Hindawi Publishing Corporation

EURASIP Journal on Wireless Communications and Networking

Volume 2008, Article ID 219570, 14 pages
doi:10.1155/2008/219570

Research Article
Energy-Efficient Bandwidth Allocation for Multiuser Scalable
Video Streaming over WLAN

Xin Ji,1, 2, 3 Sofie Pollin,1, 2, 3, 4 Gauthier Lafruit,2 Iole Moccagatta,2

Antoine Dejonghe,2, 3 and Francky Catthoor1, 2
1 Katholieke Universiteit Leuven, 3000 Leuven, Belgium
2 IMEC, Kapeldreef 75, 3001 Leuven, Belgium
3 Interdisciplinary Institute for Broadband Technology (IBBT), Ghent University, 9000 Gent, Belgium
4 UC Berkeley, CA 94720, USA

Correspondence should be addressed to Xin Ji, [email protected]

Received 27 February 2007; Accepted 9 October 2007

Recommended by Peter Schelkens

We consider the problem of packet scheduling for the transmission of multiple video streams over a wireless local area network
(WLAN). A cross-layer optimization framework is proposed to minimize the wireless transceiver energy consumption while meet-
ing the user required visual quality constraints. The framework relies on the IEEE 802.11 standard and on the embedded bitstream
structure of the scalable video coding scheme. It integrates an application-level video quality metric as QoS constraint (instead of a
communication layer quality metric) with energy consumption optimization through link layer scaling and sleeping. Both energy
minimization and min-max energy optimization strategies are discussed. Simulation results demonstrate significant energy gains
compared to the state-of-the-art approaches.

Copyright © 2008 Xin Ji et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. INTRODUCTION traffic. For instance, considering scalable video coding tech-

niques [1–4], different retransmission limits were defined for
The demand for multimedia transmission over wireless net- different MAC priority queues in [5, 6]. These approaches
works exhibits an ever growing trend. As a result, the trans- rely on scalable video coding’s inherent prioritization in the
mission of multiple video streams over a single wireless lo- compressed domain to set MAC priorities. In [7], a solu-
cal area network (WLAN) is becoming a key requirement. In tion for scheduling transmission opportunities (referred to
this context, quality of service (QoS) provisioning for real- as TXOP in the remainder of the present paper) as a function
time applications among different users is becoming more of the data type was proposed.
and more critical, as wireless networks are affected by ex- As far as energy efficiency is concerned, a substantial
tremely error-prone and time-varying conditions. Besides body of prior work focuses on energy-efficient wireless trans-
this QoS challenge, low-power consumption is imperative to mission from the viewpoints of medium access control
enable the deployment of broadband wireless connectivity in (MAC) or physical (PHY) layers [8–10]. For energy-efficient
battery-operated portable devices. wireless media systems, Goel et al. solved an image transmis-
Dynamically, adapting video packet selection and sion energy optimization problem subject to distortion and
scheduling to achieve appropriate visual quality and energy rate constraints [11]. He et al. in [12] developed a power-
efficiency for such varying wireless networks is a challenging rate-distortion analysis framework to extend the traditional
task. For simplicity, most of the WLAN transmission stud- rate-distortion analysis by including power consumption as
ies consider throughput as the most important performance a third dimension. Although hardware-specific impacts were
metric, while it is not the most appropriate choice for video appropriately considered in [12], the analysis lacked a suffi-
traffic. Some recent studies try however to improve the trans- cient consideration of channel coding and transmission with
mission performance by exploring the specificities of video respect to the time-varying characteristics of the wireless
2 EURASIP Journal on Wireless Communications and Networking

channel. Focusing on an uplink mobile-to-base station sce-

nario, Lu et al. solved in [13] a power optimization prob-
lem subject to an end-to-end distortion constraint relying on Laptop
H.263 source coding and RS channel coding in conjunction
with the Gilbert loss model. In [14], Chandra and Dey pre-
sented a technique for enabling real-time video compression IEEE 802.11 WLAN
and transmission from wireless appliances based on run-
time video adaptation, and they estimated the energy con- Hand held computer
Server
sumption based on CPU load. Yousefi’zadeh et al. formulated
a set of optimization problems in [15] aimed at minimizing
total power consumption of wireless media systems subject
to a given level of QoS and an available bitrate relying on
multiple antennas. None of the aforementioned power opti- PDA
mization works considered the video scalability influence Figure 1: WLAN access point (AP) manages several mobile termi-
on the power consumption of wireless multimedia sys- nals (MTs) in a centralized network.
tems.
In addition, to the best of our knowledge, there is no
prior work considering joint optimization of real video qual-
ity and energy efficiency for wireless media systems. This
tance learning system, when a student is studying in real time
optimization requires to take the whole protocol stack into
using a wireless device and facing battery exhausting prob-
account. Furthermore, only few researches have provided
lems, he/she may be willing to scarify some visual quality to
an analysis of the complexity of the proposed optimiza-
finish the whole studying process.
tion problems. To cope with the time-varying QoS, exist-
ing methodologies often rely on fixed or nonscalable flow- The remainder of this paper is organized as follows.
based optimizations to allocate the available network re- Section 2 provides the background for understanding the
sources across the various multimedia users. Moreover, pre- contributions of this paper. Section 3 briefly reviews the
vious researches have seldom jointly exploited the adaptation IEEE 802.11 WLAN standards and the deployed 3D wavelet
or protection techniques available at the medium access con- motion-compensated temporal filtering (MCTF) scalable
trol (MAC) or physical (PHY) layers to enhance the perfor- video coding scheme. Section 4 formulates the consid-
mance of video applications. On the one hand, we can only ered problem statement for energy-efficient video scheduling
fully benefit from new technologies if we can analyze the with rate-distortion awareness. Total energy minimization
behavior of adaptation processes acting over communica- and fairness optimization are formulated separately. Next,
tion networks, taking into account the intrinsically stochastic lightweight algorithms are designed to solve the run-time
nature of communications and observations. On the other optimization problems for practical use. Appropriate system
hand, adaptation leads to nontrivial tradeoffs among many models are used to instantiate the proposed cross-layer opti-
parameters (i.e., delay, reliability, energy cost, etc.); thus re- mization framework given the aforementioned standards. In
thinking of the entire communication systems and quality of Section 5, we examine the performance of our framework
service must be provided. through simulations. Finally, concluding remarks are pro-
The main contribution of this paper is to exploit the ap- vided in Section 6.
plication layer peak signal-to-noise ratio (PSNR) scalability
enabled by the rate-distortion properties of scalable video
2. BACKGROUD AND PRELIMINARY WORK
bitstreams and to minimize the energy consumption among
different users. Instead of using conventional communica-
tion layer QoS metrics, such as throughput or packet loss Compared to the capacity improvements of wireless trans-
probability, a proper application-level video quality metric mission techniques, there are limited advances in battery ca-
is considered in the optimization. Compared to our former pacity. Since more powerful transmission schemes cost more
work on energy-efficient video transmission over WLAN, the energy, there is an increasing energy gap between the energy
resulting solution enables to further minimize the wireless requirements of new applications and radio technologies and
transceiver energy consumptions by a factor of 2 without the energy awareness in the battery. Thus, it is critical to re-
degrading the visual quality. The considered setup consists duce the power consumption or, equivalently, to enhance the
of multiple independent users equipped with mobile termi- energy efficiency of the mobile devices.
nals (MTs) downloading video streams from the access point The goal of improving the energy efficiency of wireless
(AP) of a WLAN (see Figure 1). The video data are encoded communication devices has already triggered a lot of re-
using a scalable video coding scheme and stored on a video searches at various levels, from circuit to communication
server accessed through the AP. Therefore, no real-time en- theories and networking protocols. The energy management
coding is performed. The users receive data over a shared problem, in its most general formulation, consists in dynam-
slowly fading wireless channel. It is assumed that different ically controlling the system to minimize the average energy
users can require different video qualities. This is a very im- consumption under a performance constraint. Existing re-
portant and realistic test case. For instance, considering a dis- searches can be classified into two categories.
Xin Ji et al. 3

