Energy Efficiency Maximization for Cooperative and Non

TRANSACTIONS ON EMERGING TELECOMMUNICATIONS TECHNOLOGIES
Trans. Emerging Tel. Tech. 0000; 00:1–30
DOI: 10.1002/ett
AAA
Energy Efficiency Maximization for Cooperative and
Non-cooperative OFDMA Cellular Networks – A Survey
Álvaro Ricieri Castro e Souza1∗ , José Roberto de Almeida Amazonas1 and Taufik Abrão2
1
Escola Politécnica da Universidade de São Paulo 2 Departamento de Engenharia Elétrica da Universidade Estadual de Londrina
ABSTRACT
With the increasing data rate necessity in modern cellular networks, power consumption grows continuously for network
operators or mobile users. Under this scenario, negative implications arise, as for example economical and environmental,
and also reduce user experience quality, as battery powered devices cannot operate for long time intervals without been
charged. So, the development of energy-efficient resource allocation algorithms for modern cellular networks, as fourth
generation (4G) systems, becomes a fundamental task. Since orthogonal frequency division multiple access (OFDMA)
is the most popular multiple access technique for modern cellular communications, this survey provides a guideline to
energy-efficient approaches for OFDMA-based systems, discussing techniques and evaluation modes for energy-efficient
systems. As cooperative communication has the potential to reduce power consumption and is included as features in 4G
c 0000 John Wiley & Sons, Ltd.
standards, this technique is also investigated in this survey. Copyright ⃝
∗
Correspondence
BBB. E-mail: [email protected]
1. INTRODUCTION
and environmental issues; the first one, originated by the
pollution in generating energy, and the second one, as
In this survey, we review some techniques for energy
the operation costs become higher for telecommunication
efficiency (EE) maximization in non-cooperative and
operators[3].
cooperative OFDMA scenarios, indicating also some
As pointed out by several works, the energy con-
techniques to implement the optimization problems and
sumption for information and communications technology
documents from standardization organizations for system
(ICT) is becoming a significant percentage of the total
parameters. This way, it is possible to observe the research
power consumption [3, 4], which impacts for both users
scenario and analyze other strategies and possibilities in
and operators: At the user side, as the battery technologies
EE maximization for fourth generation (4G) systems. The
evolved in a much slow rate than the offered services
necessity of EE-based resource allocation comes with the
[5], the lifetime of the equipments becomes critical. For
increasing number of subscribers and mobile multitask
operators, the increased power consumption results in
devices, such as smartphones and tablets, which demands
higher operational costs [6]. Another perspective is the
higher capacity from cellular networks. This can be seen
environmental issue, since electricity production results
by the increasing spectral efficiency (SE) for the downlink
in pollution, as the Vodafone case pointed in [6]. In
case, which evolved from 0.05 bps/Hz in the GSM (2G)
order to balance the increasing data rate and the energy
systems [1] to a peak SE of 30 bps/Hz for the LTE-
consumption, recently the energy efficiency metric has
A 11 (4G) systems [2], which implies in significant
been proposed as an important figure of merit for efficient
power consumption increase while creates economical
wireless communication systems and networks. Defined
c 0000 John Wiley & Sons, Ltd.
Copyright ⃝
Prepared using ettauth.cls [Version: 2012/06/19 v2.10]
1
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
as the ratio between data rate and power consumption,
deep fading while reducing power consumption, which
this metric indicates how efficiently the system transform
can be translated in EE improvement. When considering
power in transmitted data. Obviously, this metric could
EE optimization for cooperative OFDMA, various aspects
impact negatively on important system parameters [1], as
can be optimized, such as time/frequency sharing, relay
for example the spectral efficiency.
placement, detection protocol and so on.
As well known, the orthogonal frequency division
With this multiple possible approaches for EE
multiple access (OFDMA) technique is the most popular
maximization, this survey reviews basic concepts
system model for high data rate communication systems,
in
such as WiMAX and LTE-Advanced (LTE-A), two
and energy-efficiency, in order to effectively discuss
candidates for 4G systems [4]. This is due to several
representative EE-maximization techniques, providing
advantages of OFDMA, such as the robustness to the
an interesting tool to discuss energy efficiency in fourth-
inter-symbol interference (ISI), caused by multipath
generation cellular system. The survey is organized as
propagation, in which it is specially impactful when
follows: In section 2 we present the OFDMA system
symbol rate is higher, and the higher diversity dimensions,
model and cooperative networks techniques, as well as
as we can consider frequency, multiuser and time
terminology deployed in this survey. In section 3, we
dimensions in the resource allocation strategies [7, 8].
describe methods for EE maximization in non-cooperative
In OFDMA, the total bandwidth is split in narrowband
OFDMA networks, while in section 4 we do the same for
subchannels, which turns possible to lower the symbol
cooperative OFDMA systems. In section 5, we discuss
rate maintaining the user data rate. With this, it is
methodologies proposed for simulation and evaluation
possible to obtain a period of symbol higher than the
methods for EE problems, also listing documents
channel delay spread, reducing the multipath problem and
from different standardization workgroups for system
increasing communication reliability. In the context of EE
parameters and scenarios choices. Finally, in section 6 we
optimization, in general the literature considers the power
conclude this survey and present some research challenges
and subcarrier allocation as optimization variables, while
and topics of interest.
OFDMA
systems,
cooperative
communications
inserting other features, as sleep time and modulation
order.
Besides all the features provided by OFDMA systems,
2. SYSTEM DESCRIPTION
one problem that any wireless communication system
faces is the channel conditions, mainly represented by
This section intends to give an overview of the main
fast fading and path-loss, which are inherent to the
techniques used in 4G systems and discussed in this paper.
propagation environment [8, 9]. The problem of path-loss
Although the energy efficiency paradigm is proposed
becomes even worst in high-frequency communications,
for several multiple access and network topologies, only
as the 5 GHz carrier frequency presented in both LTE-A
the OFDMA model is presented since the focus of this
and WiMAX, and in densely constructed areas, such as
survey is 4G based systems. In this way, we overview
metropolitan areas. For the fading case, when the channel
OFDM/OFDMA systems, cooperative networks and the
experiences deep fading, it is impossible or impractical
energy efficiency definition.
to maintain communication. In order to reduce these
problems, one of the most promising techniques is the
cooperative communications. Under this paradigm, relay
stations (RSs) are placed at the cell to improve coverage
and/or capacity, mainly for users in coverage holes or celledge area, retransmitting the received signal from mobile
stations (MSs) or base-stations (BSs) to destination. This
way, it is possible to reduce the effects of path-loss
and form a virtual antenna array (like virtual MIMO)
[8] to provide spatial diversity, providing robustness to
2
2.1. OFDM/OFDMA
The orthogonal frequency division multiplexing (OFDM)
technique consists in splitting a user data stream into
several sub-streams, which are sent in parallel on several
subcarriers, obtained by splitting the total bandwidth in
narrower channels. Considering that X is the set of
symbols to be transmitted, each symbol Xi modulates the
ith subcarrier, with |X| ≤ N , where N is the number
of available subcarriers. The rationale of this approach
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
is to increase the individual symbol time (Ts,i ) in each
symbol, there are N samples that are ISI-free at the cost
subcarrier without increasing the total time to transmit X
of a data rate penalty imposed by the redundancy. In
(Ts ), in such a way that Ts,i is higher than the channel
summary, the CP is inserted after the IDFT at source-side
delay spread Td , which is fundamental to reduce the effects
and removed before the DFT at receiver-side.
of inter-symbol interference (ISI), without affecting the
The main parameters associated to the OFDM system
data rate as Ts remains the same [7]. The symbol time in
are: a) Number of subcarriers (N ), which must satisfy the
each subcarrier is given by Ts,i = N Ts , resulting that N
channel delay spread constraint; b) Cyclic prefix (CP) size,
is defined in order to achieve Ts,i ≫ Td , i.e. N ≫ Td /Ts .
which must be at least equal to the number of channel
Figure 1 shows an OFDM block diagram and illustrates
multipath copies.
the overall process. The symbols are modulated in
Besides all the advantages described, there are two
the frequency domain by taking the N -point inverse
main drawbacks in OFDM systems [7, 10]: the highly
discrete Fourier transform (IDFT) of X. As the data
precise frequency synchronization needed and the peak-to-
stream X is serial, first it passes through the serial-to-
average power ratio (PAPR). The first one is caused by the
parallel converter, then the IDFT is applied, obtaining
sinc functions in frequency domain at the detection, since
the time-samples x, which are serialized again by the
the inter-carrier interference (ICI) is zero only when the
parallel-to-serial converter. After that, the data stream is
frequency is perfectly synchronized; this synchrony can
transmitted over a wireless channel with impulse response
be lost due to imperfect oscillators and the Doppler effect,
given by h = [h0 , h1 , · · · , hj , · · · , hv ], where v is the
generated by user mobility. The second one is intrinsically
length, in samples, of the channel delay spread. At the
caused by the multicarrier strategy. In time-domain, the
detector side, the received time signal y passes through
OFDM symbol is composed by many narrowband signals,
a serial-to-parallel converter, then the DFT is applied
which could result in peaks of power much larger than
and the frequency samples Y are then serialized and
b Note
detected, generating the estimated symbols vector X.
the average power, making the power amplifier to operate
that OFDM transmission technique is that the various
of efficiency. As described in the following, the OFDMA
subcarrier signals are generated digitally and jointly by
technique results to be the solution, at least at the user’s
an IFFT algorithm in the transmitter and that their
side, of the second problem.
in the nonlinear region, incurring in distortion and loss
spectra strongly overlap on the frequency axis. As a
As OFDM is a multiplexing strategy, it can be combined
result, generating the transmit signal is simplified and the
with different multiple access (MA) dimensions, as time
bandwidth efficiency of the OFDM/OFDMA systems is
(TDMA), code (CDMA) or frequency (FDMA). As the
significantly improved.
4G standards consider the multiple access in frequency,
As the DFT/IDFT is used, it is necessary that the signal
we will focus on the OFDMA system model. In OFDMA,
and channel convolution be circular, which implies that
for each user is allocated a number of subcarriers in
x or h has to be periodic. On the other hand, despite
a given time slot, which are updated in the following
Ts,i ≫ Td some symbols can still be affected by the
slot [7]. In this way, it is possible to recognize at least
multipath propagation [7]. In order to overcome the effects
two diversity dimensions: time and multiuser. The time
of these two situations a cyclic prefix (CP) is inserted,
dimension comes from the fact that if the channel quality
which consists of copying the Ng ≥ v last (first) samples
is poor for a user in all subcarriers in a given time-slot, it is
of x to the beginning (ending) of the OFDM symbol, which
possible to make this user wait for another time slot. The
equates the linear and circular convolutions results. As
multiuser dimension can be explored from the possibility
a result, the CP insertion makes the received signal ISI-
to choose which users allocate which subcarriers given a
free. Considering that the signal at the source has N + Ng
specific metric, as, for example, the instantaneous channel
samples, the convolution with the fading channel results
state information (CSI).
in a signal with N + Ng + v samples. By discarding the
Since in OFDMA users share the subcarriers, it is
first Ng samples, which are corrupted by the delayed
necessary to define from a time-slot basis which users
samples of the previous signal, and the v last samples of
use which subcarriers in which time-slot and with which
the received signal, which interfere with the next OFDM
power. On the downlink, the base-station (BS) defines
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
3
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Figure 1. OFDM system block diagram, where X is the set of symbols to be transmitted, h is the multipath channel coefficients and
b is the set of estimated symbols.
X
this mapping and informs users about it in order to
power than if all the subcarriers were assigned for only one
detect the correct subcarriers. On the uplink, if distributed
user at each time-slot.
solutions to the allocation problem are implemented,
each user must inform the BS the allocated subcarriers;
if a centralized solution is considered, the BS must
proceed as in the downlink case. Depending on the
metrics to be optimized, as data rate maximization,
power consumption minimization or fairness provision,
different subcarrier/time/power allocation algorithms can
be deployed. The algorithms are not specified along the
OFDMA [7], and each implementation decides how to deal
with this problem. If we consider K users, N subcarriers,
time slot equal to the symbol time Ts and power p ∈
[0, Pmax ], where Pmax is the maximum allowed power,
the allocation problem becomes extremely complex to be
solved in real time. That being so, commercial systems
such as LTE-A and WiMAX use time-slots with multiple
symbols and group subcarriers to form subchannels, to
reduce the allocation problem complexity [7].
The OFDMA system also solves another problem seen
in OFDM [7], i.e., the uplink’s PAPR problem. As the
power amplifier of user devices cannot be as efficient as of
BS’s due to cost reasons, the PAPR on the uplink becomes
even worse. However, in OFDMA systems as each user
2.2. Cooperative Cellular Networks
In order to overcome the destructive nature of wireless
channels, one of the most promising techniques is the
cooperative communications paradigm. In this scenario
one or more equipments, called relays, retransmit the
signal received from the source to the destination,
providing spatial diversity if the source-destination link is
available or improving coverage in the other case. As the
path-loss is inversely proportional and non-linear to the
source-destination distance∗ [9, 11], if the user uses the
relay(s) in a multi-hop communication to the destination
the power spent with propagation losses can be reduced,
increasing the efficiency in terms of power consumption.
