Professional Documents
Culture Documents
com
ScienceDirect
ICT Express 4 (2018) 203–208
www.elsevier.com/locate/icte
Abstract
Device-to-device (D2D) communication is expected to play a significant role in upcoming cellular networks as it promises ultra-low latency
for communication among users. This new mode may operate in licensed or unlicensed spectrum. It is a novel addition to the traditional cellular
communication paradigm. Its benefits are, however, accompanied by many technical and business issues that must be resolved before integrating
it into the cellular ecosystem. This paper discusses the main characteristics of D2D communication including its usage scenarios, architecture,
technical features, and areas of active research.
⃝c 2017 The Korean Institute of Communications and Information Sciences (KICS). Publishing Services by Elsevier B.V. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Device-to-device communication (D2D); Cellular network; 5G; Resource management; LTE direct
Contents
∗ Corresponding author.
E-mail addresses: uditnarayankar@gmail.com (U.N. Kar), debarshisanyal@gmail.com, debarshi.sanyalfcs@kiit.ac.in (D.K. Sanyal).
Peer review under responsibility of The Korean Institute of Communications and Information Sciences (KICS).
https://doi.org/10.1016/j.icte.2017.08.002
2405-9595/⃝ c 2017 The Korean Institute of Communications and Information Sciences (KICS). Publishing Services by Elsevier B.V. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
204 U.N. Kar, D.K. Sanyal / ICT Express 4 (2018) 203–208
1. Introduction
3. Architecture
In terms of spectrum usage, D2D communication is primar- 3GPP Rel. 12 of the LTE-Advanced standard specifies a
ily classified into two types. They are inband and outband [1]. general concept of proximity-based services (ProSe) that allows
physically close devices to discover themselves and commu-
nicate via direct links [8]. ProSe is meant for public safety
3.1.1. Inband D2D communication communication as well as commercial applications although
Here, cellular communication and D2D communication use the emphasis in Rel. 12 is on the public safety only. D2D
the same spectrum licensed to the cellular operator. The li- discovery and D2D communication are defined as a support
censed spectrum may be either divided into non-overlapping for ProSe. It is also known as LTE Direct since it supports
portions for D2D and cellular communication respectively direct communication between UEs using licensed spectrum
(overlay) or may not be divided at all (underlay). Overlay and the global LTE ecosystem. Three scenarios for D2D
scheme is easier to implement but underlay scheme leads to communication are considered: (1) all UEs involved in D2D
opportunistic and hence, more efficient spectrum use and more communication are within network coverage, (2) only some of
profit to operators. the UEs in D2D communication are within network coverage,
and (3) none of the UEs in D2D communication are within
network coverage.
3.1.2. Outband D2D communication
A highly simplified model for D2D communication based
Here, D2D communication uses unlicensed spectrum on ProSe reference architecture (non-roaming case) is shown
(e.g., the free 2.4 GHz ISM band or 38 GHz mm Wave in Fig. 2. The BS or eNB, as it is called in 3GPP, connected
band) where cellular communication does not occur. It helps to the Evolved Packet Core (EPC), can communicate with a
in eliminating the interference between D2D and cellular users UE directly using cellular communication. Additionally, UEs
although interference is still present from other electronic can communicate via direct D2D links. In terms of channel
devices (like Bluetooth and WiFi) operating in this band. In structure, the direct link between two UEs is called a sidelink
fact, operators can control interference when using licensed which can operate by frequency division duplex or time division
spectrum but that is infeasible for outband scheme. Outband duplex. The UEs on being powered up first synchronize with the
technology is further divided into controlled and autonomous eNB or other UEs. For this purpose, several synchronization
types. In the former, the radio interface for D2D communica- signals are defined in 3GPP Rel. 12. Coming back to the
tion is controlled by the cellular network while in the latter, architecture in Fig. 2, various ProSe applications (APPs) can be
the cellular network controls only the cellular communication installed in a UE and they may exchange data with the ProSe
leaving the control of D2D communication to the users. APPs in a remote ProSe APP server. When a UE wants to
206 U.N. Kar, D.K. Sanyal / ICT Express 4 (2018) 203–208
communicate with its peer UE, the ProSe APP in it requests more complex algorithms (in terms of computational and power
for expression codes of itself and its target peer from the budgets) in UEs unlike the case in resource-constrained sensors.
server. Alternatively, a UE can obtain the expression codes Researchers have proposed some physical layer [9] and MAC
from the Proximity Function in the eNB. After the expression layer schemes [10] for synchronization in D2D communication.
codes are retrieved, the UE initiates the discovery procedure by
announcing its own expression code or inquiring if the target 4.2. Peer discovery
UE (identified by the retrieved expression code) is present.