1 2 3 ··· L system state and relying on the tradeoff characterized in the

design-time phase, the server/access point searches the trade-
Figure 2: Directed acyclic graph of embedded bitstream. off curves of the different users in order to minimize the total
energy cost subject to a fixed and bounded transmission time
delay. It then allocates the corresponding configuration to the
(i) Top-down approaches: approaches that are intrinsi- different user devices.
cally utilization- and hardware-aware but communi- In this paper, we introduce the rate-distortion property
cation-unaware are categorized as top-down. The of the video bitstreams into the proposed cross-layer frame-
communicating device is treated as any electronic cir- work and show that significant energy gains can be achieved
cuit, and general-purpose techniques like dynamic by exploiting this property. Besides all the scalability existing
power management and energy-aware design are ap- in the PHY and MAC layers, a significant amount of scala-
plied. The first technique is defined as dynamically bility is available in the video bitstream. A directed acyclic
reconfiguring an electronic system to provide the re- graph is often used to express the interdependencies between
quested performance levels with a minimum number the different data units. A typical dependence graph of an
of active components and minimum loads on those embedded coded bitstream is sequential, as shown in Fig-
components [16, 17]. The second technique can be ure 2 [26]. The arrow directions show that a data unit can
defined as designing systems that present a desirable be correctly decoded only when the dependent data units are
energy-performance behavior for energy management also correctly decoded. From the graph, we know that the
[8, 18, 19]. loss of different data units can result in varying decoded vi-
sual qualities. Many unequal error protection schemes have
(ii) Bottom-up approaches: approaches that are in-
been developed based on this observation. By introducing
trinsically communication-aware but hardware- and
this property into the proposed cross-layer framework, we
utilization-unaware are categorized as bottom-up.
show that significant energy gains can be achieved. The pro-
They rely on the fundamentals of information and
posed scheme is practical and can be integrated within exist-
communication theories to derive energy-aware trans-
ing wireless and multimedia standards.
mission techniques and communication protocols. We
find here, for instance, the transmission scaling tech-
niques which exploit the fundamental tradeoff that 3. WLAN VIDEO STREAMING SYSTEM OVERVIEW AND
exists between transmission rate/power and energy ENERGY-PERFORMANCE MODELING
[20, 21]. Network power management techniques also
The use of IEEE 802.11 WLANs is growing at a rapid
fall in this category, targeting the minimization of the
pace. With the substantial increase in the available bitrates,
transmission power under QoS constraints [22].
the transmission of real-time audio/video applications over
(iii) Top-down and bottom-up approaches can easily re-
WLANs becomes a reality. In this section, we first briefly in-
sult in a fundamental contradiction. A good example is
troduce the IEEE 802.11 standard and the scalable video cod-
the conflict between transmission scaling at the phys-
ing scheme that are considered in the present work. It is how-
ical (PHY) layer (bottom-up) and sleeping schemes at
ever important to emphasize that the cross-layer algorithms
the MAC layer [23] (top-down). Scaling tends to min-
proposed in this paper can be deployed with any video cod-
imize transmission energy consumption by transmit-
ing scheme where the bitstream can be organized into data
ting with the lowest power over the longest feasible
units with embedded structure (see Section 3.3). Based on
duration, whereas sleeping tends to minimize the duty
this description, we show how to calculate the energy con-
cycle of the radio circuitry by transmitting as fast as
sumption, the transmission delay, the error probability of the
possible. Clearly, the two techniques are contradictory
data, and the expected quality of the received decoded video.
when it comes to defining the optimal transmit rate
and power allocation.
3.1. PHY modes of 802.11a OFDM and channel model
In [24, 25], we showed that a cross-layer combination of
both approaches can significantly decrease the energy con- The IEEE 802.11a [27] PHY layer is based on orthogonal
sumption in a multiuser scenario. A framework was pro- frequency division multiplexing (OFDM), and it provides
posed for allocating the network resources energy efficiently. eight different modes with different modulation schemes and
The framework is subdivided into two steps and it focuses code rates resulting in data transmission rates ranging from
on the PHY and MAC layers for which the energy, packet er- 6 to 54 Mbps. The corresponding data rate and the associ-
ror rate (PER), and transmission time are considered. First, ated power level requirements are provided in Table 1, where
during the design-time phase, the performance-energy scala- NDBPS denotes the number of data bits per symbol.
bility resulting from the available controllable parameters of
the system is analyzed. Cost-resource-quality tradeoffs, tak- 3.1.1. PHY layer performance model
ing into account energy cost, PER quality, and transmission
time resource requirements of each user, are fully character- We consider a direct-conversion radio transceiver architec-
ized for each possible system state (i.e., a finite set of possi- ture [28]. Four control dimensions have significant impact
ble realizations of external variables tracking system dynam- on energy and performance for these OFDM transceivers:
ics). Second, during the run-time phase, knowing the current the modulation order (Nmod ), the code rate (Bc ), the power
4 EURASIP Journal on Wireless Communications and Networking

Table 1: Multiple PHY modes for IEEE 802.11a.

Mode Data rate (Mbps) Min (dBm) Modulation Code rate (R) NDBPS
1 6 −82 BPSK 1/2 24
2 9 −81 BPSK 3/4 36
3 12 −79 QPSK 1/2 48
4 18 −77 QPSK 3/4 72
5 24 −74 16-QAM 1/2 96
6 36 −70 16-QAM 3/4 144
7 48 −66 64-QAM 1/2 192
8 54 −65 64-QAM 3/4 216