The spacial diversity can be seen as virtual multiple input
multiple output) (MIMO) [9], where the relays, plus the
direct transmission, act as virtual antennas. This is an
interesting method to provide spatial diversity in cellular
communications, allowing the increase of the cell coverage
(macro-diversity); moreover, since the required antenna
separation to provide uncorrelated signals for MIMO
uses only a portion of the available subcarriers and the
PAPR is proportional to the number of used subcarriers,
users experiment lower PAPR and can be able to use less
4
∗
Considering that the path-loss is commonly defined as Lo d−n , where Lo is a
constant depending on system parameters, d is the communication distance and
n is the path-loss exponent, which depends of the propagation scenario and in
general is given by 2 ≤ n ≤ 6
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Figure 2. Basic cooperative strategies.
systems is at least half wavelength (micro-diversity) [7],
coverage to otherwise uncovered holes or to provide higher
in cases where carrier frequency is in the hundreds of
data rates where it is necessary.
megahertz range or less, it becomes unrealistic to obtain
Based on Fig. 2, it is possible to infer that the relay
uncorrelated signal replicas in mobile cellular devices,
can be recognized as a network equipment by BSs
such as cell phones.
and mobile stations (MSs) or not [13, 14]. The first
A simple example of a cooperative communication can
condition, presented in Figs. 2.b and 2.d, is called non-
be seen in Figure 2, where the source S communicates with
transparent relaying. In this way, the RSs are considered
the destination D using a single relay RS. There are several
as a BS, communicating directly with the mobile users.
ways to do it [8]:
This mode is used to extend coverage, since the link
a) S transmits to D in the first time-slot, and RS overhears
and then retransmits in the second time-slot (Fig. 2.a);
b) S transmits to RS in the first time-slot, and then RS
transmits to D in the second time-slot (Fig. 2.b);
c) S transmits to D in the first time-slot, RS transmits
the overheard message in second time-slot, and S also
transmits a new (or copied) message in the second timeslot (Fig. 2.c);
MS-BS is unavailable [15]. The second mode, presented
in Figs. 2.a and 2.c, is called transparent relaying
and corresponds to the case where MSs and the BS
communicate directly, and the RSs just overhear the
transmission and then collaborate by retransmitting. This
model is used to increase throughput/reliability or to
reduce power consumption, since the MSs are already
covered by the BS [15].
To retransmit the information, there are also two types
of protocols [8, 16]. The first one refers to the case in which
d) S transmits to RS in the first time-slot, and both RS and
the RSs decode the received signal before retransmitting
S transmits to D in the second time-slot (Fig. 2.d);
it, called regenerative protocols, while in the second case
Another classification of relay stations (RSs) concerns
the type of equipment [12]. If the equipment is dedicated
to retransmission and installed as part of the cell
infrastructure, it is called fixed relay station (FRS). On the
other hand, if the user equipments are able to retransmit
information of other users, they are called mobile relay
stations (MRSs). Although both implementations are
possible, the fixed approach is the choice for the 4G
standards, given that there are dedicated power supply and
dedicated equipment for relay operation, while the mobile
devices would have to share limited power and time to
act as relay, despite the higher diversity that could be
achieved. Another advantage of fixed relays is that they
can be installed in a planned way, aiming to provide better
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
the signal is analogically processed before retransmission,
being called non-regenerative protocols. The best known
implementations are the decode-and-forward (DF) and
amplify-and-forward (AF) protocols, respectively. The
main advantages of regenerative protocols are the
possibility of more sophisticated processing, including remodulation and coding changes, aiming to not propagate
noise from the received signal, while the drawbacks are
the processing cost, extra latency and possibility of wrong
detection. As pointed by [17], the AF protocol presents a
higher diversity order than the DF protocol.
Finally, there is the method to separate the backhaul
and access links. As shown in Fig. 3, there are three
different links: the direct link, where MS and BS
5
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
communicate directly, the backhaul link, where RS
This evaluation changes the paradigm from the data rate
and BS communicate, and the access links, where MS
efficiency, mainly measured by the spectral efficiency
and RS communicate. As pointed out in [18], the
(SE), to the cost of the data rate in terms of power
backhaul and access links cannot operate in the same
consumption. The basic EE metrics units are bits per Joule,
time/frequency/space due to self-interference. In this way,
i.e., the number of transmitted effective bits per Joule
it is necessary to separate these links in one dimension. The
consumed, and Joules per bit, i.e., the amount of energy
possible separations are [19]:
consumed to transmit one effective bit. In this context, an
• Inband relaying: in this case, backhaul and access
links are separated by sharing the resources in time
domain, or by sufficient antenna separation at relay
stations, to avoid self-interference;
effective bit refers to a bit that carries effective information
from the source to the destination, discarding protocol
headers, signaling data and redundant bits inserted by error
detection codes. The EE metric can be evaluated as
• Outband relaying: one RF channel is used for each
link, or the original RF channel is split in two fixed
portions of the total bandwidth.
[
f (γ)r
G
L
ξ=
=
P
M ϱp(γ) + pc
bits
Joule
]
(1)
where G is defined as the effective throughput, or goodput;
P is the total power consumption; L is the number
of information bits per packet; M is the number of
bits transmitted per packet, including protocol headers,
signaling data, redundant bits, and information bits as
well; f (γ) is the efficiency function; γ is the signal to
interference plus noise ratio (SINR); r is the achievable
data rate in bits per second; ϱ is the power amplifier
inefficiency; p(γ) maps the SINR into transmission power
in Watts; and pc refers to the circuit consumption in Watts.
The efficiency function is used to discard reception
Figure 3. Link scheme for cooperative communications.
errors, since the EE numerator refers to goodput. In
general, this function approximates the package success
Regarding the commercial scenario, both LTE-A and
rate (PSR) [22], since using the exact formulation incurs in
WiMAX consider only fixed relays and regenerative
the trivial solution p(γ) = 0 [23]. As f (γ) is a cumulative
protocols [19, 20, 16], since the adaptive modulation
density function (CDF), the two main constraints are:
and coding (AMC) feature is only possible with the
{
deployment of regenerative protocols. For LTE-A, both
f (γ) =
transparent and non-transparent modes are allowed, and
0,
γ<0
1,
γ→∞
(2)
also inband and outband relaying, the last with a second RF
channel. Just to clarify the common LTE-A nomenclature,
in this case there are two types of RSs: The Type-I RSs
are non-transparent equipments, deploying regenerative
protocols and operating at the MAC layer, just like a
regular BS from the MSs point of view, while the TypeII RSs are transparent equipments that can operate with
regenerative or non-regenerative protocols [21].
To avoid the trivial solution p(γ) = 0, it is defined
that f (0) = 0. The parameters of f (γ) must reflect the
system model in order to obtain the correct results. When
considering low order phase-modulation (BPSK, QPSK),
the approximation f (γ) = (1 − e−γ )M is a well-known
choice [23],[24].
The data rate r can be considered fixed or a function
of the achieved SINR; a common approach to set r is by
2.3. Energy Efficiency
means of the Shannon capacity equation [25], obtaining
(3):
The energy efficiency (EE) metric is used to evaluate
the efficiency with which the communication system
r(γ) = W log2 (1 + γ)
[bits/second] ,
(3)
converts consumed energy into effective transmitted data.
6
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
where W is the system bandwidth.
Last, the power parameters ϱ and pc are system depen-
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
considering a group of subcarriers or even the resource
blocks (RBs), which are portions of subcarriers and time.
dent, and can be obtained in standards or approximated
As an example of suboptimal approaches, the authors
once some specifications, as for example cell size and
of [31] describe the optimal RB/power allocation problem
deployment scenario, are provided.
The earlier works in EE focused in maximizing
for the uplink of a single-cell OFDMA system and find
(
)
K
that the complexity is about O KNRB
, where NRB is the
the energy efficiency without taking into account QoS
number of resource blocks per allocation interval. As this
constraints [23],[24],[22], such as data rate and delay. As
complexity is prohibitive for commercial implementations,
QoS is fundamental in modern cellular standards, recent
the idea behind the paper is to develop iterative suboptimal
works include such metrics as constraints, investigating
algorithms, which allocate one RB per time, repeating
the impact of them and proposing techniques to optimize
the process until there are no RBs and available users to
EE, considering QoS aggregation and QoS over time,
allocate to them. Two algorithms are developed to iterate
i.e., maintaining QoS over time-slots and not necessarily
over the available RBs, one considering a determinate
instantaneous QoS. The impact of using EE metrics is also
RB order and the other evaluating what is the best RB
analyzed in terms of trade-offs between well-established
to be allocated in the current iteration. Numerical results
metrics, as spectral efficiency (EE-SE) [26, 27, 28],
demonstrated that the proposed algorithms can provide a
operational/deployment cost, allocated bandwidth [1] and
sub-optimal EE with lower complexity, but a gap higher
delay [29].
than 10% is obtained when the number of users and
subcarriers are increased (mainly for the second parameter)
2.4. Notation
Tables I and II present the terminology used in this survey
for the acronyms and symbols, respectively.
in the EE achieved by the optimal solution and the
proposed algorithms.
In [4] authors formulated the EE maximization problem
for single-cell OFDMA systems in both uplink and
downlink cases. In the downlink description, weights
are used in data rate to provide fairness/priority, while
3. EE IN OFDM/OFDMA
As OFDMA is the main multiple access technique for 4G
cellular systems and EE is becoming an important system
evaluation criterion, there are several works in literature
considering EE maximization in OFDMA. The works
mainly focus on allocating subcarriers and transmission
power to obtain higher EE, using frequency/multiuser/time
diversity in the algorithms, which is in general named
scheduling algorithm. As the power/subcarrier (or even
subchannel) allocation problem is NP-hard [30] [31], in
general sub-optimal strategies have been usually proposed.
Some insights about resource allocation in OFDMA
networks can be found in [32], where a survey on uplink
resource allocation for OFDMA systems is carried out.
Despite the fact that EE is not taken into account, some
of the conclusions can be used to plan the EE resource
allocation framework. Examples of that include the impact
of buffer model, instantaneous or ergodic QoS metrics,
how to define system data rate (if continuous or given by
the AMC), as well as individual subcarrier allocation or
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
a max min problem in the uplink case is considered
to optimize the lower individual EE, while the QoS
metrics are represented by minimum rate criterion in both
cases. The problem is presented with some sub-optimal
algorithms, which virtually allocate the worst subcarrier
to it user and measuring the urgency in allocating more
subcarriers. The results obtained with the suboptimal
approach results are close to the obtained with optimal
solution, which has been obtained testing all possible
configurations of subcarrier allocation.
Maintaining the single-cell scenario and considering the
downlink, the authors of [33] describe an alternative way
to increase EE. Including the circuit power from MSs in
receiving mode on the EE formulation, eq. (1), the problem
now consists in minimizing the time that the MSs are
in active mode, so that the transmissions for each user
is concentrated in few time slots and the power spent
in receiving mode is saved. After time-slot allocation, a
power control algorithm allocates the necessary power to
each user, which can be changed to consider other metrics
without affecting the first algorithm, as pointed out by the
7
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Table I. Table of Acronyms.
Acronym
AF
AMC
ARQ
BER
BS
BPSK
CAPEX
CDF
CDMA
CF
CoMP
CP
CSI
DF
DFT
DS-CDMA
EE
FDMA
GSM
HARQ
ICI
ICT
IDFT
IRI
ISD
ISI
LTE(-A)
MA
Description
Amplify and forward
Adaptive modulation and coding
Automatic repeat request
Bit error rate
Base station
Binary phase-shift keying
Capital expenditure
Cumulative density function
Code division MA
Compress and forward
Coordinated multipoint
Cyclic prefix
Channel state information
Decode and forward
Discrete fourier transform
Direct sequence CDMA
Energy Efficiency
Frequency Division MA
Global System for Mobile Communications
Hybrid ARQ
Inter-carrier interference
Information and communications technology
Inverse DFT
Inter-relay interference
Inter-site distance
Inter-symbol interference
Long Term Evolution (-Advanced)
Multiple access
Acronym
MAC
MAI
MCS
MIMO
M-QAM
MRC
MS
NP
OFDM
OFDMA
OPEX
PAPR
PSR
QoS
QPSK
RB
RF
RO
RS
SE
SER
SISO
SF
SG
SINR
SNR
TDMA
WiMAX
Description
Media Access Control
Multiple access
Modulation and coding scheme
Multiple input multiple output
M -symbol quadrature amplitude modulation
Maximal-ratio combining
Mobile station
Non-polynomial
Orthogonal Frequency Division Multiplexing
Orthogonal Frequency Division MA
Operational expenditure
Peak-to-average power ratio
Package success rate
Quality of service
Quadrature phase-shift keying
Resource block
Radio frequency
Relay ordering
Relay station
Spectral Efficiency
Symbol error rate
Single-input single-output
Shorten and forward
Stochastic geometry
Signal to interference plus noise ratio
Signal to noise ratio
Time division MA
Worldwide Interoperability for Microwave Access
Table II. Table of symbols.
Symbol
α
γ
γTH
ξ
ϱ
d
f (γ)
G
h
hi
K
L
Lo
M
N
n
Ng
NRB
Ns
Description
Factor of relay placement adjust
SINR or SNR, depending on definition
Minimum γ to cooperate
Energy-efficiency utility function
Power amplifier inefficiency
Distance between two nodes
Efficiency function
Effective throughput (Goodput)
Channel impulse response vector
ith channel impulse response
Number of users
Number of information bits per packet
Path-loss constant
Number of bits per packet
Number of subcarriers
Path-loss exponent
Size of the cyclic prefix
Number of resource blocks
Number of subchannels
Symbol
O(.)