After device discovery, the UEs can communicate directly. In Looking at the demand of D2D network, there should be
terms of air interface for D2D signal and data transmission, re- an efficient method for discovering peers. This means a UE
sources are allocated either by the eNB or by the UEs randomly should be able to discover other nearby UEs quickly and with
from a preconfigured pool of resources. D2D communication low power consumption. From user perspective, there are two
occurs using open-loop communication in layer 1, i.e., a D2D types of peer discovery techniques, restricted and open [11].
receiver does not send any feedback (including channel state In the first case, devices cannot be discovered by the end users
information and acknowledgments) to a D2D transmitter [8]. without their permissions. In the second case, devices can be
discovered whenever they lie in the proximity of other users.
3.3. Single-hop and multi-hop networks From network perspective, peer discovery can be controlled
lightly or tightly by the BS [11]. In a multicell network, it is
Generally a D2D link connects a transmitter UE with its very difficult to get cooperation from adjacent BSs, making
intended receiver UE resulting in a single-hop communication. peer discovery a challenging job [12]. Incentive-based schemes
One can also have a multi-hop network composed of D2D links, (i.e., game-theoretic frameworks) may be investigated as a
reminiscent of a mobile ad hoc network (MANET). In a multi- probable solution.
hop D2D network, the intermediate UEs act as relays either
between a BS and a UE or between two UEs (refer to Fig. 1). 4.3. Mode selection
A variant of the first scenario could be a cooperative cluster of
UEs in which the BS transmits a data item to the cluster head A pair of UEs that have discovered themselves are potential
which then groupcasts it to other UEs in the cluster (perhaps candidates for a D2D communication. But performance-wise,
with network coding to improve throughput). 3GPP Rel. 13 cellular communication may be more preferable if, for example,
enables UE-to-network relay while 3GPP Rel. 14 adds support the direct channel is more noisy. Mode selection is concerned
for vehicular communication (i.e., high speed and high density with choosing the right mode – cellular or D2D – for communi-
of nodes) based on D2D technology. cation between two UEs to achieve some performance objective
like high spectral efficiency, low latency, or low transmit power.
4. Challenges and ongoing research Mode selection can be done by the network or by the UEs.
To formulate the mode selection problem, one may associate
We will now discuss the various technical aspects and the a decision variable with each UE, that captures the selected
corresponding challenges of D2D communication in wireless mode and then add a variety of objectives and constraints. A
networks. simple objective could be that the channel gain of the selected
mode should be higher than that of other possible modes. More
4.1. Synchronization sophisticated objectives like optimal spectrum reuse, weighted-
sum-rate maximization, etc. could also be used. Constraints
In a typical cellular network, UEs achieve time and fre- could be minimum QoS at receiver, maximum transmit power,
quency synchronization using periodic broadcasts from the BS. etc. Thus mode selection is generally coupled with power
Devices in D2D communication can also synchronize with the control. The analysis could be done using instantaneous system
same broadcasts so long as they belong to the same BS. The information (which may be difficult to acquire in practice)
situation gets complicated in the following cases: (1) UEs be- or statistical system information leading to decisions that are
long to different BSs that may not be themselves synchronized, respectively optimal at a given instant or optimal in an average
or (2) some of the UEs are in the coverage of the network and sense over a longer duration [13].
some outside the coverage, and (3) all UEs lie outside network
coverage [9]. Synchronization among UEs is beneficial for D2D 4.4. Resource allocation
communication because it helps a UE to use the right time
slot and frequency for discovering and communicating with its Radio resource (e.g., subcarriers) allocation is an important
peer and thus engage in more energy-efficient communication. step (especially, in inband mode) in creating and maintaining
Note that global synchronization among all UEs in a network direct links between D2D pairs in a cellular network. A simple
may not be required for D2D communication; rather local syn- but general resource allocation framework is proposed in [14]
chronization among neighboring devices is sufficient. Although for inband multicell architecture: in overlay, uplink spectrum is
one may adopt the synchronization protocols proposed for divided into two orthogonal portions with fraction η assigned
MANETs and wireless sensor networks, D2D communication to D2D communication and 1 − η to cellular communication;
usually requires more accurate synchronization and also allows in underlay, the spectrum is divided into B bands and D2D
U.N. Kar, D.K. Sanyal / ICT Express 4 (2018) 203–208 207
UEs can randomly and independently access β B (β ∈ [0, 1]) models in the context of D2D communication include how
of them. The optimal values of η and β are computed by D2D UEs in a cluster may buy or sell data items [19], how
assuming that the UEs are distributed according to a random cellular users can sell their bandwidth to D2D UEs [20] and
spatial Poisson point process and mode selection is based on how D2D-capable UE pairs can auction their cellular resources
a UE’s distance from its intended receiver, and then optimiz- to other waiting cellular users and themselves opt for D2D
ing some performance objective like a joint function of the communication [21].
rates achieved by cellular and potential D2D users. Different
resource allocation schemes can be designed by changing the 4.8. Security
optimization objectives and adding various constraints.