amplifier transmit power (PTX ), and its linearity specified by needed to send an MSDU can be, respectively, expressed as
the backoff (b). For a given data rate, communication per- EMSDU (K) and TXOPMSDU (K) [24, 25]. The energy cost and
formance is determined by the bit error rate (BER) at the re- time of transmitting an application layer packet p are then,
ceiver. Adding nonlinearity distortion to the received signal respectively, defined as E p (K) and TXOP p (K), and these val-
power, the BER can be expressed as a function of the received ues depend on the number of fragmented data units that
signal-to-noise and distortion ratio (SINAD) which can be need to be transmitted or retransmitted for successful packet
expressed as transmission. The retransmission scheme details of 802.11
MAC can be found in [32]. As the total energy and time
PTX × A needed to transmit a packet p are the sum of the energy and
SINAD = , (1)
A × D(b) + kT × W × N f time needed to transmit its fragments, E p (K) and TXOP p (K)
can be, respectively, expressed as
where A denotes the channel attenuation, the constants k, T,
W, and N f are the Boltzman constant, working temperature, E p (K) = (m + y)EMSDU (K),
channel bandwidth, and noise figure of the receiver, respec- (3)
TXOP p (K) = (m + y)TXOPMSDU (K),
tively, and the relation between the power amplifier back-
off b and the distortion D(b) has been characterized empir- where m denotes the number of MSDU fragments for the
ically for the Microsemi LX 5506 [29] 802.11a PA. The con- considered packet p, and y denotes the allowed number of
sidered BER-SINAD relation follows the model provided in MSDUs that can be retransmitted for the given packet p.
[30]. The BER-SINAD curves for different channel states for Similarly, the loss probability of a single MSDU is de-
all the considered PHY modes have been shown in Figure 3. noted as PMSDU (K), and it is computed based on the PHY
performance model introduced before. Since the probabil-
ity that a given packet p is received correctly depends on the
3.1.2. PHY layer energy model
probabilities that each of its fragments is received correctly,
Our energy model assumes the implementation detailed in We compute the packet error rate PERm y (K) at application
[31]. The corresponding parameters are provided in Table 2. layer according to
The time needed to wake up the system is assumed to be 100 y

microseconds. Denoting PPA as the power consumption of PERm
y (K) = 1 − Pemj (K),
the power amplifier, PFE as the power consumption of the j =0
 (4)
front end (FE), PBB as the power consumption of the base- m
Pey (K) = Cim (PMSDU )i (1 − P MSDU )(m−i) P iy−i (K),
R
band, and EDSP as the digital signal processor energy con-
sumption for decoding a single bit of a turbo-coded packet,
m
Pe0 (K) = (1 − P MSDU )m .
we obtain the following expressions for the energy needed to We refer to [24, 25] for more details on the wireless chan-
send or receive a MAC service data unit (MSDU) of length nel model and the link layer scaling (adapting the modu-
LMSDU under bit rate Bbit : lation order and code rate to spread the transmission over
 T T T time) and sleeping (introducing as much as possible trans-
PPA + PFE + PBB
ETX = × LMSDU , mission idle period) optimization schemes.
Bbit
R R  (2)
PFE + PBB R 3.3. Distortion, energy, and delay of
ERX = + EDSP × LMSDU .
Bbit scalable video bitstream

3.2. Error probability, energy consumption, Embedded scalable video coding has been an active research
and transmission delay of the IEEE 802.11 MAC topic in recent years. It has the attractive capability of re-
constructing lower resolution or lower quality videos from
Considering a possible transmission configuration vector a single bitstream, hence providing simple and flexible so-
K (each specific control dimension listed in Table 2 cor- lutions for transmission over heterogeneous network condi-
responds to an entry in this vector), the energy and time tions and easier adaptation to a variety of storage devices and
Xin Ji et al. 5

BPSK QPSK
100 100

10−2 10−2

10−4 10−4

0 10 20 30 40 0 10 20 30 40
(a) (b)

QAM16 QAM64
100 100

10−2 10−2

10−4 10−4

0 10 20 30 40 0 10 20 30 40
(c) (d)

Figure 3: BER-SINAD relations.

Table 2: Parameters of the energy model.

Performance model Energy model MAC model Control dimensions

W = 20 MHz T
PFE = 200 mW LMSDU = 1500 B Backoff (dB) 6 to 6
B = 250 kBaud R
PFE = 200 mW TACK = 52 μs Pout (dBm) 0 to 20
Nc = 48 T
PBB = 50 mW TPLCP = 20 μs Modulation BPSK, QPSK, 16-QAM, 64-QAM
T = 198 K R
PBB = 50 mW TSIFS = 16 μs Code rates 1/2, 2/3, 3/4
N f = 10 dB R
EDSP = 8.7 nJ/b Block = 288 PSNR

terminals. Accordingly, many recent video codecs, such as the ficient manner [33]. After the removal of the temporal re-
Scalable Video Coding (SVC) extensions of H.264/AVC [3], dundancies, the produced low-pass and high-pass frames are
MPEG-4 FGS [4], and so forth, enable embedded scalable decomposed spatially by discrete wavelet transform (DWT).
coding. In a typical MCTF-based video compression, the rate allo-
cation of the scalable bitstream is possible for a maximum
3.3.1. Architecture of the considered scalable video encoder granularity of one group of pictures (GOP). Encoder and de-
coder thus process the video sequence on a GOP-by-GOP ba-
We consider a scalable video codec based on motion- sis, which creates naturally independent data units group.
compensated temporal filtering (MCTF) and a wavelet trans- An important feature of wavelet transforms is the inher-
form [2]. MCTF aims at removing the temporal redundan- ent support of scalability in the compressed domain. Cou-
cies of video sequences. Unlike predictive coding schemes, it pled with the embedded coding techniques, wavelet video
does not employ a closed-loop prediction scheme. Instead, it coding achieves continuous rate scalability. After applying
uses an open-loop pyramidal decomposition to remove both the wavelet transform, the resulting subband coefficients are
long-term and short-term temporal dependencies in an ef- coded using bitplane coding and a global rate-distortion
6 EURASIP Journal on Wireless Communications and Networking

optimization. As a result, the final bitstream is constructed where TXOP pi (Ki ) denotes the transmission time under con-
to satisfy the bitrate constraint and minimize the overall dis- figuration Kl .
tortion [2].
To achieve quality scalability, a multilayer bitstream is 4. ENERGY-EFFICIENT MULTIUSER CROSS-LAYER
formed where each layer represents a quality-level improve- OPTIMIZATION
ment. The fractional bitplane coding ensures that the bit-
stream is embedded with fine granularity. In this work, we 4.1. Problem formulation
distribute the rate of the layers inside a GOP in a way that
In this paper, we focus on techniques that efficiently
every enhancement quality layer contributes to a similar dis-
adapt the transmission strategy in order to minimize the
tortion decrease. The resulting embedded bitstream has a se-
transceiver energy cost while meeting the required end-to-
quential dependency; each layer can only be decoded under
end distortion and delay. Most of the existing solutions do
the condition that all the previous layers have been received.
not take into account the rate-distortion properties of video
Note that in our simulations, no error concealment is used.
bitstreams, and therefore they often lead to inferior network
In the next section, we will explain in detail how to estimate
efficiency and suboptimal qualities for the video users.
the distortion in the case of packet losses for these coding as-
As we operate in a very dynamic environment, the sys-
sumptions.
tem behavior will vary over time. Both the energy cost func-
tion and the resources required for transmission will depend
3.3.2. Distortion, energy, and delay calibrations of on this run-time behavior. In the considered wireless video
video bitstream streaming environment, the system state is determined by the
Commonly used quality measurements of reconstructed im- current channel state and the rate-distortion property of the
ages and videos are mean squared error (MSE) and peak video bitstream. Each GOP can then be associated with a set
signal-to-noise ratio (PSNR). Typical PSNR values should of possible system states S, which determines the mapping of
range from 30 to 40 dB. Taking only quality scalability into the transmission strategies K to the energy cost (K →EGOP,S )
account and assuming a stable channel during one GOP time and the required bandwidth resource (K →TXOPGOP,S ). Each
period, it is possible to calculate the expected distortion con- user experiences different channel and rate-distortion dy-
tribution of each quality layer on a GOP-per-GOP basis. We namics, resulting in different system states over time, which
focused on a GOP-based approach instead of the more fine may or may not be correlated with other users. It is this im-
granular ones to limit overhead and complexity. portant characteristic which makes it possible to exploit mul-
Let us assume that each GOP is encoded into L quality tiuser diversity for energy efficiency.
layers and that a quality layer is the smallest application layer From the former analysis, and under the assumption that
data unit. Let Dl denote the distortion corresponding to the all video users can require their own end-to-end quality, the
reception of layers 1 to l (1 < l < L), and let D0 denote the optimization problem is formulated with video quality as
distortion associated with losing the first layer. Denoting the one of the constraints. We consider two different objectives:
error probability of layer l under transmission configuration minimizing the total energy cost of all users, and the max-
Kl as PERKl , the probability of correctly receiving the qual- imum energy cost among all users (fairness rule). For both
 objectives, we provide a low-complexity run-time optimiza-
ity layers until layer l is lj =1 (1 − PERK j ). Relying on the
tion algorithm. The advantage of the proposed solutions will
sequential dependency of the embedded bitstream structure,
be analyzed and discussed in Section 5.
the expected average distortion De over one GOP can then be
calculated as
 i
−1 

