applications can be achieved. The increased complexity
of GFDM will be manageable with the evolution of electron-
ics. A flexible, customizable Field Programmable Gate Array
(FPGA) platform [47] has been used to develop a GFDM
proof-of-concept and testbed for experimental research.
In addition to the waveform design, further enhanced tech-
niques are needed at PHY layer to cope with Tactile Internet
requirements. Hereby, high reliability can be realized by effi-
ciently using channel coding techniques to exploit diversity
levels [24], [38].
To enable low latency in Tactile Internet applications, fast
decoding techniques are desirable. Compared to channel codes
with iterative decoding, convolutional codes can start decoding
as the data arrives, so that they offer faster receiver processing.
In [48], [49], for example, low-density parity-check convolu-
tional (LDPCC) codes with stringent latency constraints which
allow to combine operation close to the channel capacity with
a low structural latency by using windowed decoding has been
analyzed. Based on optimized decoding schedules for LDPCC
codes, a decoding delay of only 100 ns was observed.
In addition, for achieving very high reliability convolutional
codes do not have error floors (like turbo codes) and they per-
form similarly well for short message sizes, so that they are
often applied to control applications which are also discussed
with the Tactile Internet applications.
Finally, the frame structure for Tactile Internet shall be
designed in a way that it supports fast decoding. This can be
realized by short transmission time intervals and by placing ref-
erence symbols and control information at the beginning of the
frame, so that the channel estimation can lead to quick decoding
and data can be decoded with smallest decoding delays when it
is received.
VI. W
IRELESS ACCESS
Significant changes to the way the wireless access is handled
are also needed. Notably, radio access protocols as well as r adio
resource management require a fresh approach. To this end, we
outline some recent developments in both domains.

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 467
A. Radio Access Protocols
Radio access protocols need to support the range of relevant
frequencies for 5G wireless access. This spectrum flexibility
comprises that different duplexing schemes can be used, like
frequency division duplex (FDD) and time division duplex
(TDD) where the uplink-downlink slit can be configured very
dynamically. Furthermore the access protocols need to support
operation in licensed spectrum, as well as in shared-licensed or
license-exempt spectrum use.
For Tactile Internet applications, a substantial role of radio
access protocols is to provide very low latencies and at the
same time provide very high reliability and availability [50].
A high level of diversity in space and frequency is needed
to provide high reliability levels [24], [38]. Multi-connectivity
is a way to provide this diversity, where the dimensions of
multi-connectivity are frequency and space. Connectivity for
the device (potentially simultaneously) can be established via
multiple frequency layers and / or via multiple sites. On dif-
ferent frequency layers different configurations of the 5G radio
interface can be used, e.g. according to the spectrum proper-
ties. It is desirable that LTE carriers can be tightly integrated
into the multi-connectivity with 5G due to their wide deploy-
ment and high availability, in particular during the early 5G
deployments [51]. Multi-connectivity via multiple sites shall be
flexible to work independent of the type of backhaul connection
that exists between the sites (e.g. fibre cable, wireless links or
copper cables), and should thus be able to handle various back-
haul latencies. A common radio resource control protocol layer
is well suited to integrate different frequency layers and spatial
transmission paths. Besides joint radio resource management
over multiple layers, it provides control plane diversity and fast
control plane switching for robust signaling and radio link man-
agement procedures like mobility. For the use plane it enables
use plane aggregation (for high peak rates) and fast user plane
switching (for reliability). By coordinating connectivity states
on multiple frequency layers (and RATs) via a common con-
trol plane, devices can have optimized sleep modes with very
low energy consumption and fast connectivity activation [52]. A
common control also allows to separate control signaling from
user plane transmission, so that for example system informa-
tion does not need to be transmitted per frequency layer or at
all sites. The overall benefits of an integrated common protocol
layer are increased reliability and diversity, increased user plane
data rates, high spectral efficiency, and better energy efficiency
for the device and the network).
The latency of the radio transmission depends on how
quickly radio resource can be allocated for a device when a data
packet arrives at the radio interface. Network based scheduling
has proven to be an efficient solution, but it comes at the cost
for uplink transmission of a scheduling request and scheduling
grant phase prior to the actual data transmission. Contention-
based transmissions can allow quicker uplink access to the
radio channel. However, if collision probabilities are high large
delays can occur due to backoff and retransmission schemes.
An instant uplink access method is desirable where certain
radio resources can be instantly used by devices while col-
lision probabilities are controlled to remain sufficiently low.
For periodic traffic types persistent scheduling is desirable.
Network based scheduling remains an efficient resource allo-
cation scheme for traffic that does not have prohibitively low
delay requirements.
B. Radio Resource Management
Once the access protocols and mechanisms are determined,
resources need to be allocated which is typically the role of the
RRM protocols. The RRM challenge w.r.t. the Tactile Internet is
that the outlined use cases will require round-trip latencies of as
little as 1 ms as well as high reliability and capacity (data rates).
In some cases, wired access networks are partly meeting the
requirements, but wireless access networks are not yet designed
to match these needs. Scaling-up research in this area will be
essential.
As established above, the 5G access network needs to cope
with one-way latencies of only about 100 µs (Fig. 3). The
resource allocation of the available physical blocks needs to
be done up to 10-times faster than in LTE. The transmission
errors inherent to wireless systems will need to be ironed-out
through careful design. Applications in industrial environments
for example, where potentially huge numbers of robots and
machines will work in close proximity, will create challenging
interference conditions not satisfied by current wireless sys-
tems. Classical approaches for medium-access control to be
reconsidered in such an environment, and new techniques may
be needed to drive latency down to a bare minimum.
New ideas and concepts to boost access networks’ inher-
ent redundancy and diversity need to be researched to address
the stringent reliability requirements of Tactile Internet applica-
tions. Multi-connectivity, as well as tight integration with LTE,
is one approach to increase reliability, in particular if some
of the access layers have limited availability due to challeng-
ing propagation conditions as in high frequencies. Simplified
and fast resource-access schemes as well as efficient signal-
ing protocols need to be designed to optimize the use of the
underlying physical resources. Further, radio resource manage-
ment has to control the interference levels, so that low latency
communication can also be provided with high reliability and
availability.