D2D communication affords stronger anonymity and data
4.5. Interference management
privacy compared to conventional cellular communication since
the data are not stored at a central location. However, various
In inband communication, cellular and D2D links may inter-
common attacks like eavesdropping, denial of service, man-
fere with each other based on how they share the frequencies.
in-the-middle, node impersonification, IP spoofing, malware
In outband communication, D2D links suffer interference from
attack, etc. can paralyze D2D links. Users would also like to
each other as well as from other devices operating in the
protect their privacy, e.g., by restricting the availability of their
same band [15]. Interference can be reduced if UEs transmit
sensitive personal data. The same lack of a central authority
at lower power levels which might, however, affect the QoS
makes it difficult to implement security and privacy measures.
at the receiver. Thus interference-aware resource management
Authors in [3] model threats in a three-dimensional space:
is an involved optimization problem. Often it is cast as a
weighted-sum-rate maximization problem subject to maximum (1) whether the attacker is internal or external, (2) whether the
transmit power and minimum QoS constraints or as a trans- attacker is active (e.g., it modifies in-transit data) or passive
mit power minimization problem subject to a minimum QoS (e.g., it only snoops on data), and (3) whether the attack is local
constraint [16]. Power control, in addition to interference mit- or extended across the network. Several proposals to safeguard
igation, leads to energy-efficient operation which is one of the D2D networks are reviewed in [3,22].
goals of next generation wireless networks. Careful scheduling
of transmissions also helps to minimize interference. Suitable 5. D2D communication and 5G networks
modulation and coding schemes (adaptively chosen based on
The upcoming 5G network is expected to support aggregate
channel quality) and hybrid automatic repeat request (which
data rates (i.e., total amount of data the network can serve,
is the combination of automatic repeat request and forward
measured in bits/sec/area) 1000 times that of the current 4G
error correction) increase the robustness of the transmitted
network [23]. The spectral efficiency (number of bits transmit-
signal against noise. Mode selection, resource allocation and
ted per Hz) and energy efficiency (number of bits per Joule)
interference minimization are closely related and often jointly
optimized. For these three kinds of problems, several central- should be 10 times higher than those of 4G network, data
ized, distributed and hybrid algorithms have been proposed but rates for mobile users should be multi-Gbps and end-to-end
research is still active [16]. latencies around 1 ms. 5G network will essentially contain
an umbrella of technologies including heterogeneous network
4.6. D2D with mobility (HetNet) (i.e., use of various radio access technologies, multiple
backhaul techniques and a hierarchy of cells—macro, pico,
Most D2D-related research has focussed on static users femto), massive MIMO (i.e., large antenna arrays at BS to
while cellular networks essentially cater to mobile users. serve many users concurrently), cognitive radio network (CRN)
More analysis is needed to understand how the performance (where secondary users opportunistically use the primary users’
gains auger in dynamic scenarios (from pedestrian to vehic- spectrum), millimeter wave (mmWave) spectrum (works over
ular speeds) and what interference handling and handover 30–300 GHz frequency) and D2D communication [23]. The last
mechanisms are needed as UEs move within and across one is fundamentally important in 5G despite the challenges of
cells [17]. Multi-hop D2D communication also poses many implementation [4]. The mmWave spectrum in 5G can be used
challenges [18]. to form short-range D2D links between UEs. Since mmWave
suffers low multi-user interference, many mmWave D2D links
4.7. Pricing can operate concurrently, thus improving network capacity.
Secondary users in a CRN can also use D2D communication
This is one of the most pressing issues to cellular operators. to avoid interference to primary users. D2D communication
The difficult question is how to control the direct link between complements HetNets and massive-MIMO-enabled BSs in im-
the devices and how to charge the users. Various pricing models proving spectral efficiency and data rates. Additionally, MIMO
are explained in [4]. For example, operators may use UEs as antennas embedded in UEs will increase noise resilience and
relays for other users and may give financial incentives to the system capacity by exploiting diversity and multiplexing gains.
relay UEs [4]. Operators can also provide chargeable services D2D-based relays with MIMO-enabled devices can also en-
like security during D2D communication [4]. Other economic hance system capacity significantly [1].
208 U.N. Kar, D.K. Sanyal / ICT Express 4 (2018) 203–208