4.1.1. Optimization towards total energy minimization

L

De = PERKl × D0 + 1 − PERK j The optimization consists in finding for each user u, u ∈
i=1 j =1

(5) (1, . . . , N), the configuration Ku∗ that minimizes the overall

L
 energy cost, subject to radio resource and video distortion
× PERKi+1 × Di + 1 − PERKi × DL . constraints. Such configuration is applied at the beginning
i=1
of every GOP transmission interval, considering the current
The energy EGOP of the whole GOP can be expressed as channel conditions and video rate-distortion properties:
the sum of its layers:  

N


L
 Ku∗ = min EGOPu Ku , (8)
EGOP = E p i Ki , (6) u=1
i=1
subject to
where E pl (Kl ) denotes the associated energy cost under con-
figuration Kl . Deu ≤ Dur , u ∈ (1, . . . , N),
Similarly, the transmission time TXOPGOP of the whole 
N (9)
GOP is TXOPGOPu ≤ Tr ,
u=1

L

TXOPGOP = TXOP pi Ki , (7) where Dur
and Tr denote the distortion and time constraints,
i=1 respectively.
Xin Ji et al. 7

4.1.2. Optimization towards fairness 0.035

In this approach, we consider how to allocate the bandwidth 0.03

and transmission strategies to achieve more fair energy cost
among all the users and formulate this problem as a min-max
problem. For N users inside the network, the optimization 0.025

Energy (J)
problem is formulated as a min-max problem to find for each
of the users u the configuration Ku∗ such that 0.02
 
Ku∗ = argmin maxEGOPu Ku , u = 1, . . . , N, (10)
0.015
subject to
Deu ≤ Dur , u ∈ (1, . . . , N), 0.01

N (11)
TXOPGOPu ≤ Tr , 0.005
u=1 0.014 0.016 0.018 0.02 0.022 0.024 0.026 0.028 0.03 0.032
TXOP (s)
where Dur and Tr denote the distortion and time constraint,
respectively. Figure 4: Energy versus TXOP Pareto curve example.

4.2. Two-phase solution approach

ing to their minimal cost and resource consumption. In this
Each of the above formulated problems is a multidimen- paper, the system states are denoted by different channel sta-
sional assignment problem, which is known to be non- tuses and the dynamic rate-distortion properties of video
polynomial (NP) time hard problem. To obtain a tractable traffic loads. To that end, we consider the Pareto concept for
run-time complexity, we proposed a two-phase solution ap- multi-objective optimization [34].
proach; at design time, for each possible system state, the op- Let us consider the following multi-objective program-
timal operating points (namely, Pareto sets) are determined ming problem:
according to their minimal energy cost and resource (TXOP)
consumption. At run time, a low-complexity algorithm is MIN f1 (X), f2 (X), . . . , fM (X) (12)
X ∈Ω
provided for the formulation of each of the problems rely-
ing on the design time calibration.  
To solve the optimization towards total energy minimiza- f i X1 ≤ f i X2 , ∀ i ∈ 1, 2, . . . , M,
  (13)
tion, we convert the problem into a Lagrangian relaxation f j X1 < f j X2 , ∀ j ∈ 1, 2, . . . , M,
problem. The main steps are as follows.
A solution X1 is strictly better than a solution X2 if X1
(i) At the design time, the optimal operational points are is at least as good as X2 with respect to all the M objectives
determined for each possible system state according to (the first condition of (13)), and X1 is strictly better than X2
their minimal energy cost, resource (TXOP) consump- with respect to at least one objective (the second condition
tion, and distortion. The operational points are gener- of (13)). A Pareto optimal solution is defined as if there is
ated to reduce the search space from the initial prob- no other solution strictly better than X1 . A multi-objective
lem. optimization problem may have multiple Pareto optimal so-
(ii) At the run time, the bisection algorithm is used to solve lutions, and different decision makers with different prefer-
the optimization problem. ences may select different Pareto optimal solutions. The set of
To solve the min-max problem, the main steps are as fol- all possible Pareto optimal solutions constitutes a Pareto fron-
lows. tier in the objective space. A two-dimensional Pareto fron-
tier is also called a Pareto curve. Figure 4 shows an example
(i) At the design time, the derivation of the optimal oper- of Pareto curve considering energy and network resources as
ational points is performed for the original Min-Max objectives.
problem after the system states of all users are known. At design time, for each possible system state, we com-
(ii) At run time, a lightweight water-filling scheme is pro- pute the three-dimensional Pareto frontier, considering the
posed to assign the optimal system configuration to optimization objectives of the distortion D, the network re-
each user. source TXOP, and the energy E. The Pareto frontier can be
In the Sections 4.2.1–4.2.4, the design-time and run-time ap- found by any global optimal algorithm since complexity is
proaches will be introduced, respectively. The two proposed not the concern at the design-time step.
algorithms will be detailed in the following. From the video side, the design-time calibration can be
provided for different GOP sizes. This enables adaptation to
4.2.1. Design-time phase channel violations by choosing a smaller GOP size for the
next coherence period. Depending on the channel state (how
The goal of the design-time phase is to determine, for each long it is stable), we may adapt the number of frames with-
possible system state, the optimal operating points accord- out influencing the later part of the video bitstream—thanks
8 EURASIP Journal on Wireless Communications and Networking

(1) Initialization:
min
(a) allocate to each of the u users the lowest cost possible for the given state EGOP .
u
(2) If Nu=1 TXOP0u > Tr ,
initialize λmax , λmin , and λtrying .
Do
restore previous λtrying :
λtrying = (λmax + λmin )/2.
For each user,
if λtrying is higher than the highest or lower than the lowest,
jump out of the loop
or else find λ > λtrying > λnext .
If the total delay is lower than the constraint
λmax = λtrying ,
restore the difference between the total delay and the constraint
or else λmin = λtrying .
While (the difference between total delay and constraint converges to the same point),
(3) for each of the users,
the Pareto energy TXOP set will be searched
till finding the configurations which have a λ just lower or equal to the resulting λ,
and these configurations are the optimal output settings.