P
p
p(γ)
pc
Pmax
R
r
Ts,i
Ts
Td
v
W
X
x
Xi
Y
y
Description
Asymptotic complexity order
Total power consumption
Allocated transmission power
Power consumption to achieve γ
Circuit power
Maximum power allowed for a given user
Cell radius
Data rate
Time of each symbol Xi
Time to transmit X
Channel delay spread
Length, in samples, of the channel delay spread
Available bandwidth
Set of symbols to be transmitted
Signal to be transmitted in time domain
ith individual symbol to be transmitted
Signal received in frequency domain
Signal received in time domain
authors. So, it is possible to adapt a algorithm like the one
coding scheme (MCS) between two subcarriers allocated
in [4] to further improve EE.
to the same user, in order to reduce the consumed power
In [30], it is considered a multicell downlink OFDMA
and, consequently, the interference generated. Reducing
scenario with AMC, where the allocated subcarriers are
the interference, other users can reduce transmission
swapped in an intermediate step in order to guarantee QoS
power over those subcarriers, which can lead to an
to more users, and then the power allocation is rerun,
overall power reduction, called as ”virtuous loop“. The
verifying if it is useful to change the modulation and
algorithm presents significant gains in terms of spectral
8
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
and energy efficiencies regards conventional approaches,
order to investigate if that is the optimum scenario for
such as round-robin scheduling.
EE maximization, in [37] is considered the downlink of
Maintaining the multicell scenario, the authors of
a single-cell single-input single-output (SISO) OFDM
[34] also consider the problem of subcarrier and power
system where the users are able to share the subcarriers.
allocation for maximizing EE, but with the N subcarriers
Analyzing the proposed problem, the authors conclude
grouped in Ns subchannels and deploying pricing
that even with the possibility of subcarrier sharing,
strategies. As pointed out in [35], pricing mechanisms
the optimum solution is that only one user uses each
are an effective way to reduce transmission power and
subcarrier at a given time, which corresponds to the
therefore multicell interference, resulting in higher EE.
OFDMA approach. With this information, the problem is
The subchannel/power allocation is separated in two steps:
then simplified to the OFDMA case and numerical results
in the first one, equal power is considered, and each
is obtained by simulation, without any QoS guarantee. It
subchannel is allocated to the user with higher SINR.
is also presented a simplification to obtain the optimal
To avoid underserved users and improve fairness, at the
power/subcarrier allocation, using bisection method over
final of this step an adjustment is made to distribute
the transmission power to solve the optimization problem.
subchannels from users with two or more subchannels
There are also other techniques that could be deployed
to users with any assigned subchannel. The second
with OFDMA to further improve EE. One of them is the
step consists in allocation of power to the assigned
coordinated multipoint (CoMP), which allows the base
subchannels, considering the bits per Joule metric with
stations in neighbor cells to jointly define which users each
pricing coefficient.
one will cover and also which subcarriers will be allocated,
The Joule-per-bit metric is used in [3], where a single-
aiming to improve coverage and reduce interference,
cell downlink OFDMA system case is investigated. As
while providing ways to reduce transmission power and
the authors discuss, referencing [36], the bit-per-Joule
improve EE. Another common technique is multiple-input
metric presents “lack of linearity between the consumed
multiple-output (MIMO), where the stations are equipped
energy and the transmitted power”, which can lead
with multiple antennas, for transmission and/or reception,
to incorrect conclusions; despite of this, the bit-per-
in order to provide spatial diversity and/or multiplexing
Joule metric is widely adopted. The problem consists
gains.
in allocating subcarriers and bits (consequently power)
When considering the CoMP technique, system
to each subcarrier in order to provide a minimal SE
capacity is directly proportional to the backhaul link
to the system and minimal data rate for each user,
capacity, as discussed in [38], which implies that the best
considered as the QoS metric. The problem is formulated
performance is obtained with infinite capacity backhaul
using a set of modulations, i.e. discrete data rate,
link to exchange the necessary information about users,
and the fractional problem developed is solved using
including CSI and data to be transmitted. Obviously,
Dinkledbach’s parametric approach, in an iterative way.
the deployment and energy cost of such backhaul links,
The interesting result presented is that Joule-per-bit
even if they are dedicated to this function, is prohibitive,
decreases when system loading increases, i.e. the system
so it is necessary to determine the impact of limited
becomes more efficient when system loading increases.
backhaul capacity. Thus, in [38] each BS has a limited
This can be explained by: a) more users more data rate is
backhaul link and needs to decide with which BSs is
necessary, which reduces the impact of circuit power at the
best to cooperate in order to reduce inter-cell interference
BS; b) multiuser diversity improve with more users; and c)
and maximize capacity. By considering power/subcarrier
the low number of users considered initially. The EE-SE
allocation, zero-forcing precoder coefficients, fairness and
tradeoff is investigated by increasing minimum data rate
backhaul allocation, an heuristic algorithm is developed,
per user and the overall SE, and the numerical results show
which first allocate the backhaul link and then iterates over
that increasing the rate/SE requirements the EE is reduced.
the user scheduling considering power/precoder fixed and
All
the
aforementioned
works
consider
an
power/precoder optimization for fixed scheduling.
OFDM/OFDMA system where the users do not share the
From the EE perspective, this is done in [39]
same subcarrier/subchannel at the same cell and time. In
considering multicell downlink OFDMA and a central
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
9
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Table III. Representative papers for EE in OFDM/OFDMA non-cooperative systems
Year
2011
Paper
[4]
2011
[36]
2012
[30]
2012
[31]
2012
[32]
2012
[33]
2012
[3]
2013
[37]
Contribution
Power/subcarrier allocation for single-cell downlink/uplink case, considering EE fairness with minimum rate constraint.
Suboptimal iterative solutions are presented for both cases;
Framework for energy-efficiency communication systems, with a realistic BS power model and also a traffic model that
allows to analyze the energy efficiency algorithms in large scale areas, such as a country, which has different user densities
and data necessities;
Smart allocation algorithm, which makes a subcarrier rearrangement between users to admit more users, and also a
power/bit reallocation to reduce power consumption;
Suboptimal iterative algorithm that allocates the subcarriers/power in a fixed or optimized order to the user that provides
the higher EE gain in each iteration;
Survey on uplink resource allocation for OFDMA networks, considering centralized/distributed scheduling, multicell
scenario and describing open trends;
The scheduler concentrate the MSs RBs in fewer time slots, in order to make MSs sleep in some slots to save the power
spent in receiving mode;
EE maximization by allocating subchannels and bits to each user with overall SE and individual data rate constraints. The
EE-SE tradeoff is investigated by increasing the rate/SE constraints;
Demonstrate that OFDMA has higher EE than OFDM systems for downlink case, i.e., sharing the subcarriers do not
increase EE, and a bisection method to solve subcarrier/power allocation;
unit that processes the information received from all
increased for the same SNR, which results in lower EE.
BSs and then communicates to each one, exchanging
As multiple antennas in mobile devices can be hard
the necessary data through limited backhaul channels.
or even impossible to implement, considering multiple
With this limitation, the tradeoff between EE, backhaul
antennas only for the BS equipment is also an interesting
capacity and network capacity is analyzed considering
scenario, as the case described in [42]. Considering the
that for each user must be assured a minimum data rate
EE for the MSs, the numerical results show that when the
and the backhaul link presents a maximum data rate
number of antennas is increased, the EE are improved,
constraint, considering as optimization parameters power,
mainly when the number of MSs increases. Also, the
subcarriers (user selection) and zero-force beamforming.
jointly power/subcarrier optimization increases the EE
The optimization problem is solved in three steps, one
when compared with only power allocation.
for each variable, and a iterative algorithm executes these
three steps. It is analyzed the EE behavior under different
Table III summarizes representative works and results
analyzed in this section.
maximum transmission power, backhaul capacities and
number of active users, and is demonstrated that EE is
not directly proportional to backhaul capacity, due to the
power consumption cost necessary to increase backhaul
capacity that overcomes the capacity gain.
For MIMO systems, there are several considerations
on energy-efficient system design. For example, in [40]
the authors analyze the SNR gains obtained with multiple
antennas in both receiver and transmitter equipments,
considering different cell radius scenarios and number of
antennas. As pointed out by the numerical results, when the
number of antennas is increased, the SNR gain is increased
4. EE IN COOPERATIVE OFDMA
As already mentioned, cooperative communications have
several benefits that can result in higher energy efficiency,
not only in terms of power consumption but also
in deployment costs, since relay stations have less
functionalities and processing power. When considering
the deployment of cooperative networks in cellular
scenarios, there are several aspects to consider, for
example:
for the same cell radius. However, since the energy efficient
design must consider operational costs, as circuit power
and computational complexity, these results require a more
accurate analysis. One case is analyzed in [41], considering
MIMO deployment in pico-cell scenario. In this case,
the results demonstrated that the power consumption is
• relay station placement, which includes the placement location and the number of stations installed;
• cooperation
protocol:
regenerative
or
non-
regenerative;
• inband or outband channel operation;
• which relays cooperate and how to assign then;
10
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Besides these four basic aspects, there are others
be made in both domains, amplifying the received signal in
that can be included, as frequency reuse patterns and
the time domain or taking the DFT to access the received
CoMP, and the considered optimization variables from the
signal in each subcarrier/subchannel and adaptively
OFDMA case, as transmission power, subcarrier and time
amplifying each one, retransmitting the signal after
allocation, now including the RS nodes in the optimization
redoing the IDFT. From the energy efficiency perspective,
problem. In this way, there are several opportunities in
it is necessary to investigate if the obtained gain can
the EE maximization for cellular cooperative OFDMA
compensate the complexity/energy cost associated to the
networks. In this section, we will discuss the four aspects
both DFTs application. When considering the regenerative
mentioned and then present salient techniques that can
protocols, the only possibility is the frequency domain
be combined with cooperative networks to provide higher
operation, since the information is modulated in frequency
energy efficiencies.
domain.
There are few papers comparing regenerative and
4.1. Retransmission protocols and operation
modes
Analyzing first the retransmission protocols, we point
out two possibilities: non-regenerative and regenerative
protocols. If we consider only the commercial aspect, the
LTE-A standards only consider the regenerative protocols
for implementation, as some features like the adaptive
modulation and coding (AMC) is only possible at the
relays when they are able to fully decode the received
signal and then adapt the MCS to the conditions of the
next hop. As the proposal of this survey is the literature
overview, we also consider the non-regenerative protocols
as a possible choice.
The two most common protocols are the amplify-andforward (AF) and decode-and-forward (DF), respectively
a non-regenerative and a regenerative protocol. The basic
approach for both protocols is to process the received
signal and then retransmitting the information to the
destination node. As pointed out in several papers, this
approach can result in poor retransmitted information if
the received signal is highly corrupted, which is in general
caused by poor SINR. If we consider AF, under low SINR
the amplified signal consists mainly of interference and
noise, while for DF the bit error rate (BER) becomes
significantly high and the retransmitted data does not
correspond to the information sent by source. One possible
way to overcome this problem is to adopt a threshold to
decide if the RS is able to cooperate. In this case, the QoS
metric that the RS must obey is mapped in SINR threshold
γTH and if the achieved SINR is greater than γTH , the RS
can cooperate.
non-regenerative protocols for OFDMA from the energy
efficiency perspective. Looking at the simplest possible
case, as described in figure 2, there are some results. For
example, in [44] the energy efficiency of AF and selective
DF, which is a protocol that only relays when the MS-RS
link succeeds† , is compared, considering the existence or
absence of return channel, which indicates the necessity
of RS cooperation. In the proposed scenario, there is not a
best protocol for all cases, and this choice depends on some
factors, as equipment distances and network topology.
The authors of [45] consider a similar case, but using
only AF and selective DF and a circuit power consumption
model that reproduces the extra cost of DF operation.
It is proposed the optimization of the modulation in
order to minimize the energy-per-bit metric in a threenode system, for different node distances. Under this
model, the selective DF outperforms AF protocol, which
is the opposite result obtained of the first case. This fact
demonstrates that both protocols and their variants must be
further investigated in the OFDMA cellular case in order to
provide accurate results, since the topology is not restricted
to the three node case.
In addition to classical AF and DF, there are other
relaying protocols that can be used. For example, an
extension to the AF protocol called shorten-and-forward
(SF) protocol has been proposed in [46]. The main concern
in SF protocol is that when the signal is transmitted over
two hops and no detection is performed by the RS, the
cyclic prefix can be not enough to overcome the delay
spread inserted by both hops. In this way, finite response
(FIR) filters are used at the RS to reduce the increased
When choosing the relaying protocol we also define if
the relay operation is made in time or frequency domain
[43]. For the non-regenerative protocols, the operation can
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
†
In general, the failure or success of a given link is observed in terms of the
received SINR, for a given threshold, which results in the model described in the
second paragraph of subsection 4.1
11
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
signal length. The SF approach can result in EE gains
since it is possible to use a shorten CP, which reduces the
throughput penalty in OFDMA, and also be determinant
Frequency
for multihop AF systems, to avoid ISI for relayed users.
When considering the separation of backhaul and access
links, there are also two modes: inband or outband.