Radio resource allocation is a key component of radio
resource management. It has direct impact on throughput,
latency, reliability, and QoS for various services. With the
introduction of tactile applications into the 5G ecosystem,
resource allocation becomes particularly challenging as the
available resources would be shared between tactile applica-
tions and other human-to-human (H2H) or machine-to-machine
(M2M) applications, having different and often conflicting
service requirements.
In state-of-the-art LTE networks, packet scheduling is a key
resource allocation technique. Packet scheduling takes into
account the QoS requirements, buffer status reports, and chan-
nel quality of users to maximize the spectral efficiency. Since
the nature of tactile applications is different than H2H and
M2M applications, the scheduling requirements are different as
well. Therefore, using one scheduler for different applications
may not result in optimal resource allocation decision.

468 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
Due to stringent service requirements, radio resources must
be provided on priority for tactile applications. It is particularly
desired that there is no external competition on radio resources
for tactile applications. A robust approach to guarantee this is
to allocate a separate end-to-end “slice” of available resources
to tactile applications, which remains dedicated for any ongo-
ing operation. Since the resource requirements will change over
time, such an end-to-end slicing must be allocated dynamically.
Against this background, a dynamic and flexible resource
allocation scheme is desirable in 5G that maximizes the util-
ity of various applications by ensuring an efficient utilization
of radio resources. Such dynamic and flexible resource alloca-
tion ought to be achieved through end-to-end “virtualization”
of resources, i.e. using the virtualization methods in the SDN
domain and seamlessly connect it to the advanced RRM 5G
schedulers. Such end-to-end “virtualization” enables flexible
slicing, isolation, and customization of resources across differ-
ent vertical applications and user devices. It has the following
key benefits in general and specifically in context of the Tactile
Internet.
• It comes as a natural solution for cloud-RAN approach,
which is increasingly gaining popularity for 5G.
• It ensures resources are allocated to tactile applications on
priority with no external competition.
• It facilitates application of tactile-specific scheduling
algorithms for maximizing the utility of tactile applica-
tions.
• It enables secure resource allocation owing to isolation of
allocated slice from rest of the radio resources.
Based on such virtualized approach and illustrated in Fig. 4,
we propose a resource allocation scheme that consists of the
following key steps:
• Dynamic Resource Slicing. In this step, resources are
allocated to different “slices” as in traditional RRM. For
efficient utilization of scarce radio resources, a combi-
nation of bandwidth-based and resource-based provision-
ing can be used. In bandwidth-based approach, resource
allocation is defined in terms of aggregate throughput
that will be obtained by the flows. On the other hand
in resource-based approach, a fraction of base station’s
resources are allocated to each “slice”. Depending on the
number of active devices and exploiting the channel state
information, the number of radio resources allocated can
be determined statistically.
• Resource Isolation. We propose an end-to-end isola-
tion of radio resources between tactile applications/users
and other applications/users, and connect it to the slic-
ing in the SDN. For other applications/users, resources
can be managed in two distinct ways: (a) isolation of
resources across user devices but not across applica-
tions, and (b) isolation of resources across applications
but not across user devices. Such isolation s cheme not
only guarantees end-to-end availability of resources for
haptic applications, but also allows for customization of
resources for other applications.
• Resource Customization. In this step, resources are cus-
tomized according to the service requirements of different
applications. For example, tactile applications typically
Fig. 4. An illustration of the proposed resource management and end-to-end
resource reservation scheme.
generate a very high packet load and therefore, dynamic
scheduling schemes may not be feasible due to dispropor-
tionally large signaling compared to tactile data. Hence
persistent scheduling schemes are needed to tactile appli-
cations wherein radio resources are allocated for a given
set of sub-frames.
VII. N
ETWORK AND CLOUD DESIGNS
Subsequently, we discuss non access stratum infrastructure
requirements. In the context of the Tactile Internet, the most
important are core networking as well as cloud designs.
A. Core Network
In the Tactile Internet ecosystem, the core network must pro-
vide adequate QoS as well security for tactile applications.
Overall, a thin core network with substantial decrease in the
protocol overhead is desirable. The thinning of the core network
can be achieved by its functional decomposition and moving
some of core functionalities to the access network. This will
reduce the number of nodes in the data path and hence reduce
the end-to-end latency.
The SDN paradigm is particularly attractive for the mobile
core network. The SDN-enabled core network will introduce
the programmability and hence the flexibility to tailor the data
flow inside the core network with respect to the requirements of
tactile applications. This will result in improved service quality
and user experience.
Regarding security, the current IP Security (IPSec protocol
functionalities are sufficient for providing the required secu-
rity in the Tactile Internet. However, the placement of IPSec
far from the tactile edge significantly increases the end-to-end
delays. Therefore, innovative approaches are needed to provide
adequate security for tactile applications with minimal delays
(see Section III-C). Once more autonomous components of the
Tactile Internet are considered (e.g. a self-driving car), then
other security issues will need to be resolved; these are beyond
the scope of this paper.
An important issue from the protocol stack perspective is the
small packet header to payload ratio. For example, the pay-
load of one packet for a 3-DoF haptic stream is only 6 bytes.

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 469
With the transition towards IPv6, the issue becomes particularly
challenging as the header size doubles. Current header com-
pression techniques [53], [54] have been designed for specific
Internet protocols like RTP and might suffer significantly from
unreliable wireless links. Besides, there is an inherent trade-off
between header compression and the inherent delay this process
might entail. Hence, this issue needs to be revisited.
The core Internet latency must be reduced which is vari-
able and largely dictated by queuing delays and geographical
routing policies. Faster forwarding techniques improving the
packet forwarding efficiency by using high-speed cache and
data-flow-based technology can be applied to reduce the latency
of the core Internet. In addition, anycast addressing (as dis-
cussed in IPv6) can be considered for proximity issues in
Tactile Internet. This can be for highly reliable transmissions,
especially when the closest interface fails, or to reduce latency
through proximity.