Algorithm 1: Run-time bisection search algorithm to find the Lagrange multiplier and the optimal configuration.

to the open-loop temporal decomposition of the MCTF the slope (λ sets) of the convex hull. In contrast, we define
scheme. the λ sets to be the slope EGOPu /TXOPGOPu of each opera-
tional point, and we find the lowest λ∗ which satisfies the
4.2.2. Run-time phase constraint. From the definition of λ, we know that it repre-
sents the energy cost compared to the resource. And from the
Once the system states of all users have been known at run Pareto property, for each specific user, the λ(Ku ) is increas-
time, lightweight schemes are proposed to assign the best sys- ing with EGOPu (Ku ). The lower the λ is, the lower the EGOPi
tem configuration to each user. will be. Thus, if all the users choose configurations with a λ
The 3D Pareto frontier is firstly converted to a 2D Pareto lower than λ∗ , the constraint will not be satisfied. And if all
curve according to the QoS constraint. This step can also the users choose configurations with a λ larger than λ∗ , the
be incorporated in the design-time phase by providing sev- energy cost will be more than the resulting one.
eral QoS constraint levels (2D Pareto curve) for the run-time A bisection search is proposed to find the appropriate λ∗ .
choice. The Pareto frontier is first pruned by deleting those The first step of the initial solution is to include the lowest
settings that cannot satisfy the QoS constraint. The remain- cost point from each user (the highest resource requirement
ing cost-resource tradeoffs are further explored to extract a according to Pareto property). The amount of the resources

Pareto curve. used by this initial solution is TXOP0 = Nu=1 TXOP0u . In the
After the Pareto pruning, n Pareto cost-resource sets are next step, if TXOP0 is higher than the resource constraint
available for each user. The run-time algorithms for both Tr , we use the bisection search until finding the appropri-
problem formulations are discussed in the next sections. ate λ satisfying the resource constraint. Without loss of gen-
erality, we assume that each of these u users maintains a U
4.2.3. Proposed algorithm for minimizing total energy cost-resource Pareto setting. In this case, the complexity of
this step is O(NU log (NU)). From the Pareto property, λ is
The optimization problem expressed in (9)–(10) is reformu- strictly increasing with the energy. After finding the appro-
lated so as to introduce a Lagrangian multiplier λ [35]: priate λ for each user, the Pareto set will be searched. The

N 
N configurations which have a λ just lower or equal to the re-

minimize Jtot = EGOPu Ku + λ TXOPGOPu , (14) sulting λ are the optimal output settings. The complexity
u=1 u=1 of this step is O(NU). The pseudocode of the algorithm is
subject to shown in Algorithm 1.


N
4.2.4. Proposed algorithm for minimizing
TXOPGOPu ≤ Tr . (15)
u=1
the maximum energy

The conventional solution consists in constructing a con- A greedy water-filling algorithm is proposed to solve the
vex hull of the operational points first, and then searching run-time searching for this problem. The first step of the
Xin Ji et al. 9

(1) Initialization:
min
allocate to each of the N users the lowest cost possible for the given state EGOP u
.
Construct an N-value energy level vector,
with each of these values corresponding to the energy cost of one of the users.

(2) If Nu=1 TXOP0u > Tr ,
for the user who requires the lowest energy cost in this step,
sort out its energy TXOP tradeoff curve,
until a setting whose energy cost exceeds the second lowest energy cost level
is found or the resource constraint is satisfied.
(3) If the resource constraint is not satisfied,
update the energy level vector and repeat step 2 until the resource constraint is satisfied.

Algorithm 2: Run-time greedy water-filling algorithm.

initial solution is also to include the lowest cost point from Knapsack problem where the goal is to pack different dis-
each user (the highest resource requirement according to the crete items with different resource constraints and values to
Pareto property). Suppose that there are N users and the re- the user. If an infinite set of items would be present, with in-
source requirement of each of these N users composes U finitesimally small differences in terms of resource cost and
water-filling level vectors. The amount of the resources used value, the problem would be easy to solve. The discrete na-

by this initial solution is TXOP0 = Nu=1 TXOP0u . ture however makes it NP-hard. Many approximations how-
In the next step, if TXOP0 is higher than the resource ever exist that allow to find a close-to optimal solution that
constraint Tr , for the user which achieves the lowest energy works well enough in practice.
cost among others, we reallocate the setting until its energy The difference between the maximum energy cost
cost exceeds the second energy cost level or the resource con- achieved by the algorithm and the optimum one lies how-
straint is satisfied. If the resource constraint is not satisfied ever for sure between the maximum and minimum energy
by this step, we update the water-filling level vector and re- cost achieved by the last adaptation. In theory, this differ-
peat the last step until the resource constraint is satisfied. ence is hence bounded by the largest step size found in the
The resulting outputs are the optimal settings for all users. Pareto curves that are the possible optimal configurations for
The complexity of the water-filling algorithm is O(NU2 ). The the system. Practically, the convergence of the algorithm pro-
pseudocode of the algorithm is shown in Algorithm 2. vides a solution close to the optimal solutions with reason-
If the step sizes of the Pareto curve axes are infinitesimally able complexity. The reason is that in practice, the step sizes
small, the attentative reader might indeed observe that the between the different points on the curve are small enough.
Ku∗ we find is the optimum configuration to achieve the min-
max energy cost among users. 5. NUMERICAL RESULTS
∗ ∗ ∗
Proof. For configuration set Ku , for all u, Eu ≤ max Eu .
Based on the proposed two-phase approaches and the con-
If there exists a configuration set K u , which results for all
sidered transceiver system models, we now verify the energy
u in max Eu < max Eu∗ , then for all u, K u < max Eu∗ .
savings over a range of practical scenarios.
From the descending searching style of step 2, we have
for all u, Eu ≤ Eu∗ , and there exists at least one u such that
Eu < Eu∗ . 5.1. Simulation setup
From the definition of Pareto property, we have
TXOPu ≥ TXOP∗u , and there  In the experiments, a GOP size of 16 is assumed. Four se-
one u such that
exists at least quences (bus, city, foreman, mother and daughter) are con-
TXOPu > TXOP∗u . Hence, TXOPu > TXOP∗u .
From the water-filling searching of step 2, we know that sidered here as examples of video with different levels of mo-
tion activities, thus resulting in different bitrate versus dis-
for all the resulting TXOP higher than the TXOP∗u , the
constraint cannot be satisfied. Thus, there is no configura- tortion. All the sequences have CIF (352 × 288, 4 : 2 : 0)
tion set that can satisfy the constraint while achieving a max resolution and 30 frames per second. The number of qual-
energy cost lower than that of Ku∗ . ity layers is set to 5. Empirically, for an image/video of CIF
size, PSNR value of 25–35 dB corresponds to an acceptable
Due to the discrete step size of the possible configura- visual quality for most of the users. We therefore encoded ev-
tions that form the Pareto curve, there might exist other con- ery sequence with a visual quality of approximately 35 dB for
figurations that achieve less max energy cost. This is espe- the full-length bitstream and 25 dB for the base layer portion
cially likely to happen if the step sizes are very irregular. This of the bitstream. The intermediate bitstream rates of every
is a problem inherent to the discrete nature of the system, quality layer of each video sequence are shown in Table 3.
and it is well known that for such problems finding the op- Since network congestion influence on the performance-
timal solution can be very hard. This is similar to the known energy tradeoff is not the focus of the current paper, we limit
10 EURASIP Journal on Wireless Communications and Networking

Table 3: Bitrate settings of a different video sequence.