In the inband method, the time/frequency resources are
t1
divided by the two links in the time domain, resulting
that the existing infrastructure can be maintained. The
Backhaul/direct Access/direct
links
links
a)
t2
Time
main drawbacks are synchronization issues and possible
interference if this task was not well solved. On the other
hand, the outband mode deploys a complete separation
of the links, as for example the use of a second
carrier frequency for one of the links. In this case, the
synchronization problem is not an issue, since both links
are separated, at the cost of new equipments and, in the
case of an extra carrier frequency, the cost of this extra
frequency band. These two schemes are presented in Fig. 4,
with a third model which is proposed in [47] and described
in this section.
b)
In [47], the authors investigate the backhaul/access
links separation considering capacity enhancement for
a dual-carrier OFDMA system with DF relays. There
are four proposed carrier allocation modes, two inband
and two outband, and an interesting result is that both
Backhaul/direct Access/direct
links
f2
links
Frequency
schemes achieve almost the same throughput CDF and
relative gain. This result comes from the fact that the
allocated resources in each case are almost the same:
regarding the inband case, all carriers are allocated half
of the total time, while in the outband case only one
carrier is available, but is allocated all the time. Besides
this equality, it is also pointed out that some factors
can result in capacity loss for the inband case, given
c)
t2
f
0
Backhaul/direct
links
f1
Time
Figure 4. Schemes for access/backhaul links separation: a)
inband; b) outband; c) mixed inband/outband proposed in [47].
by synchronization, backhaul/access switching and the
necessity of transmitting control data from BS to RS using
resources that are reserved for user data. A capacity loss of
about 8 ∼ 12% is expected. In order to enhance EE, it is
also investigated the resource sharing optimization, given
by dynamically sharing the resources for the backhaul link.
Numerical results demonstrates the effectiveness of this
optimizations, mainly for the worst users.
In [48], the authors consider the downlink of a multicell
OFDMA system, in which each cell is aided by some
RSs. The proposed relay protocol separates the backhaul
and access links in time, splitting equally the access time,
and the two-hop and one-hop communication in frequency,
12
by allocating orthogonal subchannels to each mode.
The optimization problem considers the transmission
mode (one or two-hops), subchannel allocation and relay
processing design, while the BSs transmission power is
considered fixed. The optimal solution consists in trying
all the possible modes and subchannel allocations, and
then evaluating the relay processing design, which has
been demonstrated no loose in optimality compared to
trying all the combinations of these three variables at the
same time. Besides this simplification, the complexity is
still prohibitive; hence, these problems can be sequentially
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Table IV. Representative papers for retransmission protocols and operation modes
Year
2010
Paper
[45]
2011
[47]
2011
[52]
2012
[44]
2012
2012
[46]
[48]
2012
[49]
2012
[50]
Contribution
Energy-per-bit minimization by optimizing the modulation order for AF and selective DF deploying or not MRC for both
protocols;
Analysis of inband and outband modes with different combinations of carrier usage, resource sharing optimization and
proposition of a mixed inband/outband mode;
Demonstrate that the two-hop model (Fig. 2.b) results in higher EE than the multicast two-hop model (Fig. 2.c), while for
capacity the opposite occurs;
EE for AF and selective DF with network coding and with/without return channel and fairness, and the EE contours, which
are the data rates that results in the same EE;
Application of shorten-and-forward to reduce ISI due to the extra delay spread caused by the two-hop communication;
Optimization of subchannel allocation, transmission mode and relay processing to increase power efficiency, comparing
AF and DF protocols for single- or multi-relay selection;
Bandwidth/time sharing scheme for type-I relays in LTE-A systems for power minimization, considering equipment
prioritization and data rate requirements;
Dynamic time allocation for backhaul and access links and resource reuse in the access link for downlink aiming to
maximize capacity under fairness constraint;
solved in a single-cell way by a heuristic strategy. It
share the subchannels, i.e., spatial reuse of resources is
is demonstrated that the DF protocol results in higher
permitted only in the access link. The numerical results
power efficiency, measured in bits per Watts, than the two
demonstrate the achievable gains with the combination
protocols based on AF protocol with multiple or single
of these techniques, which can be maximized if relay
relay selection strategies.
placement optimization is also considered. In [51] the time
Besides the pure inband/outband strategies, in [47]
sharing is also mentioned, but no numerical results are
and [16] a mixed inband/outband separation model is
provided. When considering the outband mode with an
proposed. As the backhaul link can become the system
exclusive carrier to the backhaul link, one possible problem
bottleneck, mainly when the number of relays/relayed
arises when a few number of users uses RSs, resulting in
users increase, it is defined that the second carrier is
resource wasting. To avoid this, in [47] is proposed that
used only for backhaul and direct links, while the first
some direct users can be allocated in the backhaul link if it
carrier is used by these two links and the access link,
is underloaded.
using time division between backhaul and access links.
In [49] it is discussed relaying techniques based in
This alternative is proposed to increase the efficiency
bandwidth sharing and time sharing, initially considering
of link usage, as in pure outband systems if one link
cases with only one relay and then a problem with multiple
is underloaded the spectral resource is wasted, and to
relays are considered, but only the single-relay case is
prevent backhaul bottleneck, providing extra bandwidth.
considered. An interesting result pointed out is that the
As pointed out in [47], this method outperforms both
time sharing system can be equivalent to the bandwidth
inband and outband cases.
sharing if average power metrics are considered, while if
Basically, the only decision to make in outband mode
peak power constraints are considered, the performance
is which link to allocate in each available carrier. The
can be worst in terms of power minimization. All
common approach is to define that one carrier supports
the proposed problems include equipment prioritization,
the direct/access links and the second is used only for
which turns possible to increase the power cost of specific
the backhaul link, as they can limit system performance.
equipments, as for example the MSs which has limited
On the other hand, the inband mode has some parameters
batteries. The case of multiple relays are simulated and a
to choose, as for example the backhaul/access links
flexible bandwidth sharing model is proposed to allocate
proportion, as discussed in [49] and [50]. In [50], it
resources for BS-RS links, which is demonstrated to
is analyzed the dynamic time allocation for backhaul
reduce power consumption over fixed strategies.
and access links and resource reuse for the downlink of
In terms of the cooperation strategies presented in Fig.
a single-cell OFDMA macro-cell aided by RSs aiming
2, it is discussed in [52] the EE of the schemes presented
to maximize capacity under fairness constraint. In the
in Figs. 2.b and 2.c, named two-hop half-duplex and
proposed method, MSs from different RSs are able to
multicast cooperative scheme, respectively. Considering
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
13
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
the downlink of an OFDMA system and optimized
resource sharing between the access/backhaul links, the
numerical results show that the two-hop protocol results
in higher EE, while the multicast strategy maximizes the
ϴ
average data rate. Another result comes from [53], in
αR
which the strategies from Figs. 2.a and 2.b are compared
in terms of EE maximization, and it is demonstrated by
R
numerical results that the later one provides the higher EE.
Table IV summarizes representative works and results
analyzed in this section.
a)
4.2. RS Deployment Problem
The distance from the RSs to the BS, the number of RSs
and even the distribution of the RSs in the cell could
result in different energy efficiency gains. When the RSs
are close to the BS, more users are assisted by relays,
β
but the backhaul link could be saturated, and when the
α
RSs are next to cell edge, few users can exploit the
benefit of relaying. With few RSs, the MS-RS distance
can be higher, which results in higher power consumption
to compensate path-loss. Increasing this number results
in higher probability of finding a better RS, but again
the backhaul link may limit the EE gain. Finally, if the
b)
distribution of RSs is considered for practical scenarios,
taking into account, for instance, coverage holes and
zones with increased number of users, tends to obtain
increased EE than considering stochastic placement. The
main models for RS deployment problem are demonstrated
in Fig. 5.
The relay deployment problem can also consider the
economical cost of the installation, as described in [54].
In this paper, the authors considered the cost factor
R
varying the deployment density of relay stations and BS
stations, considering system capacity normalized by cell
area and deployment of type-I and type-II relays. To solve
the cost minimization problem, all the combinations of
type-I and type-II RS and BS densities that result in
the same normalized capacity are found, and then the
point of tangency to linear cost line, which represents
equal cost for BS-RS density, is taken as minimum
c)
Figure 5. Relay deployment strategies, where: a) RSs are
uniformly distributed over a circumference of radius αR
centered at BS; b) RSs are placed inside a ring area delimited
by αR and βR; and c) stochastic placement. It is considered
NRs = 6 in a) and b), while NRs = 8 in c).
deployment cost point. The RS deployment is compared
to micro-BS deployment, and is shown that more RSs are
necessary to obtain the same performance as the micro-
of [55] introduce user experience satisfaction metrics to
BS case, but the costs are lower for RS deployment,
the problem, indicating that not necessarily the best-
which implicates that RSs are the best choice. Extending
cost deployment results in best experience satisfaction
this work and only considering type-II relays, the authors
for the mobile users. In [56], the authors compare the
14
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
deployment of relay stations with femtocells deployment,
defined as the users that cannot achieve a target SINR.
and it is demonstrated that RSs can achieve lower power
Solved using non-linear programming, it is demonstrated
consumption than femtocell-based deployment, while
by numerical results that the optimized deployment of
reducing operational and capital expenditure (OPEX and
relays reduce the probability of low data rates over the
CAPEX, respectively) when compared to macro-BS only
uniform placement and non-cooperative scenario.
scenarios.
In [62], an iterative RS placement algorithm for
The simplest approach to a relay-based network is to
multicell WiMAX systems is presented, in which the RS
consider that the RS is between the BS and all mobile
that provides the higher capacity gain at each iteration is
users, so that all users are relayed, as can be seen in
deployed. Instead of choosing a placement radius, the RSs
[57]. For the cellular network case this is not a suitable
can be placed in candidate positions in an annular section
model, since we can have users closer to the BS, which can
of each cell, which begins at the limit of BS coverage
communicate direct with it. A more sophisticated approach
area, which can be seen as the area where the BS are
to the relay placement problem is to define that the RSs
able to provide sufficient quality for the MSs. Despite
are placed in a circumference of radius αR centered at
discussing the frequency reuse technique, there are no
the BS, where R is the cell radius and 0 < α < 1, where
information about the model considered. Numerical results
the number of RSs and the parameter α are modified
demonstrate throughput gain when deploying RSs in the
to optimize EE. This method is well established in the
multicell system.
literature, and has even been used for second generation
When considering coverage extension, the authors of
systems, as CDMA [58]. In [59], the relay placement
[63] develop a method to determine the non-transparent
optimization is investigated for the uplink of direct-
RS placement radius in order to achieve the higher
sequence (DS) CDMA systems, with non-regenerative
coverage radius. Defining coverage as the area where
outband RSs and outage probability constraint. It is
a SINR detection threshold is maintained, the coverage
demonstrated that even for the most inefficient detector the
is maximized in such a way that both the BS-RS and
RSs bring significant gains for EE and outage probability
RS-MS links are able to achieve the SINR threshold,
reduction, which combined to filter optimization at least
and the number of RSs is determined in order to obtain
double the EE normalized by bandwidth, given the extra
non-overlapping coverage between two RSs, without
carrier for outband relaying. For the OFDM/OFDMA case,
optimizing this variable, and they are placed in a
this BS-centered approach without placement optimization
circumference centered at the BS. This approach is tested
‡
can be seen in [48] and [60], while in [49] and [50] the
for both single- and multi-cell cases, where the first one
placement distance and/or number of relays are analyzed
is optimally solved and the second one is solved by
to obtain power saving and capacity gain, respectively.
an iterative algorithm. Numerical results demonstrate the
The benefits of a well planed RSs deployment can
coverage gains obtained with RSs deployment.
be confirmed analyzing the results from [61], at least in
Instead of deploying RSs for each cell, in [64] the
terms of system capacity. The authors consider a multicell
authors discuss a two-cell case and propose to place
scenario where only in the central cell the relays can be
only one RS between the two cells. The whole idea in
placed in candidate locations in a ring area centered at
the paper is to reduce the inter-carrier interference (ICI)
the BS, while in the other cells the RSs are located using
and the inter-relay interference (IRI) resulting from the
uniform distribution over a circumference also centered at
standard cooperative approach by using the unique relay
the BS. For comparison purpose, simulation results include
for both cells. In the first time-slot, each BS transmits to
a second approach, where all cells use the BS-centered
the users, and the relay overhears this transmission from
uniform distribution. In this context, two optimization
both cells. In the second time-slot, the RS chooses one of
problems are proposed: maximize overall system capacity
the received signals in each subcarrier and then retransmits
or maximize the capacity for cell-edge users, which are
to all covered users. In this way, each relayed user receives
its own information or the information of the other cell
user. If it receives its own information, the user can use
‡
In this work, the relay placement is done in a ring area and not in an
circumference, but the method is almost the same.
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
combination techniques to obtain diversity gain; in the
15
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Table V. List of papers for deployment optimization
Year
2008
Paper
[54]
2011
[55]
2011
2011
[56]
[63]
2011
[64]
2011
[51]
2012
[61]
2012
[62]
2012
[65]
Contribution
Optimization of deployment cost of RSs and BSs compared to micro-BSs and BSs, given a target normalized capacity and
type-I or type-II RSs;
Optimization of deployment cost of type-I RSs and BSs in a similar manner of [54], but considering also the analysis of
user experience satisfaction and demonstrating a tradeoff between cost reduction and user satisfaction;
Comparison of CAPEX and OPEX for RS-based and femtocell-based heterogeneous networks;
Coverage maximization by optimizing RS placement radius, where the RSs are placed in a circumference centered at the
BS. The number of RSs is determined by the placement radius;
Placement of one RS between two adjacent BSs to reduce interference in the RSs, where the relayed signal can be used to
obtain diversity gain or to cancel interference;
Analysis of a realistic cellular metropolitan case of RSs versus micro-BS deployment, using a coverage map provided by
a cellular operator;
Placement of RSs in candidate positions inside a ring area for capacity maximization, considering as metric the capacity
of all users or only cell-edge users;
Iterative placement of RSs, deploying the relay that results in the highest gain at each iteration aiming to maximize system
capacity;
Stochastic geometry for the placement problem, where the cell area, BS, MS and RS positions and quantity are considered
as random variables;
other case, the relayed information can be used to cancel
highly non-regular, which results that SG approach is
the ICI, also resulting in SINR gain.
an adequate framework for cellular network evaluation.