B. Ultra-Reliable Edge-Cloud
As we will see, the Tactile Internet profits from cloud
technology being natively embedded into the networking tech-
nologies. The advantage of cloud computing and storage is that
resources are not only shared by multiple users but also dynam-
ically reallocated per demand [55]. The move away from heavy
dedicated infrastructure to a shared one maximizes computing
capabilities whilst minimizing power consumption, rack space,
software licenses, among others. This, in turn, allows one to
get applications up and running faster, with improved man-
ageability and less maintenance, and the ability to adjust to
fluctuating and unpredictable data flow demands during run-
time. Technical enablers for cloud capabilities are low-cost
computers/servers; low-cost storage; high-capacity networks;
and secure software capabilities in form of virtualization and
service-oriented architectures [56], [57]. This is really impor-
tant to ensure a scalable and trusted up-take of Tactile Internet
technologies [58].
As discussed before, the distance from the cloud to the wire-
less edge can maximum be in the order of a few kilometers
so that the 1ms challenge can be met. This, however, requires
the cloud application server to be hosted at the edge of the
operator’s core network (at best). This offers some important
advantages: notably, the applications are stored and executed
in a very trusted and secure environment which is particu-
larly paramount to critical Tactile Internet applications. On the
downside, scalability is an issue and so is delay if the server is
placed deep into the CN.
To circumvent this issue, the notion of Cloudlets has recently
been proposed [59]. Here fairly distributed cloud storage and
computing capabilities are placed at the very edge of the
network. In the case of the cellular network, this would be
equivalent to putting it into the RAN or at edge between
RAN and CN. That has interesting implications since con-
tent (Tactile Internet AI engines, see below) and network
management (Cloud-RAN) can be merged into a single infras-
tructure [60]–[62]. Separation in functionality is ensured via
different virtualization approaches [63]. This approach bodes
well for the Tactile Internet since infrastructure investments
Fig. 5. Possible placement options of cloud functionalities.
to upgrade to Cloud-RAN capabilities are already underway.
Again, the downside is the dependence on the operator since
the networking infrastructure will need to be used.
The cloud community, however, has been pushing the
edge-cloud concept even further through the Fog Computing
and Nebula paradigms [ 64]–[67]. The latter refers to a dis-
persed cloud infrastructure that uses (voluntary) edge resources
for both computation and data storage. Early trials were
able to validate an architecture that enables distributed data-
intensive computing through a number of optimizations includ-
ing l ocation-aware data and computation placement, replica-
tion, and recovery. The system was shown to be robust to a
wide array of failures and substantially outperform other cloud
approaches. If security and trust are solved to the same degree
as cellular networks, then the Nebula approach could power
Tactile Internet applications at the haptic user equipment (UE).
The different cloud instantiations are shown in Fig. 5. In
general, it has become obvious that there is an increasing
demand for edge-cloud capabilities, partly because of scal-
ability and partly because of end-to-end delay reasons. For
the cellular 5G networking architecture to scale and support
that trend, 3GPP will need to further relax some requirements
about injecting non-3GPP traffic into the core. First steps had
been done with femtocell technologies, i.e Home eNB (HeNB),
HeNB-Gateway, etc; that work would need to be refined to
facilitate the trusted and authenticated nebula paradigm whilst
not being bound to the current embodiment of the CN. Further
recent work has concentrated to incorporating very hetero-
geneous trusted/untrusted real-time/non-real-time traffic into
3GPP mobile systems [68]–[73].
VIII. A
RTIFICIAL INTELLIGENT CAPABILITIES
Content and skillset data will be transmitted over a signif-
icantly more powerful 5G core network as well as the next
generation Internet. The finite speed of light, however, is the
biggest adversary in facilitating the required real-time experi-
ence advocated by the Tactile Internet. Whilst the advances on
hardware, protocols and architecture are paramount in dimin-
ishing end-to-end delays, the ultimate limit is set by the finite
speed of light.
Breaking the laws of physics not being an option, other
more sophisticated–techniques need to be invoked to facili-
tate the required paradigm shift. In essence, so we argue, this

470 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
will be provided by unprecedented edge artificial intelligence
(AI) engines which are cached and then executed in real-time
close to the tactile experience. The respective components are
discussed subsequently.
A. Advanced Content-Caching
With a proper cloud technology in place, the Tactile Internet
application content needs to be loaded or ported. A typical
example would be an AI algorithm (see below) which is tai-
lored to work in the context of a remote dentistry operation, or
remote car servicing.
These advanced caching techniques and user-oriented traf-
fic management approaches at the edge of the network improve
network performance by de-congestion of the core network and
reduction of end-to-end latency the latter is particularly impor-
tant to the Tactile Internet. Important here is to understand that
the caching for a specific Tactile Internet application will not
help the actual application at hand but other Tactile Internet
applications which run in parallel over the same network; and
vice-versa.
Significant work has been conducted on optimum edge-cloud
caching policies [74]–[79]. Worth highlighting is the recent
work [74] which advocates for proactive caching by using
advanced predictive process capabilities as well as the mas-
sively improved cloud technologies discussed before. With this
approach, peak traffic demands are substantially reduced by
intelligently serving predictable user demands via caching at
base stations and users’ devices. Whilst the advocated approach
pertains to rather long-term windows and file structures, it
forms the foundation for predictive Tactile Internet caching, as
will become apparent in the subsequent section.
B. Tactile Artificial Intelligence Engines
Arguably the most important content to be stored are AI
engines which predict the haptic/tactile experience, i.e. accel-
eration of movement on one end and the force feedback on
the other. That allows to spatially decouple the active and reac-
tive end(s) of the Tactile Internet since the tactile experience is
virtually emulated on either end; this, in turn, allows a much
wider geographic separation between the tactile ends, beyond
the 1ms-at-speed-of-light-limit.
The algorithmic framework is currently based on simple lin-
ear regression algorithms which are able to predict movement
and reaction over the next tens of milliseconds, mainly because
our skillset driven actions are fairly repetitive and exhibit strong
patterns across the six degrees of freedom. When the predicted
action/reaction deviates from the real one by a certain amount
, then the coefficients are updated and transmitted to the other
end allowing for corrections to be put in place before damage is
done at e.g. a deviation of δ. This has been illustrated in Fig. 6.