Sequence name Bus City Foreman Mother and daughter

Bitrate of the 1st layer 256 kbps 64 kbps 96 kbps 64 kbps
Bitrate of the 2nd layer 384 kbps 128 kbps 112 kbps 80 kbps
Bitrate of the 3rd layer 448 kbps 256 kbps 192 kbps 96 kbps
Bitrate of the 4th layer 512 kbps 384 kbps 288 kbps 112 kbps
Bitrate of the 5th layer 1028 kbps 448 kbps 384 kbps 128 kbps

Table 4: Different PER’s influence to received video quality.

PER BUS encoded at 32.4 dB Mobile encoded at 32.7 dB Foreman encoded at 34.3 dB
0.05 31.0406 31.8462 32.8051
0.01 32.1733 32.3564 34.1034
0.005 32.1845 32.695 34.1606
0.001 32.3422 32.7484 34.2982

the number of users in the network so that their require- 0.1

Transceiver energy (J)

ment can be satisfied. In the real-time variant channel simu- 0.08
lation, we use the best possible quality transmission config- 0.06
uration when the channel state is very bad and therefore the
0.04
quality requirement can almost never be reached. The aver-
age quality results turn out to always match the requirement 0.02
well. 0
Each quality layer of the bitstream is encapsulated into City Bus Foreman Mother
a separate network packet. Thus, if one network packet is Sequence names
dropped, the corresponding quality layer is lost. Every net-
Expected PSNR
work packet is further fragmented in MAC Service Data
Constant PER
Units (MSDU) of 1500 Bytes at link level. The maximal num- SoA
ber of retransmission is limited to 10 times.
An indoor channel model for the 5 GHz band [36] was Figure 5: Impact of video content.
used assuming a terminal moving uniformly at a speed be-
tween 0 to 5.2 km/h (walking speed). A set of 1000 time-
varying frequency channel response realizations (sampled (ii) Design-time + run-time approach 1: “constant PER”:
every 2 ms over one minute) were generated and normalized this is the approach introduced in [24, 25]. With this
in power. The bitstream was modulated using a turbo-coded strategy, every video packet is transmitted with a con-
802.11a OFDM PHY. The resulting PHY dynamics were then figuration resulting in a PER lower than 1e-2 until the
mapped to an 8-state Markov model, as detailed in [28]. transmitted bitstream reaches the quality required by
In Table 4, we show the network packet error rate (PER)’s the users. Instead of always transmitting the packets
influence to the received video quality. From the comparison with the highest possible data rate, an optimal sched-
of values in this table, we reach the conclusion that with PER ule exploring the tradeoff brought by link layer scaling
lower than 1e-2, the video can be regarded as correctly re- and sleeping is introduced.
ceived. When calculating the configuration at design-time, (iii) Design-time + run-time approach 2: “expected PSNR”:
to further assure the stable visual quality, the first quality this is the approach introduced in this paper. In this
layer is always given as a configuration with error probability transmission strategy, we introduce the expected vi-
lower than 1e-2. The sequence has been iteratively transmit- sual distortion into the design-time Pareto frontier cal-
ted more than 10 times to get relevant statistics. culation. By emphasizing differently the total energy
We consider in the sequel the following three transmis- minimization and fairness improvement for the run-
sion strategies. time algorithm, this transmission strategy can be fur-
ther differentiated into the following two schemes.
(i) SoA reference point: the server uses the highest feasi-
ble modulation in addition to code rate that enables to (a) Min total energy: the total energy consumption
transmit the packets with a loss probability lower than of the users’ terminal transceivers is minimized.
1e-2 (transmit as fast as possible). After successfully re- (b) Min-max energy: the maximum energy con-
ceiving and decoding the required video bitstream, the sumption among the users’ terminal transceivers
mobile devices are switched to sleep mode. This ap- is minimized.
proach aims at maximizing sleep duration. It is pro- The detailed results for the two proposed run-
posed in commercial 802.11 interfaces [37]. time schemes are discussed in Section 5.2.4.
Xin Ji et al. 11
Transceiver energy of foreman (J)

0.18 5.2.2. Impact of the channel status

0.16
0.14 The impact of the channel state is shown in Figure 6. The
0.12 channel state is again assumed to be constant during the
0.1 whole transmission. The foreman sequence is considered
0.08 here. The PER constraint is fixed to 1e-2 for the “constant
0.06 PER” approach. The QoS constraint is fixed to 35 dB for the
0.04 “expected PSNR” approach. Seven channels are used in this
0.02 test, where channel 1 is the best and 7 is the worst. From the
0 results, it is clear that the “constant PER” approach outper-
1 2 3 4 5 6 7
forms the SoA approach for all the channel states. The “ex-
Channel states
pected PSNR” approach enables further energy savings, be-
Expected PSNR cause video packets are treated differently based on their vi-
Constant PER sual relevance.
SoA For bad channel states, the energy cost of the “con-
stant PER” approach tends to be similar to that of the SoA
Figure 6: Impact of channel status. scheme. But for bad channel states, the “expected PSNR” ap-
proach provides the biggest savings. This is because the “ex-
0.35 pected PSNR” approach relaxes the PER requirement of low-
importance video packets.
Transceiver energy (J)

0.3
0.25
In Figure 7, we consider a time-varying channel and eval-
uate the energy cost of several video sequences (with differ-
0.2
ent rate-distortion properties) transmitted simultaneously.
0.15
The channel varies independently over all the users on a
0.1
GOP-by-GOP basis. For the “constant PER” approach, the
0.05
PER constraint is 1e-2. For the “expected PSNR” approach,
0 the QoS requirement used is 35 dB. Total energy consump-
Bus City Mother Foreman
tion after transmission is shown in Figure 7. Compared to
Sequence name
the static channels (see Figure 6), the time-varying channels
Expected PSNR cause a further increase in energy cost. In addition, the en-
Constant PER ergy cost further increases because of the multiuser scenario.
SoA Each packet has to be transmitted with lower TXOP to share
the bandwidth with other users, thus increasing the energy
Figure 7: Impact of a time variant channel and multiple user to the
transceiver energy of different schedulers. consumption to maintain the QoS requirement. Neverthe-
less, even under these conditions, the energy cost for each
user has been reduced approximately by a factor 2, compared
to the SoA approach. The “Expected PSNR” outperforms the
5.2. Results analysis “Constant PER” approach by another factor of 2.