Unlike the hypothetical models considered above, the
Another advantage of the solution in [65] is that the EE is
model described in [51] considers the coverage map of
determined by the expectation of the stochastic definition
one area in London, and discusses RS deployment in
of the problem, resulting in an analytical solution, at the
order to provide better coverage and higher capacity, and
cost of considering fixed power allocation.
also compares the obtained results with the deployment
Relevant works and results compiled in this section for
of micro-BSs. RSs and micro-BSs can be installed in any
relay deployment optimization problem are listed in Table
location of the studied region, and this decision is made
V.
employing a metric defined by the authors that considers
the outage probability, backhaul link quality, coverage
and system capacity, which can be weighted to provide
4.3. Relay Assignment
different objectives. As the micro-BSs have a dedicated
Given that the placement and the physical design of the
link with the BS, the backhaul link bottleneck does not
relay stations have been done, it is necessary to define
affect the result and no access/backhaul resource sharing
how they cooperate. As the system diversity order is
is necessary, which results in higher coverage and capacity
directly proportional to the number of relay stations, since
gain when compared to RS deployment. In this way, we
they ideally provide uncorrelated copies of the original
can see a possible trade-off between cost efficiency and
signal, one can conclude that using all the available relays,
coverage/capacity enhancement when comparing [54] and
as shown in Fig. 6.b, is the best choice. Besides this
[51].
possible gain, one clear bottleneck is the backhaul link
A different metric from all above is the stochastic
limitation, which cannot support the increased amount of
geometry (SG) approach. Used in several resource
data. Another clarification comes from [66], which states
allocation problems and performance analysis, it is
that if the best relay is selected, the diversity-multiplexing
employed in the RS placement problem in [65]. As
gains tradeoff is the same as if all the relays are considered,
the problem is considered with a stochastic perspective,
which significantly reduces the amount of data to be
the placement distance and geometry are also stochastic,
transmitted without losing performance. The best relay
allowing us to analyze the effects of RS and BS densities
is the one that provides the best link between MS-RS-
on the system’s EE. The stochastic approach does not
BS given the policies presented in [66], where the ith
seem to reflect comercial deployment situations, but
RS is chosen if a) it maximizes the worst channel gain
the authors remember that the existent deployments are
of the two paths (BS-RSi , RSi -MS) or b) it maximizes
the harmonic mean of the two paths. Since these policies
16
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
are not necessarily optimal from de point-of-view of
of the relay selection problem since no channel state
energy-efficiency maximization, other techniques must be
information (CSI) is needed to perform selection, but only
analyzed.
the user location. Despite its simplicity, this approach can
result in overloaded/underloaded RSs and the selection of
non-optimal RSs, since qualitative metrics have not been
S
considered. Another simple procedure in relay assignment
has been suggested in [62]; indeed, when a RS is deployed
in a given iteration, the users that lay in the RS coverage
RS
RS
RS
RS
RS
area are allocated to it, and then removed from the
allocation algorithm to avoid coverage by a second RS.
Most of the RS selection algorithms focus on capacity
enhancement, as can be seen in [60], [50], [62], [49]
D
and [61]. For example, in [61] users only use the RSs if
a)
the achieved spectral efficiency is lower than a specified
threshold, while in [60] the users that cannot achieve the
minimum SE defined as the QoS metric use the relay
S
stations; herein, the criterion of the RS selection is based
on channel gain quality. In [60], the BS split equally the
power among the subcarriers and define the subcarrier
allocation, and then the problem is to select the RSs and
RS
RS
RS
RS
RS
then optimize the power allocation at each RS. The relay
selection is simplified to choose the RS that guarantee
the highest channel gain for the allocated user/subcarrier
pair, and the power allocation can consider fairness. As
D
the capacity maximization does not necessarily imply
b)
higher energy efficiency, it is necessary to investigate other
approaches.
In order to reduce power consumption, in [67]
S
the authors formulate the relay assignment problem
considering the energy costs to transmit, receive and
operate network equipments with a reliability constraint
RS
RS
RS
RS
RS
(inverse of the outage probability), enabling transmissions
to occur in direct or cooperative mode independent
of user’s position. The network consists of multiple
pairs of sources and destinations and multiple DF-based
D
c)
Figure 6. Relay assignment strategies, considering the selection of a) one RS; b) all available RSs; and c) a subset of the
available RSs.
RSs, which only retransmit if required by destination.
It is considered that each source-(RS)-destination link
is orthogonal, which can be seen as the subcarriers
in OFDMA, and each relay serves only one link,
which is a limitation to apply this model to cellular
A common simplification to the relay selection problem
is to just allocate the MSs to the nearest RS, and to
consider that if the MS-BS distance is less than the RSBS distance, then the mobile user communicates directly
with the BS. Such approach has been described in several
papers [58],[50],[49], and is able to reduce the overhead
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
case. To simplify the cooperative/direct transmission
selection, it is introduced the concept of virtual RSs,
and then a iterative max-min reliability fairness problem
is modeled to determine the relay and power allocation,
which is transformed in a proportional fairness algorithm
and solved by Lagrange decomposition. The numerical
17
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
results demonstrate significant gains in terms of fairness
the common two-hop method. On the other hand, since
and maximum transmission reliability, with significant
more time-slots are required to implement the multi-hop
complexity reduction to the exhaustive search method.
communication, it is expected a spectral efficiency loss.
When the optimization metric is the energy efficiency,
In terms of EE, the multihop communication can result
there are specific relay selection. For example, in [68] the
in lower power consumption, given that the path-loss is
authors formulated a metric to determine the best relay
reduced, but the expected loss in SE can null the power
selection based on the distance of all candidate relays to
reduction; in this case, probably exists a tradeoff between
the place where an optimal relay must be placed to achieve
the number of hops and EE.
the higher EE, given in energy-per-bit consumption under
In [71], the previous research is extended to the DF
maximum tolerable symbol error rate (SER) restriction.
protocol, but as pointed out by [72] and corroborated by
Considering only the path-loss effect and user’s position,
numerical results, the MRC diversity does not provide the
the optimum placement of an hypothetical optimum relay
same gain for DF relaying as for AF relaying. Furthermore,
that minimizes the energy per bit ratio is calculated and the
depending on the distance between the terminals and path-
nearest deployed RS is then selected. The theoretical RS
loss coefficients, one technique can be better than the
can be placed in the point which results in maximum EE
other in terms of outage probability, where the outage
for the network, to transmitter or to the RS, depending on
event means that the target SE has not been reached.
the parameter of interest. Despite the optimization problem
This way, the authors propose a problem where the
have not been defined for OFDMA, the approach can be
nodes can use either two-hop communication or the RO
easily converted into single-cell OFDMA systems.
mode, considering that only two relays are available, by
A similar technique is considered in [69], in which the
selecting the method that maximizes the end-to-end mutual
authors determine energy-efficient relay regions, where the
information. In this mixed case, the gains in terms of
achieved EE is higher than the direct transmission EE.
outage probability are remarkable, even varying the path-
The novelty presented in the paper is that asymmetric
loss exponent and position of the network nodes.
traffic flowing between downlink and uplink is considered
Another possible approach to energy-efficient relay
to determine relay selection, since the relay allocation is
assignment is demonstrated in [73], where the cooperative
not changed while switching from downlink to uplink (and
transmission is chosen only if the achieved energy
vice-versa) in order to reduce the overhead. In this model,
efficiency is higher than the obtained with direct
it is considered DF relaying and SE constraint, and the
transmission, and in this case the relay with higher EE
metric deployed is the minimization of the energy-per-bit
is selected. Another important consideration made in this
consumption. All the MSs can use or not the RSs, with
paper is the rate constraint. As the EE optimum selection
relay/mode selection being defined by the most energy
does not consider the possible capacity loss that can
efficient one.
result from using the cooperative mode, the authors use
In [70], the authors propose the relay ordering (RO)
a multiplicative factor to limit the minimum acceptable
technique for multi-branch multi-hop networks, i.e., where
capacity: if no loss is permitted, the cooperative capacity
all the RSs can overhear the source and also the other
must be the same of the direct link, while if EE is
RSs. The RO technique consists in defining an order
the only concern, this multiplicative factor can be set to
to the multi-hop transmissions, in a way that making a
zero. Considering also the optimization of time-sharing
maximal ratio combining (MRC) of the earlier transmitted
for access/backhaul links and half-duplex DF case, it is
signals from source and other relays maximizes the SINR
demonstrated that the proposed technique outperforms
at destination in each hop, which is also combined
the one presented in [66] for EE, and that time-sharing
via MRC at the destination. As pointed out, testing
optimization enables even higher gains.
all the possible RSs orderings can result in elevated
Operational cost for relay assignment have been
computational complexity, which results in a suboptimal
considered in [74]. The relay selection is performed in two
algorithm that order the RSs according to the SINR in the
steps, considering distinct operational costs at MS, RS and
source-RS path. According to the numerical results, the
BS to transmit and receive data, including digital signal
RO strategy results in lower symbol error rate (SER) than
processing costs. In the first step, all relays that achieve
18
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Table VI. List of papers for relay assignment
Year
2006
Paper
[66]
2008
[75]
2009
[70]
2010
[69]
2012
[68]
2012
[71]
2012
[73]
2012
[74]
2013
[60]
2013
[67]
Contribution
It is presented the opportunistic relaying approach, which consists of selecting one of the available RSs under two proposed
policies;
Selection of the optimum number of DF-based RSs to beamform and transmit data to the destination, considering the cost
of acquiring/transmitting the CSI for source-to-RSs and RSs-to-destination;
The relay ordering (RO) technique is presented for multi-branch multi-hop AF-based networks, maximizing the SINR
received at each slot of multi-hop communications;
Unique relay selection for both downlink and uplink to minimize energy-per-bit consumption. Presents also the energy
efficient cooperation regions, where the cooperative mode outperforms direct communication;
The RS which is closest to an theoretical optimal RS for a given user is selected, where this optimal RS can be chosen to
optimize the EE for overall cell, to the RSs or to the MSs;
The RO technique presented in [70] is extended to the DF case, and a mixed two-hop/RO method is developed to maximize
the end-to-end mutual information;
EE-based opportunistic relaying, where only one RS can be selected and only if the EE is higher than the non-cooperative
EE, considering also maximum data rate loss with cooperative mode;
The RS with highest channel gain to the destination is selected, considering also transmission bandwidth and constellation
size optimization;
Only users that cannot achieve the target SE uses the RSs, and for the predefined subcarrier allocation the RS that provides
the highest channel to the MS is selected;
Transmission mode, power allocation and relay selection to minimize power consumption given a minimum reliability
constraint with fairness for multi-source/destination network;
a minimum SNR threshold to meet BER requirements
MSs connect to the nearest RS, it will be necessary to
are included in a subset of possible candidates, while in
obtain the CSI of all MS-RS subchannels for the allocated
the second step, the RS with higher channel gain to the
RS and the CSI of all RS-BS subchannels, which results
BS is selected for relaying. It is demonstrated that when
in an excessive overhead, mainly for the first case, since
the cost of circuit power consumption becomes higher
the RSs must estimate the CSI and then transmit to the
than the transmission power consumption, as in low-range
BS. To solve this problem, a simplified feedback model
communication, the extra energy consumption of a relay
is proposed, which depends only on the number of RSs
station overcomes the relaying gain, which can result in
and subchannels. This problem is also extended to the case
lower EE. A predecessor of this paper is [75], where a
where each user can allocate only a limited number of
beamforming of RSs at the second step is considered,
subchannels, in order to avoid underserved users.
instead of the selection of only one or all RSs, considering
For the first optimization problem described above, it
the cost to obtain and feedback the necessary feedback, but
is also proposed the spatial reuse of subchannels between
discarding the effects of circuit power as it considers only
the RSs. This is done by allowing that part of the RSs
long range transmissions.
operate in the backhaul link while the other part operates
Table VI summarized the main results considered in this
subsection.
in the access link using the same subchannels, which
allows capacity enhancement at the cost of the interference
generated. In order to reduce interference and also reduce
4.4. Other Variables in EE Optimization
the complexity of the problem, since all the possible RSs
partitions could be considered as candidates, the partition
Another formulations for EE optimization consider
is done balancing system load and maintaining contiguous
swapping part of the allocated subcarriers between the
sets of RSs. Other example of spatial reuse can be found in
access and backhaul links, in order to consider diversity
[77], where six RSs are placed in a circumference centered
since the BS-RS and RS-MS links are uncorrelated, which
at the BS and the subcarriers used in the access link by one
is known as subcarrier pairing and illustrated in Fig. 7.b.
RS can be used in the backhaul link for other RS, since a
This way, in [76] it is proposed a mechanism that allocates
sufficient spatial separation is provided; in the considered
pairs of subcarriers or subchannels, one for the access link
work, both RSs with common subcarriers deployed for
and other to the backhaul link, in order to provide better
access and backhaul links are placed 180 degrees away
use of the multiuser and frequency diversities, aiming to
from each other.
maximize end-to-end throughput with fairness. Since the
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
19
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Access link
MS
subcarriers
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Backhaul link
RS
subcarriers
RS
subcarriers
BS
subcarriers
RS to not transmit data from some users in a given timeslot, waiting for a better opportunity to send the data. The
1
1
1
1
symbol-wise or resource block-wise processing refers to
2
2
2
2
the granularity of the proposed solution, corresponding to
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
process the entire OFDM symbol or each RB individually;
the second provides more flexibility at the cost of extra
complexity. For the vehicular scenario discussed in [43],
it is demonstrated that a remarkable gain can be obtained
by combining some of these techniques.
a)
Another advance discussed in [43] is the ability of the
RSs in buffering the received data in order to overcome
deep fading in access or backhaul links. This way, the RSs
Access link
MS
subcarriers
Backhaul link
RS
subcarriers
RS
subcarriers
BS
subcarriers
are able to deal with bottleneck situations, by adjusting
transmission rates and even the time sharing between
those two links. In the context of energy efficiency
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
Other important metric that impacts directly on the
6
6
6
6
energy efficiency and in the development of efficient
maximization, this feature can be useful, as the energy
presumably deployed to transmit in deep fading when no
buffering is possible can be saved to be used in the next
slots. Obviously, QoS metrics and buffer violation must be
analyzed, to avoid starving users.
b)
Figure 7. Illustration of subcarrier pairing method in b),
compared to the standard approach of MS-RS-BS link described
in a).
resource allocation algorithms is the adopted traffic model.