More sophisticated algorithms have become available, rang-
ing from [80]–[86]. For instance, [82] employed a prediction
method for three-dimensional position and force data by means
of an advanced first-order autoregressive (AR) model. After an
initialization and training process, the adaptive coefficients of
the model are computed for the predicted values to be pro-
duced. The algorithm then decides if the training values need
Fig. 6. Illustration how predictive edge-AI gives the perception of a 1ms delay
whilst the actual latency due to communications can be much larger.
to be updated either from the predicted data or the current real
data.
The AI algorithms, ported through intelligent edge-caching
and stored on appropriate cloud technologies, aid in enabling
the perception of real-time interaction, stabilizing the tactile
system, and consequently enhancing the QoE of the Tactile
Internet user.
IX. E
CONOMIC IMPACT
The Tactile Internet is characterized by extremely low
latency, ultra-high reliability, availability and security. Prior
Internets are only supporting content delivery, whereby content
can be multimedia content or static content, e.g. video streams,
data files, voice, or email. With the Tactile Internet real and
virtual objects can be steered directly at real-time interaction
speeds. Hence, not only content will be transmitted, but con-
trol information. The paradigm shift of the Tactile Internet is
therefore that communication is built for enabling steering and
control. This is a big difference to content delivery in today’s
technologies.
This opens up completely new opportunities for existing
and new applications in many fields, of which some are listed
in Section II. Hence, the Tactile Internet will have a marked
impact on business and society, introducing numerous new
opportunities for emerging technology markets and the delivery
of essential public services.
Being an enabler of skillset delivery, the Tactile Internet
is a very timely technology for service and skillset driven
economies like the ones predominantly found in Europe.
Especially, in the business-to-business ecosystem the Tactile
Internet will be a strong enabler to drive markets for
autonomous cars, remote medical care, and many other indus-
tries. For consumers it will revolutionize the way they interact
with their environment and surroundings. A preliminary mar-
ket analysis as presented in [4], shows the potential of the
Tactile Internet in different markets such as in mobility, man-
ufacturing, event organization, education with entertainment
(edutainment), (health) care, smart grid and further emerging
markets like agriculture, drones, constructions, etc. It has been
shown, that a preliminary market analysis revealed that the
potential market could extend to 20 trillion US dollar world-
wide. This is around 20% of today’s worldwide gross domestic
product (GDP) [87].

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 471
The Tactile Internet will also have a r epercussion on the
telecommunications ecosystem. Notably, as illustrated in Fig. 5,
the content-bearing cloud infrastructure needs to be brought
really close to the edge so as to enable the 1ms latency require-
ment. That, in turn, allows the telecommunications operators
to charge the content providers and thereby opening an addi-
tional stream of revenues. Whilst this happens to a limited
degree with companies like video-streaming service Netflix
today, the Tactile Internet will see an explosion of edge-content
caching through cloud and thereby an exponentially increasing
business-to-business (B2B) market opportunity for operators.
Furthermore, telecommunications vendors may also play an
increasingly direct role in this ecosystem. The reason is because
(at least early) Tactile Internet applications will be less con-
sumer but more B2B driven. B2B customer acquisition however
is more in-line with vendors, and indeed currently vendors push
hard for alliances with various B2B companies, like oil/gas,
transport, etc. Once locked into these markets, the vendors will
be able to procure the best operator deals which in a sense flips
the business model of cellular communications.
Overall, the Tactile Internet opens up massive business
opportunities for the operators, vendors, over-the-top-content
providers; and society at large.
X. C
ONCLUSIONS AND FUTURE WORK
The different embodiments of the Internet will be dwarfed by
the emergence of the Tactile Internet that will be able to deliver
real time control and physical tactile experiences remotely. It
will revolutionize almost every segment of the society. It is
expected that the next generation 5G mobile communications
systems will underpin the Tactile Internet at the wireless edge.
It was thus the aim of this article to investigate the interest-
ing area of 5G and Tactile Internet intersection. After discussing
the exciting Tactile Internet applications, key technical require-
ments have been identified for the Tactile Internet. The paper
covered several technical issues and challenges pertaining to 5G
networks that must be addressed to enable the Tactile Internet
including innovations in end-to-end architecture, revolutions
in hardware and PHY layer transmission novel approaches to
radio access and core protocols as well as to resource manage-
ment and unprecedented edge-cloud and AI capabilities. The
biggest challenge–given the unprecedented requirements on
delay and reliability–will be to ensure tight whilst at the same
time scalable integration of the various technology components
into a single, seamless end-to-end networking experience. Once
achieved, the Tactile Internet will have a massive impact onto
business and society. It will create new opportunities for ven-
dors, operators, content providers and other members of the
service chain.
To sum up, research on the Tactile Internet is still in its
infancy. Besides generating interest in the research community,
the paper aims to open several areas for future research includ-
ing ultra reliable network connectivity in the wireless domain,
reducing end-to-end latency through innovations in air inter-
face, backhaul, core networking, and the Internet, efficient ways
of coding tactile information, and overcoming the physical limit
due to finite speed of light.
A
CKNOWLEDGMENT
The authors thank the researchers at the Collaborative
researches Center 912 “Highly Adaptive Energy-Efficient
Computing” supported by the German Research Foundation
(DFG) for their supports [35]. That work was also partially
funded by the European Commission H2020 5GPP projects
NORMA and VirtuWind. Finally, we would like to thank Stefan
Parkvall, Johan Torsner and Icaro Da Silva for their valuable
comments.
R
EFERENCES
[1] L. Atzori, A. Iera, and G. Morabito, “The Internet of Things: A survey,”
J. Comput. Netw., vol. 54, no. 15, pp. 2787–2805, 2010.
[2] ITU, “World telecommunications/ICT indicators database,” Int.
Telecommun. Union, Tech. Rep., Dec. 2014.
[3] ITU-T, “The tactile internet,” ITU-T technology watch report [Online].
Available: https://www.itu.int/dms_pub/itu-t/oth/23/01/T230100002300
01PDFE.pdf
[4] G. Fettweis. The Opportunities of the Tactile Internet—A Challenge
For Future Electronics [Online]. Available: http://www.lis.ei.tum.de/
fileadmin/w00bdv/www/fpl2014/fettweis.pdf
[5] E. Dahlman et al., “5G wireless access: Requirements and realization,”
IEEE Commun. Mag., vol. 52, no. 12, pp. 42–45, Dec. 2014.