The simulation results show that a significant energy de- 5.2.3. Impact of the different user requirements.
crease can already be achieved with the “constant PER” ap-
proach compared to the state-of-the-art approach. When the In this section, we present the impact of the different user
“expected PSNR” approach is used, simulation proves that QoS requirements on the energy cost. The four different se-
energy savings up to a factor of 2 can be achieved while main- quences with QoS requirements of 35 dB, 33 dB, and 31 dB
taining a uniform visual quality, thus significantly improving are tested simultaneously on the time-varying channel. The
QoS. In Sections 5.2.1–5.2.4, we show the detail results from energy cost of all these sequences after transmission is pre-
the aspects of different video content, channel status, user re- sented in Figure 8. It is clear that the lower the QoS require-
quirements together with the fairness discussion. ment is, the lower the energy consumption will be. The rea-
son is straightforward; the lower the quality is, the smaller
5.2.1. Impact of the video content the bitrate will be, and hence the lower the energy cost.
This shows once again that by taking into account the rate-
Figure 5 shows the influence of the video content on the en- distortion properties into the optimization system, we can
ergy cost for the different approaches listed above. The PER obtain more energy gains.
constraint is fixed to 1e-2 for the “constant PER” approach.
The QoS constraint is fixed to 35 dB for the “expected PSNR” 5.2.4. Fairness discussion
approach. As expected, the higher the bitrate of a sequence
is, the higher the transceiver energy cost will be. Those re- So far we have considered only one of the two proposed so-
sults are provided by delivering video on a fixed channel state lutions, that is, the total energy minimization. In this final
(channel state 2, with 40% occurrence probability). section we present the performance of the second proposed
12 EURASIP Journal on Wireless Communications and Networking

0.35 solution, that is, the fairness solution. In particular, we com-

0.3 pare the impacts of the two run-time algorithms on the en-
Transceiver energy (J)

0.25 ergy consumption.

0.2 In Figure 9, we consider 8 users simultaneously requiring
video streams with an expected PSNR of 35 dB. Each of the
0.15
4 videos is required by 2 users. No significant difference is
0.1
measured in terms of energy cost.
0.05 Figure 10 shows 16 users receiving video bitstreams si-
0 multaneously with an expected PSNR equal to 35 dB. Each

Foreman 31

Mother 31
City 31
Bus 31
Foreman 33
City 33

Mother 33
Bus 35

Bus 33
Mother 35
City 35

Foreman 35

of the 4 videos is required by 4 users. Under this setup, band-

width requirements are much more stringent than those in
the former setup, and the difference between the two run-
time approaches becomes significant. The results show that
Different visual quality requirements (dB)
the energy consumption of those users who require the max-
Expected PSNR imum energy cost decreases by about 3.5%, with the energy
Constant PER consumption of other users increasing by more than 30%.
SoA From the results presented in the previous sections, we
conclude that in general the minimal total energy approach
Figure 8: Impact of a time variant channel and multiple user with
reduces the energy cost by a factor of 2. Based on the results
different quality requirements to the transceiver energy.
presented in this section, we, therefore, recommend the min-
max energy approach unless the users are facing energy ex-
hausting issues.

0.08 6. CONCLUSIONS AND FUTURE WORK

0.07
Transceiver energy (J)

0.06 We have introduced a cross-layer optimization framework to

0.05 minimize the wireless transceiver energy consumption for
0.04 downloading multiple video streams over a WLAN. Rely-
0.03
ing on the IEEE 802.11 standard and scalable video coding,
0.02
the proposed solutions schedule the packets transmission by
both exploiting link layer scaling and sleeping trade-offs, and
0.01
integrating rate-distortion properties of the video sequences
0
1 2 3 4 5 6 7 8 into the optimization scheme. Results have shown that in
User numbers comparison with state-of-the-art approaches, the proposed
expected PSNR approach achieves stable visual quality ac-
Min total energy cording to the user QoS requirements, while largely decreas-
Min-maximum-energy
ing the energy cost. Compared to link layer optimization,
Figure 9: Energy comparison for 8 users requiring video bitstreams the proposed expected PSNR approach achieves energy gains
simultaneously. by a factor of 2. Fairness and total energy minimization ap-
proaches have also been discussed in this paper.
The work presented in this paper offers an insight view of
cross-layer optimization for energy-efficient bandwidth al-
location using scalable video coding. Our future work will
0.1
concentrate on the scalability provided by the different video
0.09
coding schemes. In particular, future work will focus on the
Transceiver energy (J)

0.08
0.07 scalable video coding (SVC) extension of H.264/AVC, the
0.06 upcoming state-of-the-art scalable video coding, which uses
0.05 different temporal and spatial decomposition schemes com-
0.04 pared to the one used in this paper. Also, in this paper, the
0.03 distortion calculation was GOP-based. Though calculating
0.02
0.01
the distortion based on packets will increase overhead and
0 complexity, it would be interesting to see how the proposed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 approaches perform under these conditions, despite the fact
User numbers that we expect that the general trends presented here will be
maintained. Additionally, we plan to investigate the perfor-
Minimize total energy
mance of the proposed approaches in uplink streaming, P2P
Minimize maximum energy
video transmission, and so forth. These scenarios have very
Figure 10: Energy comparison for 16 users requiring video bit- different timing constraints, hence requiring more interest-
streams simultaneously. ing optimization schemes and solutions. Finally, it would be
Xin Ji et al. 13