While several papers focus only on the full buffer model
[30, 52, 34], which states that there always are information
to be sent for all users all the time and the buffer has infinite
capacity, in [78] it is considered also a more practical
model, i.e., the finite buffer case, where the user is active
In addiction to the spatial reuse, it is also considered in
[77] the possibility of choosing the transmission scheme
for users located outside the one-hop area: in this way, the
MSs that can use a RS can choose between one- or two-hop
transmission with or without spatial reuse. The metric used
to decide which mode to use is the number of necessary
subcarriers to provide QoS for a given user, in order to
maximize the number of users with guaranteed QoS.
In [43] the authors investigate several improvements
that can be made for AF and DF protocols, as relaying
in time/frequency domain, slot division between access
and backhaul links, relay buffering, subcarrier pairing
and symbol-wise or resource block-wise processing. The
resource sharing between backhaul and access links
is investigated in some papers, and tries to balance
system loading between these two links to avoid over or
undercapacity. The relay buffering scheme is proposed to
avoid deep fading and overloads, turning possible to the
20
in the system while the buffer has data to be sent. In this
model, it is possible to consider data arriving to the buffer
with a stochastic distribution during time or that exists
a finite amount of data in the buffer to be transmitted.
With the latter model, it is possible to develop algorithms
that consider the amount of buffered information and
waiting time, as for example in [79]. The proportional
fair scheduler operate over the traffic models, and it is
demonstrated that the conclusions for one traffic model not
necessarily applies to the other model.
As urban/suburban cellular systems are very dense, the
multicellular interference is a very important effect to
be considered in the resource allocation algorithms. As
can be seen in several works, as for example in [80],
when the interference level grows the nature of the EE
utility function results in more users transmitting with
the maximum allowed power, which results in increasing
interference from one cell to another. One way to reduce
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
this effect is to set a price to the available power,
power consumption can be substantially reduced even
discouraging maximum power allocation. This strategy,
when more RSs are deployed, a result that, in general,
early presented in the EE context in [35], has been
cannot be achieved in conventional approaches given the
adapted to the multicell OFDMA scenario in [81], but
circuit power from the extra RSs.
not considering RS deployment. The EE maximization
If a cross-layer optimization is considered, there are
problem consists in optimizing power and subcarrier
other variables that can be examined. For example, in
allocation for downlink OFDMA, which is described
[21] the authors analyze both type-I and type-II LTE-A
as difficult to jointly solve given the strong correlation
relay configurations and propose a couple of optimization
between these two variables. This way, it is proposed a
metrics to the specifications. For the type-II (transparent)
suboptimal model that iteratively allocates the necessary
relays, it is proposed an optimization in the hybrid
power and then allocates the subcarriers until convergence
automatic repeat request (HARQ) in order to obtain
or maximum time are reached.
higher spectral efficiency for both regenerative and non-
An fundamental metric to be considered in the
regenerative protocols, and indicates that a throughput
deployment of energy-efficient relay networks is the EE-
gain about 45% can be obtained with the optimized
SE tradeoff. This importance comes from the fact that
HARQ for relay-based networks. For the type-I RSs,
the spectral efficiency cannot be completely ignored to
it is analyzed: a) access/backhaul resource allocation
achieve higher EE and that SE still being one of the
with different prioritization; b) interference handling for
most important performance metrics for modern wireless
inband relaying; and c) resource reuse for multi-hop
communication systems. Recently, the EE-SE tradeoff
transmissions.
problem is investigated in [53] for AF protocol considering
the optimization of transmission power and amplifying
Summaries for the main papers and results analyzed in
this subsection are depicted in Table VII.
factor, with/without the direct link. Is is demonstrated that
not necessarily the RS deployment results in better EE-SE
tradeoff, and mode selection is then necessary to obtain
better results. In [26], the EE-SE tradeoff is investigated
under the uplink of a OFDMA system considering the
impact of interference between the links, and it is
5. SIMULATION AND EVALUATION
METHODS
demonstrated that the EE-SE tradeoff is reduced when the
interference increases. The authors of [27] consider the
Given all the techniques described above, it is possible
EE-SE tradeoff for the downlink OFDMA case with QoS
to see that diverse combinations can be made to improve
constraints, given in this case by minimum rate criterium.
the energy efficiency of cellular OFDMA networks. Since
It is demonstrated that the EE is quasiconcave regarding
most of the literature considers simulation methods to
the SE for a sufficient large number of subcarriers, which
analyze performance, another considerable challenge is to
means that not necessarily the target or the maximum SE
develop a realistic simulation scenarios. Such definition
results in the higher EE.
can consider several parameters, including:
Since all RSs inserts circuit power consumption, an
interesting idea is to control the activation time of the RSs,
making some of them sleep, mainly when the traffic load is
reduced. This approach is discussed in [82], where an onedimensional network is considered, when several relays are
placed in order to improve energy efficiency. The relay
stations implements a random sleep control, where each
RS have a probability to be in sleep mode, which reduces
the power consumed with circuit, at the possible cost of
coverage reduction. By combining optimized placement
and sleep probability, the authors demonstrated that the
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
• Cell topology: inter-site distance (ISD), number of
neighbor cells, frequency reuse factor, bandwidth,
carrier frequency;
• Equipment description: antenna heights and patterns, circuit power consumption, gains, coverage
area;
• Physical area: Urban/suburban/rural, channel effects like shadowing and fast fading, multipath
propagation, path-loss model, probability of LoS,
user density;
21
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Table VII. List of papers for other variables
Year
2007
Paper
[77]
2008
[76]
2011
[26]
2011
[27]
2011
[82]
2012
[43]
2012
[78]
2012
[21]
2013
[81]
2013
[53]
Contribution
Spatial reuse of subcarriers between sufficient spaced RSs and one- or two-hop mode selection for users that are after RS
placement radius;
Subchannel pairing with reduced feedback model, to avoid a possible bottleneck, and spatial reuse of subchannels to
improve system capacity;
Game-theoretic approach to EE maximization, where the scheduler can consider or not fairness. It is also analyzed the
EE-SE tradeoff behavior for different interference power levels;
States that EE is strictly quasiconcave in SE for a sufficient large number of subcarriers, demonstrating the EE-SE tradeoff,
which depends on the rate criterion;
Optimization of relay placement and active probability for an one-dimensional network, in order to reduce the power
consumption, including transmission and circuit power;
Analysis of six characteristics to be considered in the relaying protocol design, as regenerative or non-regenerative
protocols, RS buffering and time/frequency operation, and the possible combinations of these features.;
Analysis of the impact of the traffic model for a downlink non-cooperative OFDMA system with proportional fair
scheduling, under full-buffer or finite-buffer models;
This paper analyzes several techniques to improve cooperative communications, considering both transparent and nontransparent RSs and also mobile relays;
Pricing mechanism for downlink multicell OFDMA systems based on exponential pricing, different from the linear model
presented in [35], and power/subcarrier allocation;
The EE-SE tradeoff is considered as maximizing the EE given a minimum SE constraint, optimizing source transmission
power and the RS amplification factor. Comparison of cooperative strategies described in Figs. 2.a and 2.b;
• OFDMA system parameters: number of subcarri-
and also some considerations that makes possible to adjust
ers and data subcarriers, uplink/downlink opera-
the value of circuit power consumption if some feature is
tion, backhaul/access links parameters, allocation
added or removed.
period;
The numerical evaluation also presents some challenges. Despite a possible adoption of stochastic geometry
As the parameter definition is a critical decision and can
approach, which provides methods for analytical evalu-
result in completely biased and unpractical conclusions,
ation, the majority of the available techniques consider
there are some references and standards to consult in
simulation over a predefined number of trials in order
order to define suitable simulation model. For example,
to obtain the averaged results of the proposed optimiza-
most of the standardization groups, like 3GPP and IEEE,
tions, in general varying the distribution/density of mobile
have reference documents with the parameters considered
stations in order to emulate mobility while avoid bias
for their systems, LTE-A and WiMAX, respectively.
in the obtained result by considering best or worst case.
Another research groups, like WINNER and EARTH, also
Determining the sufficient number of trials is fundamental
have guidelines for parameters values choice. Table VIII
to the quality of the results, in a way that the obtained result
presents a list of these documents and a short description
can be reproduced. If the number of trials is overestimated,
of consolidated system parameters found inside them.
the results will not be affected, but extra computational
The effect of path-loss channel can be also simulated
resources are wasted without gain.
by deterministic models, as Lee, Hata, and so on, if the
To implement an energy efficient power allocation
physical description is relaxed [11]. Another important
policy, there are various approaches described in literature.
documents to be considered for parameters are the 4G
One of the most known is the game theoretic approach,
standards, as the published by 3GPP and IEEE 802.16
in which each player, i.e. mobile user, competes to grab
Workgroup [96]. In terms of circuit power consumption,
the necessary resources to achieve the target EE. In the
if the exact value is hard to obtain, it is possible to model
context of OFDMA, the users can compete for subcarriers
it with the specifications of the equipment. As described
and RS/BS transmission power. In the context of game
in [97], the circuit power consumption can be calculated
theory, users can form coalitions or selfishly compete for
by the cost of several functionalities, as analog-to-digital
and digital-to-analog converters and the signal processing
made by the BS or RSs. In the case of WiMAX relays,
[97] discusses case studies for both AF and DF protocols,
22
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Table VIII. Documents that describes system parameters for cellular (non-)cooperative OFDMA simulation
Document
ITU-R M.2135 [83]
3GPP TR 36.942 [85]
3GPP TR 36.814 [86]
3GPP TS 36.101 [87]
3GPP TS 36.104 [88]
3GPP TS 36.116 [89]
3GPP TS 36.211 [90]
3GPP TS 36.216 [91]
3GPP TS 36.826 [92]
IEEE 802.16j-06/013r3 [93]
WINNER II Channel Models
[94]
R1-104460 [95]
Description
As a guideline to evaluation of radio interface, provide several parameters for channel modeling and
discuss the evaluation model used for simulation and analytical results. This documents references other
technical notes, as for example the ITU-R M.2134 [84], which brings performance metrics, as cell-edge
spectral efficiency. Besides not considering relay equipments, this document discuss the possibility of
relay stations, providing some of the necessary changes in the considered model;
Despite the interference simulation described in this technical report is not in the scope of relay scenarios,
this document provides several parameters for BS and MS power profile, channel path-loss models and
cell arrangement;
This document defines the two types of RSs supported for LTE-A, including considerations in link
separation, which types of control messages are transmitted in each case and compatibility with earlier
LTE specifications. Brings also considerations about channel modeling and simulation model for RS
deployment, with network parameters and performance metrics;
Technical definitions for MS, including maximum power, test cases, operational bands;
Technical definitions for BS, including operational frequency bands, interfering limits and power and
performance parameters;
Technical definitions for RS, including operational frequency bands, interfering limits and power and
performance parameters;
Definition of resource block components, subcarrier bandwidth and other descriptions for
the transmission channel, as available modulations and symbol mapping, frame structure and
scrambling/precoding functionalities;
This specification provide changes for another documents of 3GPP for relay functionalities, as resource
partitioning, control channels and relay procedures. Despite being not completed, it is interesting to
follow its evolution;
This document provides system scenarios for LTE relay nodes, that are supposed to be fixed, and other
system parameters, as maximum output power, multipath channel taps and MIMO correlation matrix,
number of relay nodes, deployment layouts and BS/RS selection algorithm;
Despite the fact that this document is from 2007, the detailed description of how to compose the
simulation model, channel and traffic models (full buffer or application-based ones) make this document
an interesting suggestion to overview the necessary steps to construct the simulation case;
This document provides several channel models for cellular communications with carrier frequencies
between 2 and 6 GHz, including models designed for BS-RS and RS-MS links (as the B5 class) and BSMS for several propagation scenarios, as urban, suburban and rural. This document also provides LoS
and NLoS models, LoS probability and shadowing parameters;
In this document, simulation results are presented to demonstrate the throughput gains for type-I relays
in uplink LTE-A. Furthermore, this document present a detailed system parameter description and
references to other documents that consider the downlink case;
the resources§ . When forming coalition, the users can
OFDMA systems there are multicell interference, inter-
share information in order to decide the best strategy, i.e.
carrier interference and interference for relay access, the
the optimum resource allocation, but it involves a large
resource allocation can be properly modeled as games [26].
amount of data exchange between MSs, RSs and BS and
There are other possible implementations for the EE-based
also the exhaustive search in all the possibilities. If other
resource allocation, as can be seen in [27].
variables are included, as for example the placement radius
A fundamental aspect observed is that the scope
and number of RSs, the computational complexity can
of the problem must be well determined, in order to
be prohibitive. This way, the non-coalitional games are
provide accurate interpretation of the obtained results. For
preferred, at the cost of, in general, loss of optimality.
example, the results in [45] indicate that the selective DF
For example, in [98] it is demonstrated that when the
is the best approach in any configuration in the 3-node
multiple access interference (MAI) is present, the resultant
network, but in [99] it is proposed a SIMO model that
Nash equilibrium from the non-coalitional game can be
consider the impact of the circuit power in both protocols,
improved if all users are able to reduce their respective
including the extra components needed for DF operation,
transmission power. Since in practical cooperative cellular
and shows that the AF protocol result in higher EE than
DF when the distance between source and relay is lower,
§
In general these games are called cooperative and non-cooperative, respectively,
but in order to avoid confusion with cooperative networks, the terms coalitional
and non-coalitional are used to identify, respectively, the two approaches.