[6] P. Popovski et al., “Scenarios, requirements and KPIs for 5G mobile
and wireless system—deliverable 1.1,” ICT-317669 METIS Project, Apr.
2013.
[7] Z. Roth et al., “Vision and architecture supporting wireless GBit/sec/km2
capacity density deployments,” in Proc. Future Netw. Mobile Summit,
Jun. 2010, pp. 1–7.
[8] NGMN. (2015). 5G White Paper [Online]. Available: https://www.ngmn.
org/uploads/media/NGMN_5G_White_Paper_V1_0.pdf
[9] P. Marsch et al., “5G radio access network design—A brief overview
on the 5G-PPP project METIS-II,” in Proc. Eur. Conf. Netw. Commun.
(EUCNC), Paris, France, Jun./Jul. 2015.
[10] A. Gupta and R. Jha, “A survey of 5G network: Architecture and
emerging technologies,” IEEE Access, vol. 3, pp. 1206–1232, 2015.
[11] M. Weiner et al., “Design of a low-latency, high-reliability wireless com-
munication system for control applications,” in Proc. IEEE Int. Conf.
Comm. (ICC), Sydney, Australia, Jun. 2014, pp. 3835–3841.
[12] N. Nikaein and S. Krea, “Latency for real-time machine-to-machine
communication in LTE-based system architecture,” in Proc. 11th Eur.
Wireless Conf. Sustain. Wireless Technol. (Eur. Wireless), Vienna, Austria,
Apr. 2011, pp. 1–6.
[13] Sercos. (2014). SERCOS News, The Automation Bus Magazine [Online].
Available: http://www.sercos.com/literature/pdf/sercos news 0114 en.pdf
[14] Z. Lin and S. Pearson, “An inside look at industrial ethernet communica-
tion protocols,” Texas Instruments, White Paper, Nov. 2013.
[15] A. Varghese and D. Tandur, “Wireless requirements and challenges in
industry 4.0,” in Proc. Int. Conf. Comtemporary Comput. Informat.,Nov.
2015, pp. 634–638.
[16] J. Arata et al., “A remote surgery experiment between Japan and Thailand
over internet using a low latency CODEC system,” in Proc. IEEE Int.
Conf. Robot. Autom. (ICRA), Apr. 2007, pp. 953–959.
[17] A. Aijaz, M. Dohler, A. H. Aghvami, V. Friderikos, and M. Frodigh,
“Realizing the tactile internet: Haptic communications over next gener-
ation 5G cellular networks,” IEEE Wireless Commun., 2015. [Online].
Available: http://arxiv.org/abs/1510.02826
[18] M. Dohler, “Machine-to-Machine in smart cities & smart grids: Vision,
technologies & applications,” Keynote at WiFlex, Sep. 2013 [Online].
Available: http://bit.ly/1B4rLpY
[19] Z. Shi, H. Zou, M. Rank, L. Chen, S. Hirche, and H. J. Mueller, “Effects
of packet loss and latency on temporal discrimination of visual-haptic
events,” IEEE Trans. Haptics, vol. 3, no. 1, pp. 28–36, Jan./Mar. 2009.
[20] M. Di Luca, T.-K. Machulla, and M. O. Ernst, “Recalibration of mul-
tisensory simultaneity: Crossmodal transfer coincides with a change in
perceptual latency,” J. Vis., vol. 9, no. 12, pp. 1–16, 2009.
[21] K. M. Stanney, R. S. Kennedy, and K. Kingdon, Virtual Environment
Usage Protocols. Boca Raton, FL, USA: CRC Press, 2002.
[22] A. Osseiran et al., “The foundation of the mobile and wireless communi-
cations system for 2020 and beyond challenges, enablers and technology
solutions,” in Proc. IEEE Veh. Technol. Conf. (VTC Spring’15), Jun.
2013, pp. 1–5.

472 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 34, NO. 3, MARCH 2016
[23] ITU-T Telecommunication Standardization Sector of ITU, “Series E:
Overall Network Operation, Telephone Service, Service Operation and
Human Factors,” E.800, Sep. 2008.
[24] O. N. C. Yilmaz, Y.-P. E. Wang, N. A. Johansson, N. Brahmi,
S. A. Ashraf, and J. Sachs, “Analysis of ultra-reliable and low-latency
5G communication for a factory automation use case,” in Proc. IEEE Int.
Conf. Commun. (ICC), London, U.K., Jun. 2015.
[25] H. D. Schotten, R. Sattiraju, D. Gozalvez-Serrano, R. Zhe, and P. Fertl,
“Availability indication as key enabler for ultra-reliable communication
in 5G,” in Proc. Eur. Conf. Netw. Commun. (EuCNC), Jun. 2014, pp. 1–5.
[26] D. Ohmann, M. Simsek, and G. Fettweis, “Achieving high availability
in wireless networks by an optimal number of Rayleigh-fading links,” in
Proc. IEEE Globecom Workshops, Dec. 2014, pp. 1402–1407.
[27] R. Chaudhari, C. Schuwerk, M. Danaei, and E. Steinbach, “Perceptual
and bitrate-scalable coding of haptic surface texture signals,” IEEE J. Sel.
Topics Signal Process. (JSTSP), vol. 9, no. 3, pp. 462–473, Apr. 2015.
[28] E. Steinbach et al., “Haptic Communications,” Proc. IEEE, vol. 100,
no. 4, pp. 937–956, Apr. 2012.
[29] M. O. Ernst and M. S. Banks, “Humans integrate visual and haptic infor-
mation in a statistically optimal fashion,” Nature, vol. 415, no. 6870,
pp. 429–423, Jan. 2002.
[30] Ericsson, “5G Systems,” White Paper, Jan. 2015 [Online]. Available:
http://www.ericsson.com/res/docs/whitepapers/what-is-a-5g-system.pdf
[31] B. Han, V. Gopalakrishnan, L. Ji, and S. Lee, “Network function virtu-
alization: Challenges and opportunities for innovations,” IEEE Commun.
Mag., vol. 53, no. 2, pp. 90–97, Feb. 2015.