interesting to investigate multimedia applications over ad- [14] S. Chandra and S. Dey, “Addressing computational and net-
hoc networks and identify the energy and congestion con- working constraints to enable video streaming from wireless
trols needed to optimize the performance of these systems. appliances,” in Proceedings of the 3rd Workshop on Embedded
Systems for Real-Time Multimedia (STMED ’05), pp. 27–32,
New York, NY, USA, September 2005.
ACKNOWLEDGMENT [15] H. Yousefi’zadeh, H. Jafarkhani, and M. Moshfeghi, “Power
optimization of wireless media systems with space-time block
The work presented in the paper is partly based on results codes,” IEEE Transactions on Image Processing, vol. 13, no. 7,
published in ICASSP 2007. pp. 873–884, 2004.
[16] T. Simunic, H. Vikalo, P. Glynn, and G. De Micheli, “Energy
efficient design of portable wireless systems,” in Proceedings of
REFERENCES the International Symposium on Low Power Electronics and De-
sign (ISLPED ’00), pp. 49–54, Rapallo, Italy, July 2000.
[1] B.-J. Kim, Z. Xiong, and W. A. Pearlman, “Low bit-rate scal- [17] S. Cui, A. J. Goldsmith, and A. Bahai, “Modulation optimiza-
able video coding with 3-D set partitioning in hierarchical tion under energy constraints,” in Proceedings of IEEE Inter-
trees (3-D SPIHT),” IEEE Transactions on Circuits and Systems national Conference on Communications (ICC ’04), vol. 4, pp.
for Video Technology, vol. 10, no. 8, pp. 1374–1387, 2000. 2805–2811, Paris, France, June 2004.
[2] J. Xu, Z. Xiong, S. Li, and Y.-Q. Zhang, “Three-dimensional [18] J. T. Ludwig, S. H. Nawab, and A. P. Chandrakasan, “Low-
embedded subband coding with optimized truncation (3- power digital filtering using approximate processing,” IEEE
D ESCOT),” Applied and Computational Harmonic Analysis, Journal of Solid-State Circuits, vol. 31, no. 3, pp. 395–399, 1996.
vol. 10, no. 3, pp. 290–315, 2001. [19] A. Wang, W. R. Heinzelman, and A. P. Chandrakasan,
[3] H. Schwarz, D. Marpe, and T. Wiegand, “Overview of the scal- “Energy-scalable protocols for battery-operated microsensor
able H.264/MPEG4-AVC extension,” in Proceedings of IEEE In- networks,” in Proceedings of IEEE Workshop on Signal Process-
ternational Conference on Image Processing (ICIP ’06), pp. 161– ing Systems (SiPS ’99), pp. 483–492, Taipei, Taiwan, October
164, Atlanta, Ga, USA, October 2006. 1999.
[4] W. Li, “Overview of fine granularity scalability in MPEG-4 [20] E. Uysal-Biyikogly, B. Prabhakar, and A. El Gamal, “Energy-
video standard,” IEEE Transactions on Circuits and Systems for efficient packet transmission over a wireless link,” ACM/IEEE
Video Technology, vol. 11, no. 3, pp. 301–317, 2001. Transactions on Networking, vol. 10, no. 4, pp. 487–499, 2002.
[5] Q. Li and M. van der Schaar, “Providing adaptive QoS to [21] C. Schurgers, Energy-aware communication systems, Ph.D. the-
layered video over wireless local area networks through real- sis, University of California, Los Angeles, Calif, USA, 2002.
time tetry limit adaptation,” IEEE Transactions on Multimedia, [22] G. J. Foschini and Z. Miljanic, “A simple distributed au-
vol. 6, no. 2, pp. 278–290, 2004. tonomous power control algorithm and its convergence,” IEEE
[6] A. Ksentini, M. Naimi, and A. Gueroui, “Toward an improve- Transactions on Vehicular Technology, vol. 42, no. 4, pp. 641–
ment of H.264 video transmission over IEEE 802.11e through 646, 1993.
a cross-layer architecture,” IEEE Communications Magazine, [23] W. Ye, J. Heidemann, and D. Estrin, “Medium access control
vol. 44, no. 1, pp. 107–114, 2006. with coordinated adaptive sleeping for wireless sensor net-
[7] D. Skyrianoglou, N. Passas, and A. Salkintzis, “Traffic schedul- works,” IEEE/ACM Transactions on Networking, vol. 12, no. 3,
ing for multimedia QoS over wireless LANs,” in Proceedings of pp. 493–506, 2004.
IEEE International Conference on Communications (ICC ’05), [24] R. Mangharam, R. Rajkumar, S. Pollin, et al., “Optimal fixed
vol. 2, pp. 1266–1270, Seoul, Korea, May 2005. and scalable energy management for wireless networks,” in
[8] A. Sinha, A. Wang, and A. Chandrakasan, “Energy-scalable Proceedings of the 24th Annual Joint Conference of the IEEE
system design,” IEEE Transactions on Very Large Scale Integra- Computer and Communications Societies (INFOCOM ’05),
tion (VLSI) Systems, vol. 10, no. 2, pp. 135–145, 2002. vol. 1, pp. 114–125, Miami, Fla, USA, March 2005.
[9] A. J. Goldsmith, “The capacity of downlink fading channels [25] S. Pollin, R. Mangharam, B. Bougard, et al., “MEERA: cross-
with variable rate and power,” IEEE Transactions On Vehicular layer methodology for energy efficient resource allocation in
Technology, vol. 46, no. 3, pp. 569–580, 1997. wireless networks,” IEEE Transactions on Wireless Communi-
[10] D. Qiao, S. Choi, A. Jain, and K. Shin, “MiSer: an optimal low- cations, vol. 6, no. 2, pp. 617–628, 2007.
energy transmission strategy for IEEE 802.11a/h,” in Proceed- [26] P. A. Chou and Z. Miao, “Rate-distortion optimized streaming
ings of the 9th Annual International Conference on Mobile Com- of packetized media,” IEEE Transactions on Multimedia, vol. 8,
puting and Networking, pp. 161–175, San Diego, Calif, USA, no. 2, pp. 390–404, 2006.
September 2003. [27] IEEE Std 802.11a, “Standard for telecommunications and in-
[11] M. Goel, S. Appadwedula, N. R. Shanbhag, K. Ramchandran, formation exchange between systems—LAN/MAN specific
and D. L. Jones, “A low-power multimedia communication requirements—part 11: wireless LAN medium access control
system for indoor wireless applications,” in Proceedings of IEEE (MAC) and physical layer (PHY) specifications,” IEEE, 1999.
Workshop on Signal Processing Systems (SiPS ’99), pp. 473–482, [28] B. Bougard, S. Pollin, G. Lenoir, et al., “Energy-scalability en-
Taipei, Taiwan, October 1999. hancement of wireless local area network transceivers,” in Pro-
[12] Z. He, Y. Liang, L. Chen, I. Ahmad, and D. Wu, “Power-rate- ceedings of the 5th IEEE Workshop on Signal Processing Advances
distortion analysis for wireless video communication under in Wireless Communications (SPAWC ’04), pp. 449–453, Lis-
energy constraints,” IEEE Transactions on Circuits and Systems boa, Portugal, July 2004.
for Video Technology, vol. 15, no. 5, pp. 645–658, 2005. [29] Microsemi LX5506 InGaP HBT 4.5–6.5GHz Power Amplifier.
[13] X. Lu, Y. Wang, and E. Erkip, “Power efficient H.263 video [30] J. G. Proakis, Digital Communications, McGraw-Hill, Berk-
transmission over wireless channels,” in Proceedings of IEEE shire, UK, 4th edition, 2001.
International Conference on Image Processing (ICIP ’02), vol. 1, [31] L. Benini, A. Bogliolo, and G. De Micheli, “A survey of design
pp. 533–536, Rochester, NY, USA, September 2002. techniques for system-level dynamic power management,”
14 EURASIP Journal on Wireless Communications and Networking

IEEE Transactions on Very Large Scale Integration (VLSI) Sys-

tems, vol. 8, no. 3, pp. 299–316, 2000.
[32] IEEE Standard 802.11-1999, Wireless LAN medium access
control (MAC) and physical layer (PHY) specifications, 1999.
[33] M. van der Schaar, Y. Andreopoulos, and Z. Hu, “Optimized
scalable video streaming over IEEE 802.11a/e HCCA wireless
networks under delay constraints,” IEEE Transactions on Mo-
bile Computing, vol. 5, no. 6, pp. 755–768, 2006.
[34] J. Lin, “Multiple-objective problems: pareto-optimal solutions
by method of proper equality constraints,” IEEE Transactions
on Automatic Control, vol. 21, no. 5, pp. 641–650, 1976.
[35] H. Everett III, “Generalized lagrange multiplier method for
solving problems of optimum allocation of resources,” Opera-
tions Research, vol. 11, no. 3, pp. 399–417, 1963.
[36] European Telecom Standards Institue, “Channel models for
HIPERLAN/2 in different indoor scenarios,” ETSI 3ERI085B,
1998.
[37] Atheros White Paper, “Power consumption and energy effi-
ciency comparisons of WLAN products,” 2003.