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
and that exists a distance when this behavior changes, i.e.
DF becomes more energy efficient than AF. Other aspect
23
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
in the problem scope is to determine which techniques
It is worth noting that some techniques cannot be jointly
will be considered for investigation and how they will be
deployed, like using the uniform RS placement discussed
combined, i.e. jointly optimized or optimization in steps,
in subsection 4.2 and most of the optimized assignment
executing one method and then applying others.
strategies considered in subsection 4.3, since in general
This way, the following issues are of paramount
importance to the cooperative EE optimization problem:
only a small portion of the deployed RSs are able to serve
a specific user. Another example is the analog AF with
any optimized subcarrier pairing, as discussed in [43]. This
• Provide a realistic scenario to determine the
way, a feasible combination of optimization features, as
achievable gains when considering numerical
discussed in [43], must be considered to avoid wasting
simulation approach. If a stochastic model is
computational resources with unpractical combinations.
considered, it is necessary to determine if it is a
plausible approximation to real case scenarios;
• Define the optimization variables, in order to limit
the investigation;
• Determine a set of optimization techniques that
will be analyzed in the developed problem and
Finally, this survey shows that the research in energy
efficient systems presents several possibilities and open
issues for modern and future wireless cellular systems,
which beyond its importance and necessity results in an
important and challenging research topic. In terms of
research challenges, we can consider:
the resultant complexity, in order to verify if
• As mentioned in [43], there are several combina-
an optimal solution is feasible or sub-optimal
tions of optimization techniques that can be made,
algorithms should be deployed;
at the cost of increased computational complexity.
Establishing which combinations can be made and
the impact of these combinations to non-optimized
systems, mainly considering analytical solutions,
6. CONCLUSION AND RESEARCH
CHALLENGES
still a challenging task;
• Besides the potential to improve energy-efficiency,
several of the analyzed techniques are considered
We have discussed several techniques for energy efficiency
only for capacity enhancement or power consump-
(EE) maximization in 4G cellular communications via
tion minimization, and not necessarily can be easily
relay stations deployment, providing useful guidelines,
or usefully adapted to the energy efficiency frame-
from system definition to techniques of optimization,
work;
also pointing out and listing documents with descriptions
• The lack of standardization and the growing
for physical channels, equipments and also simulation
importance of EE provides an interesting hole to the
techniques. Furthermore, some techniques are presented
research in cooperative networks for 4G systems,
in terms of throughput maximization and/or coverage
since significant improvements can be proposed;
extension, and can be adapted or just used to inspire new
• Besides the variety of relaying protocols, in general
techniques development in order to maximize EE.
only DF and AF, with some optimizations, are
It is worthwhile to note that several of these
considered for the 4G cases. Despite being the two
optimization techniques can be combined in order to
most known protocols, it is necessary to investigate
provide higher EE, for instance combining two or more of
other possibilities to the energy-efficient design, as
those procedures discussed in section 4. Obviously, there
for example the protocols mentioned in [21];
are also the problem of computational complexity when
• As mentioned in the first item, all of the pro-
combining those techniques. Most of the research works
posed optimization techniques adds computational
describes the optimization problems as NP-hard, and sub-
complexity to the cooperative system, but these
optimal strategies have been developed to maintain the
optimization techniques are able to provide notable
complexity in suitable levels, and this problem is also
EE improvement in the wireless communication
increased given the fact that mobile devices have limited
systems. However, it is necessary to address this
processing capability.
complexity in terms of power consumption for the
24
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
network equipments, in order to accurately demonstrate the effectiveness of these techniques;
• Advanced relaying techniques as the multimode
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
5. Sun J, Wu D, Ci S. Battery capacity footprinting
and optimization analysis for wireless multimedia
communication. IEEE Global Telecommunications
relaying for higher reability networks, such as
Conference (GLOBECOM 2011), IEEE: Houston,
presented in [100], imposes new difficulties for
USA, 2011; 1–5.
resource optimization design. In this technique, a
6. Han C, Harrold T, Armour S, Krikidis I, Videv S,
BS and/or MS can operate as RS for other MSs,
Grant P, Haas H, Thompson J, Ku I, Wang CX, et al..
and in the case of the MSs they can be able to even
Green radio: radio techniques to enable energy-
operate as BS for other MSs. Coordinate all these
efficient wireless networks. IEEE Communications
possibilities in order to maximize EE, or even other
Magazine Jun 2011; 49(6):46–54.
optimization metric, can be a difficult task since
7. Andrews JG, Ghosh A, Muhamed R. Fundamentals
several combinations are possible. Other topology
of WiMAX: Understanding Broadband Wireless
for LTE-A Release 12 is the mobile relay, which
Networking. Prentice Hall: USA, 2007.
is discussed in TR 36.836 [101] to be installed, for
8. Moghaddari M, Hossain E. Cooperative communi-
example, in a high-speed train aiming to provide
cations in OFDM and MIMO cellular relay net-
coverage gain to the transported users;
works: issues and approaches. Cooperative Cellu-
• Finally, it is necessary to consider a realistic
deployment scenario, in order to not jeopardize
lar Wireless Networks. Cambridge University Press:
Cambridge, UK, 2011; 13–45.
the analysis and conclusions. Two examples of it
9. Rost P, Fettweis G. Green communications in cel-
are the impact of traffic models, as described in
lular networks with fixed relay nodes. Cooperative
[78], in which the traffic model can drastically
Cellular Wireless Networks. Cambridge University
change the optimum fairness factor, and the three-
Press: Cambridge, UK, 2011; 300–323.
node models used in some aforementioned works;
10. Hanzo L, Mnster M, Choi B, Keller T. OFDM and
besides the applicability for single-cell OFDMA
MC-CDMA for Broadband Multi-User Communica-
cases, the cellular systems are implicitly multicell,
mainly in urban scenario;
seq-lab
tions, WLANs and Broadcasting. IEEE Press, 2006.
11. Haykin SO, Moher M. Modern Wireless Communications. Prentice Hall, 2004.
12. Nosratinia A, Hedayat A. Network architectures
and research issues in cooperative cellular wireless
REFERENCES
1. Chen Y, Zhang S, Xu S, Li G. Fundamental
trade-offs on green wireless networks. IEEE
Communications Magazine Jun 2011; 49(6):30–37.
2. ETSI. 3GPP TR 36.913 version 11.0.0 release 11 requirements for further advancements for evolved
universal terrestrial radio access (E-UTRA) (LTEAdvanced). Technical Report, 3GPP 2012.
3. Haider F, Wang CX, Haas H, Hepsaydir E, Ge
X. Energy-efficient subcarrier-and-bit allocation in
multi-user OFDMA systems. IEEE 75th Vehicular
Technology Conference (VTC Spring), 2012; 1–5.
4. Xiong C, Li G, Zhang S, Chen Y, Xu S.
Energy-efficient resource allocation in OFDMA net-
networks. Cooperative Cellular Wireless Networks.
Cambridge University Press: Cambridge, UK, 2011;
3–13.
13. Shen G, Zhang K, Wang D, Liu J, Leng X, Wang W,
Jin S. Multi-hop relay operation modes. Technical
Report IEEE C802.16m-08/1429, Alcatel Shanghai
Bell, IEEE 2008.
14. Liebl G, de Moraes TM, Weitkemper P. Advanced
relay technical proposals. Technical Report 247223,
ARTIST4G - Advanced Radio InTerface TechnologIes for 4G SysTems, ARTIST4G Feb 2011.
15. Yang Y, Hu H, Xu J, Mao G. Relay technologies for
WiMAX and LTE-advanced mobile systems. IEEE
Communications Magazine Oct 2009; 47(10):100–
105.
works. IEEE Global Telecommunications Conference (GLOBECOM), 2011; 1–5.
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
25
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
16. Kim DI, Choi W, Seo H, Kim BH. Partial infor-
27. Xiong C, Li G, Zhang S, Chen Y, Xu S.
mation relaying and relaying in 3GPP LTE. Co-
Energy- and spectral-efficiency tradeoff in downlink
operative Cellular Wireless Networks. Cambridge
OFDMA networks. IEEE International Conference
University Press: Cambridge, UK, 2011; 462–495.
on Communications (ICC), IEEE: Kyoto, Japan,
17. Laneman JN. Cooperative diversity in wireless
2011; 1–5.
networks: algorithms and architectures. PhD Thesis,
28. Souza ÁRC, Abro T, Sampaio LH, Jeszensky PJE,
Massachusetts Institute of Technology, Cambridge,
Prez-Romero J, Casadevall F. Energy and spectral
MA 2002.
efficiencies trade-off with filter optimisation in mul-
18. Hoymann C, Chen W, Montojo J, Golitschek
tiple access interference-aware networks. Transac-
A, Koutsimanis C, Shen X. Relaying operation
tions on Emerging Telecommunications Technolo-
in 3GPP LTE: challenges and solutions. IEEE
Communications Magazine 2012; 50(2):156–162.
gies 2013; :??–??
29. Meshkati F, Goldsmith A, Poor H, Schwartz
19. Loa K, Wu CC, Sheu ST, Yuan Y, Chion M, Huo D,
S. A game-theoretic approach to energy-efficient
Xu L. IMT-advanced relay standards [WiMAX/LTE
modulation in CDMA networks with delay QoS
update]. IEEE Communications Magazine Aug
constraints. IEEE Journal on Selected Areas in
2010; 48(8):40–48.
Communications 2007; 25(6):1069–1078.
20. IEEE standard for local and metropolitan area
30. Reggiani L, Dossi L, Giordano L. Improving en-
networks part 16: Air interface for broadband
ergy efficiency in multi-cell OFDMA networks.
wireless access systems amendment 3: Advanced air
8th International Conference on Wireless Com-
interface. IEEE Std 802.16m-2011 (Amendment to
munications, Networking and Mobile Computing
IEEE Std 802.16-2009) 6 2011; :1–1112.
(WiCOM), 2012; 1–4.
21. Papadogiannis A, Farber M, Saadani A, Nisar M,
31. Khakurel S, Musavian L, Le-Ngoc T. Energy-
Weitkemper P, Sui Y, Svensson T, Ktenas D, Cassiau
efficient resource and power allocation for uplink
N, Moraes T. Advanced relaying concepts for future
multi-user OFDM systems. IEEE 23rd International
wireless networks. Future Network Mobile Summit
Symposium on Personal Indoor and Mobile Radio
(FutureNetw), 2012; 1–10.
Communications (PIMRC), 2012; 357–361.
22. Betz S, Poor H. Energy efficient communications
32. Yaacoub E, Dawy Z. A survey on uplink
in CDMA networks: A game theoretic analysis
resource allocation in OFDMA wireless networks.
considering operating costs. IEEE Transactions on
IEEE Communications Surveys Tutorials 2012;
Signal Processing 2008; 56(10):5181–5190.
14(2):322–337.
23. Meshkati F, Poor H, Schwartz S, Mandayam N.
33. Chu FS, Chen KC, Fettweis G. Green resource
An energy-efficient approach to power control
allocation to minimize receiving energy in OFDMA
and receiver design in wireless data networks.
cellular systems. IEEE Communications Letters
IEEE Transactions on Communications Nov 2005;
53(11):1885–1894.
24. Buzzi S, Poor H. Joint receiver and transmitter
Energy efficient cross-layer resource allocation
optimization for energy-efficient CDMA communi-
scheme based on potential games in LTE-A. 15th
cations. IEEE Journal on Selected Areas in Commu-
International Symposium on Wireless Personal
nications Apr 2008; 26(3):459–472.
Multimedia Communications (WPMC), 2012; 623–
25. Shannon CE. A mathematical theory of communication. Bell System Technical Journal 1948; 27:379–
423 e 623–656.
26. Miao G, Himayat N, Li G, Talwar S. Distributed
interference-aware energy-efficient power optimization. IEEE Transactions on Wireless Communications Apr 2011; 10(4):1323–1333.
26
2012; 16(3):372–374.
34. Ling D, Lu Z, Zheng W, Wen X, Ju Y.
627.
35. Saraydar C, Mandayam N, Goodman D. Efficient
power control via pricing in wireless data networks.
IEEE Transactions on Communications Feb 2002;
50(2):291–303.
36. Auer G, Giannini V, Godor I, Skillermark P, Olsson M, Imran M, Sabella D, Gonzalez M, Desset
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
C, Blume O. Cellular energy efficiency evaluation
19th Symposium on Communications and Vehicular
framework. IEEE 73rd Vehicular Technology Con-
Technology in the Benelux (SCVT), 2012; 1–4.
ference (VTC Spring), 2011; 1–6.
47. Gora J, Redana S. In-band and out-band relaying
37. Shi Q, Xu W, Li D, Wang Y, Gu X, Li W. On the
configurations for dual-carrier LTE-advanced sys-
energy efficiency optimality of OFDMA for SISO-
tem. IEEE 22nd International Symposium on Per-
OFDM downlink system. IEEE Communications
sonal Indoor and Mobile Radio Communications
Letters 2013; 17(3):541–544.
(PIMRC), 2011; 1820–1824.