[32] B. Nunes, M. Mendonca, X.-N. Nguyen, K. Obraczka, and T. Turletti,
“A survey of software-defined networking: Past, present, and future of
programmable networks,” IEEE Commun. Surveys Tuts., vol. 16, no. 3,
pp. 1617–1634, Third Quart. 2014.
[33] M. Arslan, K. Sundaresan, and S. Rangarajan, “Software-defined net-
working in cellular radio access networks: Potential and challenges,”
IEEE Commun. Mag., vol. 53, no. 1, pp. 150–156, Jan. 2015.
[34] G. P. Fettweis, W. Lehner, and W. Nagel, “Pathways to servers of the
future: Highly adaptive energy efficient computing (HAEC),” in Proc.
DATE Conf., 2013, pp. 1161–1166.
[35] HAEC. (2015). SFB 912: Highly Adaptive Energy-Efficient Computing
[Online]. Available: http://tu-dresden.de/sfb912
[36] J. Bingham, “Multicarrier modulation for data transmission: An idea
whose time has come,” IEEE Commun. Mag., vol. 28, no. 5, pp. 5–14,
May 2009.
[37] M. Mirahmadi, A. Al-Dweik, and A. Shami, “BER reduction of OFDM
based broadband communication systems over multipath channels with
impulsive noise,” IEEE Trans. Commun., vol. 61, no. 11, pp. 4602–4615,
Nov. 2013.
[38] N. A. Johansson, Y.-P. E. Wang, E. Eriksson, and M. Hessler, “Radio
access for ultra-reliable and low-latency 5G communications,” in Proc.
IEEE Int. Conf. Commun. (ICC), London, U.K., Jun. 2015, pp. 1184–
1189.
[39] J. Van De Beek and F. Berggren, “Out-of-Band Power Suppression in
OFDM,” IEEE Commun. Lett., vol. 12, no. 9, pp. 609–611, Sep. 2008.
[40] E. Hossain, Dynamic Spectrum Access and Management in Cognitive
Radio Networks. Cambridge, U.K.: Cambridge Univ. Press, 2009.
[41] B. Farhang-Boroujeny, “OFDM versus filter bank multicarrier,” IEEE
Signal Process. Mag., vol. 28, no. 3, pp. 92–112, May 2011.
[42] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon, “Universal-
filtered multi-carrier ttechnique for wireless systems beyond LTE,” in
Proc. IEEE Globecom Workshop, Dec. 2013, pp. 223–228.
[43] R. Ayadi, M. Siala, and I. Kammoun, “Transmit/receive pulse-shaping
design in BDFM systems over time-frequency dispersive AWGN chan-
nel,” in Proc. IEEE Int. Conf. Syst. Process Control (ICSPC), Nov. 2007,
pp. 772–775.
[44] G. P. Fettweis, M. Krondorf, and S. Bittner, “GFDM—Generalized fre-
quency d ivision multiplexing,” in Proc. 69th IEEE Veh. Technol. Conf.
(VTC Spring’09)
, Barcelona, Spain, Apr. 2009, pp. 1–4.
[45] N. Michailow et al., “Generalized frequency division multiplexing for
5th generation cellular networks,” IEEE Trans. Commun., vol. 62, no. 9,
pp. 3045–3061, Sep. 2014.
[46] M. Nekovee, “Quantifying performance requirements of vehicle-to-
vehicle communication protocols for rear-end collision avoidance,” in
Proc. IEEE Veh. Technol. Conf. (VTC Spring’09), Barcelona, Spain, Apr.
2009, pp. 1–5.
[47] I. Bravo, M. Mazo, J. Lazaro, P. Jimenez, A. Gardel, and M. Marron,
“Novel HW architecture based on FPGAs oriented to solve the eigen
problem,” IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 16,
no. 12, pp. 1722–1725, Dec. 2008.
[48] N. ul Hassan, M. Lentmaier, and G. P. Fettweis, “Comparison of LDPC
block and LDPC convolutional codes based on their decoding latency,” in
Proc. Int. S ymp. Turbo Codes Iterative Inf. Process. (ISTC), Aug. 2012,
pp. 225–229.
[49] T. Hehn and J. Huber, “LDPC codes and convolutional codes with equal
structural delay: A comparison,” IEEE Trans. Commun., vol. 57, no. 6,
pp. 1683–1692, Jun. 2009.
[50] J. Sachs, P. Popovski, A. Höglund, D. Gozalvez-Serrano, and P. Fertl,
Machine-Type Communications. Cambridge, U.K.: Cambridge Univ.
Press, ISBN: 9781107130098, 2016, to be published.
[51] I. Da Silva et al., “Tight integration of new 5G air interface and LTE
to fulfill 5G requirements,” in Proc. IEEE Veh. Technol. Conf. (VTC
Spring’15), May 2015, pp. 1–5.
[52] I. Da Silva, G. Mildh, and M. Gunnar, J. L. Pradas, “Towards an effi-
cient sleeping for 5G devices,” in Proc. IEEE Vehic. Technol. Conf. (VTC
Spring’15), May. 2015.
[53] M. Degermark, “Requirements for robust IP/UDP/RTP header compres-
sion,” Internet Engineering Task Force, RFC 3096, Jul. 2001.
[54] T. Koren, “Enhanced compressed RTP (CRTP) for links with high delay,
packet loss and reordering,” Internet Engineering Task Force, RFC 3545,
Jul. 2003.
[55] B. P. Rimal, C. Eunmi, and I. Lumb, “A taxonomy and sur-
vey of cloud computing systems,” 2009. [Online]. Available:
http://arxiv.org/abs/1007.0066
[56] Q. Duan, Y. Yan, and A. V. Vasilakos, “A survey on service-oriented
network virtualization toward convergence of networking and cloud com-
puting,” IEEE Trans. Netw. Serv. Manage., vol. 9, no. 4, pp. 373–392,
Dec. 2012.
[57] C. Saravanakumar and C. Arun, “Survey on interoperability, secu-
rity, trust, privacy standardization of cloud computing,” in Proc.
Int. Conf. Contemp. Comput. Informat. (IC3I’14), Nov. 2014,
pp. 977–982.
[58] M. Dohler and G. P. Fettweis, “The Tactile Internet—IoT, 5G, and cloud
on steroids,” Telefonica Blog, Oct. 2014.
[59] M. Satyanarayanan, Z. Chen, H. Kiryong, H. Wenlu, W. Richter,
and P. Pillai, “Cloudlets: At the Leading Edge of Mobile-Cloud
Convergence,” in Proc. 6th Int. Conf. Mobile Cloud Comput. Appl. Serv.
(MobiCASE), Nov. 2014, pp. 1–9.
[60] T. Verbelen, P. Simoens, F. De Turck, and B. Dhoedt, “Leveraging
cloudlets for immersive collaborative applications,” IEEE Pervasive
Comput., vol. 12, no. 4, pp. 30–38, Oct./Dec. 2013.
[61] S. Bohez, J. De Turck, T. Verbelen, P. Simoens, and B. Dhoedt, “Mobile,
Collaborative augmented reality using cloudlets,” in Proc. Int. Conf.
MOBILe Wireless MiddleWARE Oper. Syst. Appl. (Mobilware),Nov.
2013, pp. 45–54.
[62] G. Lewis, S. Echeverria, S. Simanta, B. Bradshaw, and J. Root,
“Tactical cloudlets: Moving cloud computing to the edge,” in
Proc. IEEE Mil. Commun. Conf. (MILCOM), Oct. 2014, pp. 1440–
1446.
[63] M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies, “The case for
VM-based cloudlets in mobile computing,” IEEE Pervasive Comput.,
vol. 8, no. 4, pp. 14–23, Oct. 2009.
[64] M. Ryden, K. Oh, A. Chandra, and J. Weissman, “Nebula: Distributed
edge cloud for data-intensive computing,” in Proc. Int. Conf. Collab.
Technol. Syst. (CTS), May 2014, pp. 491–492.
[65] U. Drolia
et al., “The case for mobile edge-clouds,” in Proc. IEEE 10th
Int. Conf. Ubiq. Intell. Comput./Auton. Trusted Comput. (UIC/ATC), Dec.
2013, pp. 209–215.
[66] Q. Li, H. Niu, A. Papathanassiou, and G. Wu, “Edge cloud and underlay
networks: Empowering 5G cell-less wireless architecture,” in Proc. 20th
Eur. Conf. Wireless, May 2014, pp. 1–6.
[67] O. Galinina, A. Pyattaev, S. Andreevy, M. Dohler, and Y. Koucheryavy,
“5G multi-RAT LTE-WiFi ultra-dense small cells: Performance dynam-
ics, architecture, and trends,” IEEE J. Sel. Areas Commun., vol. 33, no. 6,
pp. 1224–1240, May 2015.
[68] S. Andreevy et al., “Understanding IoT connectivity landscape—A con-
temporary M2M radio technology roadmap,” IEEE Commun. Mag.,
vol. 53, no. 9, pp. 32–40, Sept. 2015.
[69] M. Condoluci, M. Dohler, G. Araniti, A. Molinaro, and J. Sachs,
“Enhanced radio access and data transmission procedures facilitating
industry-compliant machine-type communications over LTE-based 5G
networks,” IEEE Wireless Commun. Mag., to be published.
[70] M. Condoluci, M. Dohler, G. Araniti, A. Molinaro, and K. Zheng,
“Towards 5G DenseNets: Architectural advances for effective machine-
type communications over femtocells,” IEEE Commun. Mag., vol. 53,
no. 1, pp. 134–141, Jan. 2015.

SIMSEK et al.: 5G-ENABLED TACTILE INTERNET 473
[71] K. Zheng, S. Ouy, J. Alonso-Zarate, M. Dohler, F. Liu, and H. Zhu,
“Challenges of massive access in highly-dense LTE-advanced networks
with machine-to-machine communications,” IEEE Commun. Mag.,
vol. 21, no. 3, pp. 12–18, Jun. 2014.
[72] K. Zheng, L. Zhao, J. Mei, M. Dohler, W. Xiangy, and Y. Peng, “10Gbps-
HetSNets with millimeter-wave communications: Access and networking
challenges and protocols,” IEEE Commun. Mag., to be published.
[73] H. Y. Lateef, A. Imran, M. A. Imran, L. Giupponi, and M. Dohler, “LTE-
advanced self-organising network conflicts and coordination algorithm,”
IEEE Wireless Commun. Mag., vol. 22, no. 3, pp. 108–117, Jun. 2015.
[74] E. Bastug, M. Bennis, and M. Debbah, “Living on the edge: The role
of proactive caching in 5g wireless networks,” IEEE Commun. Mag.,
vol. 52, no. 8, pp. 82–89, Aug. 2014.
[75] F. Sardis, G. Mapp, J. Loo, and M. Aiash, “Dynamic edge-caching for
mobile users: Minimising inter-AS traffic by moving cloud services and
VMS,” in Proc. 28th Int. Conf. Adv. Inf. Netw. Appl. Workshops (WAINA),
May 2014, pp. 144–149.
[76] Y.-T. Yu, F. Bronzino, R. Fan, C. Westphal, and M. Gerla, “Congestion-
aware edge caching for adaptive video streaming in information-centric
networks,” in Proc. 12th Annu. IEEE Consum. Commun. Netw. Conf.
(CCNC), Jan. 2015, pp. 588–596.
[77] Y. Li, “Coordinated edge caching with request aggregation in radio access
network,” in Proc. 11th Annu. IEEE Consum. Commun. Netw. Conf.
(CCNC), Jan. 2014.
[78] L. Ramaswamy, L. Liu, and A. Iyengar, “Scalable delivery of dynamic
content using a cooperative edge cache grid,” IEEE Trans. Knowl. Data
Eng., vol. 19, no. 5, pp. 614–630, May 2007.
[79] L. Ramaswamy, L. Liu, and A. Iyengar, “Cache clouds: Cooperative
caching of dynamic documents in edge networks,” in Proc. 25th IEEE
Int. Conf. Distrib. Comput. Syst. (ICDCS’05), Jun. 2005, pp. 229–238.