38. Chowdhery A, Yu W, Cioffi J. Cooperative wireless
48. Joung J, Sun S. Power efficient resource allocation
multicell OFDMA network with backhaul capacity
for downlink OFDMA relay cellular networks.
constraints. IEEE International Conference on
IEEE Transactions on Signal Processing 2012;
Communications (ICC), 2011; 1–6.
60(5):2447–2459.
39. Ng D, Lo E, Schober R. Energy-efficient resource
49. Krishnan N, Yates R, Mandayam NB, Panchal J.
allocation in multi-cell OFDMA systems with
Bandwidth sharing for relaying in cellular systems.
limited backhaul capacity. IEEE Transactions on
IEEE Transactions on Wireless Communications
Wireless Communications 2012; 11(10):3618–3631.
2012; 11(1):117–129.
40. Cardoso F, Correia L. MIMO gain and energy
50. Sundaresan K, Rangarajan S. Adaptive resource
efficiency in LTE. IEEE Wireless Communications
scheduling in wireless OFDMA relay networks.
and Networking Conference (WCNC), 2012; 2593–
2597.
Proceedings IEEE INFOCOM, 2012; 1080–1088.
51. Coletti C, Mogensen P, Irmer R. Deployment of LTE
41. Cardoso F, Torrea-Duran R, Desset C, Correia L.
in-band relay and micro base stations in a realistic
MIMO strategies for energy efficient transmission
metropolitan scenario. IEEE Vehicular Technology
in LTE pico-cell environments. 20th International
Conference on Software, Telecommunications and
Computer Networks (SoftCOM), 2012; 1–5.
42. Zappone A, Alfano G, Buzzi S, Meo M. Energyefficient non-cooperative resource allocation in
Conference (VTC Fall), 2011; 1–5.
52. Fantini R, Sabella D, Caretti M. Energy efficiency in
LTE-advanced networks with relay nodes. Vehicular
Technology Conference (VTC Spring), 2011 IEEE
73rd, 2011; 1–5.
multi-cell OFDMA systems with multiple base
53. Huang S, Chen H, Cai J, Zhao F. Energy efficiency
station antennas. IEEE Online Conference on Green
and spectral-efficiency tradeoff in amplify-and-
Communications (GreenCom), IEEE, 2011; 82–87.
forward relay networks. IEEE Transactions on
43. Riihonen T, Wichman R, Werner S. Evaluation
Vehicular Technology 2013; 62(9):4366–4378.
of OFDM(A) relaying protocols: Capacity analysis
54. Werner M, Naden M, Jesus P, Silva C, Moberg P,
in infrastructure framework. IEEE Transactions on
Skillermark P, Warzanskyj W. Cost assessment and
Vehicular Technology 2012; 61(1):360–374.
optimization methods for multi-node radio access
44. Kakitani M, Demo Souza R, Imran M. Energy
efficiency contours for amplify-and-forward and
networks. IEEE Vehicular Technology Conference
(VTC Spring), 2008; 2601–2605.
decode-and-forward cooperative protocols. 8th In-
55. Wang Y, Feng G, Zhang Y. Cost-efficient deploy-
ternational Symposium on Communication Systems,
ment of relays for LTE-advanced cellular networks.
Networks Digital Signal Processing (CSNDSP),
IEEE International Conference on Communications
2012; 1–5.
(ICC), 2011; 1–5.
45. Chen Q, Gursoy M. Energy efficiency analysis in
56. Khirallah C, Thompson J, Rashvand H. Energy and
amplify-and-forward and decode-and-forward co-
cost impacts of relay and femtocell deployments
operative networks. IEEE Wireless Communications
in long-term-evolution advanced. IET Communica-
and Networking Conference (WCNC), 2010; 1–6.
tions 2011; 5(18):2617–2628.
46. Chin WH. Energy efficiency of OFDM systems
57. Zappone A, Buzzi S, Jorswieck E. Energy-efficient
using shorten-and-forward in outdoor-to-indoor
power control and receiver design in relay-assisted
relaying scenarios with short guard intervals. IEEE
DS/CDMA wireless networks via game theory.
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
27
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
IEEE Communications Letters Jul 2011; 15(7):701–
703.
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
68. Tang Z, Wang H, Hu Q. An energy-efficient
relay selection strategy based on optimal relay
58. Badruddin N, Negi R. Capacity improvement in a
location for AF cooperative transmission. IEEE
CDMA system using relaying. IEEE Wireless Com-
International Symposium on a World of Wireless,
munications and Networking Conference (WCNC),
Mobile and Multimedia Networks (WoWMoM),
vol. 1, IEEE: Atlanta, USA, 2004; 243–248.
2012; 1–4.
59. Souza ÁRC, Abro T. Interference-limited fixed
69. Yang W, Li L, Sun W. Energy-efficient relay
relaying-aided macrocellular CDMA networks.
selection and optimal relay location in cooperative
Green Networking and Communications: ICT for
cellular networks with asymmetric traffic. CoRR
Sustentability. CRC Press: Boca Raton, USA, 2013;
165–210.
60. Farazmand Y, Alfa A. Power allocation framework for OFDMA-based relay-enhanced cellular
networks. IEEE Consumer Communications and
Networking Conference (CCNC), 2013; 534–539.
61. Elgendy O, Ismail M, Elsayed K. On the relay placement problem in a multi-cell LTE-advanced system
2010; abs/1009.0078.
70. Yi Z, Kim IM. Relay ordering in a multi-hop
cooperative diversity network. IEEE Transactions
on Communications 2009; 57(9):2590–2596.
71. Ju M, Kim IM, Kim DI. Joint relay selection
and relay ordering for DF-based cooperative relay
networks. IEEE Transactions on Communications
2012; 60(4):908–915.
with co-channel interference. IEEE 8th Interna-
72. Chen Q, Gursoy M. Energy efficiency analysis in
tional Conference on Wireless and Mobile Com-
amplify-and-forward and decode-and-forward co-
puting, Networking and Communications (WiMob),
operative networks. IEEE Wireless Communications
2012; 300–307.
and Networking Conference (WCNC), 2010; 1–6.
62. Nivedita M, Raja G. Efficient relay station
73. Amin O, Lampe L. Opportunistic energy efficient
placement strategy for broadband wireless networks
cooperative communication. IEEE Wireless Com-
- 4G. International Conference on Recent Trends In
munications Letters 2012; 1(5):412–415.
Information Technology (ICRTIT), 2012; 282–286.
74. Lim G, Cimini L. Energy-efficient best-select
63. Joshi G, Karandikar A. Optimal relay placement for
relaying in wireless cooperative networks. 46th
cellular coverage extension. National Conference on
Annual Conference on Information Sciences and
Communications (NCC), 2011; 1–5.
Systems (CISS), 2012; 1–6.
64. Son H, Lee S, Lee H. An efficient FRS-cooperative
75. Madan R, Mehta N, Molisch A, Zhang J. Energy-
strategy with no CSIT in OFDM-based networks.
efficient cooperative relaying over fading channels
IEEE 73rd Vehicular Technology Conference (VTC
with simple relay selection. IEEE Transactions on
Spring), 2011; 1–5.
Wireless Communications 2008; 7(8):3013–3025.
65. Deng N, Zhang S, Zhou W, Zhu J. A stochastic
76. Sundaresan K, Wang X, Madihian M. Low-
geometry approach to energy efficiency in relay-
overhead scheduling algorithms for OFDMA relay
assisted cellular networks. IEEE Global Commu-
networks. Proceedings of the 4th Annual Interna-
nications Conference (GLOBECOM), 2012; 3484–
tional Conference on Wireless Internet, WICON
3489.
’08, Institute for Computer Sciences, Social-
66. Bletsas A, Khisti A, Reed D, Lippman A. A simple
Informatics and Telecommunications Engineering
cooperative diversity method based on network
(ICST): ICST, Brussels, Belgium, 2008; 24:1–24:9.
path selection. IEEE Journal on Selected Areas in
77. Lee J, Park S, Wang H, Hong D. QoS-guaranteed
Communications 2006; 24(3):659–672.
67. Xie K, Cao J, Wang X, Wen J. Optimal
transmission scheme selection for OFDMA multihop cellular networks. IEEE International Confer-
resource allocation for reliable and energy efficient
ence on Communications (ICC), 2007; 4587–4591.
cooperative communications. IEEE Transactions on
78. Ameigeiras P, Wang Y, Navarro-Ortiz J, Mogensen
Wireless Communications 2013; 12(10):4994–5007.
P, Lopez-Soler J. Traffic models impact on OFDMA
scheduling design. EURASIP Journal on Wireless
28
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
Communications and Networking 2012; 2012(1):1–
13.
79. Song G, Li Y, Cimini L, Zheng H. Joint channel-
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
91. ETSI. 3GPP TS 36.216 version 11.0.0 Release 11
- physical layer for relaying operation. Technical
Report, 3GPP 2012.
aware and queue-aware data scheduling in multi-
92. 3GPP. 3GPP TR 36.826 V11.1.0 - relay radio
ple shared wireless channels. IEEE Wireless Com-
transmission and reception (release 11). Technical
munications and Networking Conference (WCNC),
vol. 3, 2004; 1939–1944 Vol.3.
Report, 3GPP 2012.
93. Senarath G, Tong W, Zhu P, Zhang H, Steer D,
80. Buzzi S, Poor H, Zappone A. Transmitter waveform
Yu D, Naden M, Kitchener D, Hart M, Vadgama
and widely linear receiver design: Noncooperative
S, et al.. IEEE 802.16j-06/013r3 - multi-hop relay
games for wireless multiple-access networks. IEEE
system evaluation methodology (channel model and
Transactions on Information Theory Oct 2010;
56(10):4874–4892.
performance metric). Technical Report, IEEE 2007.
94. Kysti P, Meinil J, Hentil L, Zhao X, Jms T,
81. Liu H, Zheng W, Zhang H, Zhang Z, Wen X.
Schneider C, Narandzić M, Milojević M, Hong A,
An iterative two-step algorithm for energy efficient
Ylitalo J, et al.. WINNER II channel models (D1.1.2
resource allocation in multi-cell OFDMA networks.
V1.2) - part I channel models. Technical Report,
IEEE Wireless Communications and Networking
IST-WINNER 2007.
Conference (WCNC), 2013; 608–613.
95. Nokia NSN. R1-104460 – type-1 relay performance
82. Zhou S, Goldsmith A, Niu Z. On optimal relay
for uplink. Agenda item 6.6.4 presented for
placement and sleep control to improve energy
discussion in the 3GPP TSG-RAN WG1 Meeting
efficiency in cellular networks. IEEE International
Conference on Communications (ICC), 2011; 1–6.
#62, Madrid, Spain.
96. IEEE standard for air interface for broadband
83. Union IT. ITU-R m.2135-1 - guidelines for
wireless access systems. IEEE Std 802.16-2012
evaluation of radio interface technologies for IMT-
(Revision of IEEE Std 802.16-2009) 2012; :1–
advanced. Technical Report, ITU 2009.
2542doi:10.1109/IEEESTD.2012.6272299.
84. Union IT. ITU-R m.2134 - requirements related
97. Dohler M, Li Y. Cooperative Communications:
to technical performance for IMT-advanced radio
hardware, channel & PHY. Wiley: United Kingdom,
interface(s). Technical Report, ITU 2008.
2010.
85. ETSI. 3GPP TR 36.942 version 11.0.0 release 11:
98. Betz S, Poor H. Energy efficiency in multi-
Radio frequency (RF) system scenarios. Technical
hop CDMA networks: A game theoretic analysis
Report, 3GPP 2012.
considering operating costs. IEEE International
86. 3GPP. 3GPP TR 36.814 v9.0.0: Further advance-
Conference on Acoustics, Speech and Signal
ments for E-UTRA physical layer aspects (release
Processing (ICASSP), IEEE: Las Vegas, USA, 2008;
9). Technical Report, 3GPP 2010.
2781–2784.
87. ETSI. 3GPP TS 36.101 version 11.6.0 release
99. Krishnan N, Natarajan B. Energy efficiency of
11 - user equipment (UE) radio transmission and
cooperative SIMO schemes -amplify forward and
reception. Technical Report, 3GPP 2013.
decode forward. Proceedings of 18th Internatonal
88. ETSI. 3GPP TS 36.104 version 11.6.0 release 11 base station (BS) radio transmission and reception.
Technical Report, 3GPP 2013.
89. ETSI. 3GPP TS 36.116 version 11.3.0 release 11
Conference on Computer Communications and
Networks (ICCCN), 2009; 1–5.
100. IEEE standard for wirelessMAN-advanced air
interface for broadband wireless access systems
- relay radio transmission and reception. Technical
–amendment
2:
Report, 3GPP 2013.
IEEE
Std
802.16.1a-2013
IEEE
Std
90. ETSI. 3GPP TS 36.211 version 11.4.0 release
11 - physical channels and modulation. Technical
Report, 3GPP 2013.
Higher
802.16.1-2012)
reliability
networks.
(Amendment
2013;
to
:1–319doi:
10.1109/IEEESTD.2013.6547982.
101. 3GPP. 3GPP TR 36.836 V2.0.2 - study on mobile
relay for evolved universal terrestrial radio access
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls
29
Energy Eff. Maxim. for Coop. and Non-coop. OFDMA
Á. R. C. Souza, J. R. A. Amazonas, T. Abrão
(E-UTRA) (release 12). Technical Report, 3GPP
2013.
30
c 0000 John Wiley & Sons, Ltd.
Trans. Emerging Tel. Tech. 0000; 00:1–30 ⃝
DOI: 10.1002/ett
Prepared using ettauth.cls