[80] N. Sakr, J. Zhou, N. D. Georganas, J. Zhao, and E. Petriu, “Network
traffic reduction in six degree-of-freedom haptic, telementoring sys-
tems,” in Proc. IEEE Int. Conf. Syst. Man Cybern. (SMC’09), Oct. 2009,
pp. 2451–2456.
[81] N. Sakr, N. D. Georganas, and J. Zhao, “Exploring human perception-
based data reduction for haptic communication in 6-DOF telepresence
systems,” in Proc. IEEE Int. Symp. Haptic Audio-Visual Environ. Games
(HAVE), Oct. 2010, pp. 1–6.
[82] N. Sakr, N. Georganas, J. Zhao, and X. Shen, “Motion and force pre-
diction in haptic media,” in Proc. IEEE Int. Conf. Multimedia Expo,Jul.
2007, pp. 2242–2245.
[83] X. Xu, J. Kammerl, R. Chaudhari, and E. Steinbach, “Hybrid signal-based
and geometry-based prediction for haptic data reduction,” in Proc. IEEE
Int. Workshop Haptic Audio Visual Environ. Games ( HAVE), Oct. 2011,
pp. 68–73.
[84] X. Xu, B. Cizmeci, A. Al-Nuaimi, and E. Steinbach, “Point cloud-based
model-mediated teleoperation with dynamic and perception-based model
updating,” IEEE Trans. Instrum. Meas., vol. 63, no. 11, pp. 2558–2569,
Nov. 2014.
[85] F. Brandi and E. Steinbach, “Prediction techniques for haptic communi-
cation and their vulnerability to packet losses,” in Proc. IEEE Int. Symp.
Haptic Audio Visual Environ. Games (HAVE), Oct. 2013, pp. 63–68.
[86] F. Guo, C. Zhang, and Y. He, “Haptic data compression based on a linear
prediction model and quadratic curve reconstruction,” J. Softw.,vol.9,
no. 11, pp. 2796–2803, 2014.
[87] Worldbank. (2014). Gross Domestic Product [Online]. Available:
http://databank.worldbank.org/data/download/GDP.pdf
Meryem Simsek received the Dipl.-Ing. degree
in electrical engineering and information technol-
ogy and the Ph.D. degree from the University of
Duisburg-Essen, Essen, Germany, in 2008 and 2013,
respectively. In 2013, she was a Postdoctoral Visiting
Scientist at Florida International University, Miami,
FL, USA. Since 2014, she has been a Group Leader
with the Vodafone Chair Mobile Communication
Systems, Technical University Dresden, Dresden,
Germany. Her research interests include radio
resource management in heterogeneous wireless net-
works, 5G wireless systems, game theory, reinforcement learning, and the
tactile internet. As of June 2015, she chairs the IEEE ETC Technical
Subcommittee “The Tactile Internet.” She was a recipient of a scholarship from
the German National Academic Foundation which is granted to the outstand-
ing 0.5% of all students in Germany. She was also the recipient of the IEEE
Communications Society Fred W. Ellersick Prize in 2015.
Adnan Aijaz (M’14) received the B.E. degree
in electrical engineering (telecom) from National
University of Sciences and Technology (NUST),
Karachi, Pakistan, and the M.Sc degree in mobile
and personal communications and the Ph.D. degree
in telecommunications engineering from King’s
College London, London, U.K., in 2011 and 2014,
respectively. After a post-doc year at King’s College
London, he moved to Toshiba Research Europe Ltd.,
where he currently works as a Research Engineer.
Prior to joining King’s, he worked in a cellular indus-
try for nearly 2.5 years in the areas of network performance management,
optimization, and quality assurance. His research interests include 5G cellular
networks, machine-to-machine communications, full-duplex communications,
cognitive radio networks, smart grids, and molecular communications. He has
served as the Symposium Co-Chair for the IEEE SmartGridComm’15. He is
serving or has served on the TPC of various IEEE Conferences and workshops
including VTC Spring’15, ICC’16, CROWNCOM’16, WCNC’16, and VTC
Spring’16.
Mischa Dohler (S’01–M’03–SM’07–F’14) is a Full
Professor of Wireless Communications with King’s
College London, London, U.K., the Head of the
Centre for Telecommunications Research, the Co-
Founder and a Member of the Board of Directors
of the smart city pioneer Worldsensing. He is a
frequent keynote, panel, and tutorial speaker. He
has pioneered several research fields, contributed to
numerous wireless broadband, IoT/M2M, and cyber
security standards, holds a dozen patents, organized,
and chaired numerous conferences, has more than
200 publications, and authored several books. He acts as policy, technology, and
entrepreneurship adviser. He has talked at TEDx and has had TV & radio cov-
erage. He is a Distinguished Lecturer of the IEEE. He is the Editor-in-Chief of
the Transactions on Emerging Telecommunications Technologies (Wiley) and
the EAI Transactions on the Internet of Things.
Joachim Sachs received the Diploma degree from
Aachen University, Aachen, Germany, and the
Doctorate degree from the Technical University of
Berlin, Berlin, Germany. Currently, he is a Principal
Researcher with Ericsson Research working on future
wireless communication systems. In 2009, he was a
Visiting Scholar at Stanford University, Stanford, CA,
USA. Since 1995, he has been active in the IEEE and
the German VDE Information Technology Society
(ITG), where he currently co-chairs the technical
committee on communication networks and systems.
Gerhard Fettweis (S’84–M’90–SM’98–F’09)
received the Ph.D. degree from RWTH Aachen
University, Aachen, Germany, in 1990. After one
year at IBM Research in San Jose, CA, USA, he
moved to TCSI Inc., Berkeley, CA. Since 1994,
he has been a Vodafone Chair Professor with TU
Dresden, Dresden, Germany, with 20 companies
from Asia/Europe/US sponsoring his research on
wireless transmission and chip design. He co-
ordinates two DFG centers at TU Dresden, namely
cfaed and HAEC. He is a Member of the German
academy Acatech. In Dresden, h e has spun-out 11 start-ups, and setup funded
projects in volume of close to EUR 1/2 billion. He has helped organize IEEE
conferences, most notably as the TPC Chair of ICC 2009 and TTM 2012,
and as the General Chair of VTC Spring 2013 and DATE 2014. He was the
recipient of the Stuart Meyer Memorial Award from IEEE VTS.