Reliable Battery-Aware Cooperative Multicasting for MBS WiMAX Traffic

doi:10.4236/ijcns.2011.410079

Paper Menu >>

Journal Menu >>

Int. J. Communications, Network and System Sciences, 2011, 4, 648-655

doi:10.4236/ijcns.2011.410079 Published Online October 2011 (http://www.SciRP.org/journal/ijcns)

Reliable Battery-Aware Cooperative Multicasting for MBS

WiMAX Traffic

Sara Moftah Elrabiei1, Moohamed Hadi Habaebi1,2

1Department of Computer Engineering, University of Tripoli (Formally Known as Alfateh), Tripoli, Libya

2Department of Electrical and Computer Engineering, International Islamic University Malaysia, Kuala Lumpur, Malaysia

E-mail: [email protected], [email protected]

Received August 22, 2011; revised September 14, 2011; accepted September 26, 2011

Abstract

Recently, relay agents selection schemes were introduced to support downlink multicasting in WiMAX sin-

gle frequency networks. Such schemes were devised to work cooperatively in order to facilitate reliable Mul-

ticast Broadcast Multimedia Services (MBMS) traffic delivery over wireless channels without any consid-

eration to mobile relay agents’ battery energy levels. In this paper, we introduce a battery-aware balancing

algorithm to operate in conjunction with these relay agents selection schemes proposed in the open literature.

A simulation model, used to present the effect of “before” and “after” the battery-awareness selection crite-

rion, highlighted the benefit of using such algorithms in prolonging network lifetime with emphasis on reli-

able delivery.

Keywords: Reliability, Multicasting, WiMAX, MBMS, Battery

1. Introduction

Cooperative communication, has gained momentum

nowadays, due to its effectiveness and importance in

assisting the communication among wireless subscrib-

ers, and its ability to enhance capacity of the mobile

wireless broadband through multi-hop forward capabil-

ity. Different users can act as cooperative partners or

relay agents (RAs) to assist each other in information

transmission, to combat fading in wireless networks,

leading to improvement in overall system performance

[1-6]. Power consumption is a significant issue and key

performance indicator in such a cooperative mobile

environment, because of the limited battery power

available for mobile termnals that constrains the con-

tinuous operation time of wireless devices [7]. Fur-

thermore, multimedia mobile application software

nowadays is more processor intensive, and most of the

WiMAX MBMS applications last for long durations of

time. Therefore, a critical challenge is how to make

efficient use of energy, and to extend the battery life-

time for those cooperative relays without jeopardizing

their QoS such as throughput and system reliability. In

[5], three different re-multicasting schemes were pro-

posed and studied, for IPTV traffic, in conjunction with

other schemes available in the open literature. All pro-

posed schemes divide the downlink frame into 2-phases

(e.g., the first phase is used to receive multicast data

from the base station (BS) and the second phase is used

to re-multicast it to neighboring mobile subscriber sta-

tions (SSs)) but they differ in how to select the RAs to

perform such process. Furthermore, none of the selec-

tion schemes considered the effect of battery drainage

on selected RAs throughout the re-multicasting duration.

The contribution of this paper is two-folds. At first, we

study the effect of using such re-multicasting schemes

without any consideration to the SS battery state and

highlight the effect of the selection process on average

node energy consumption levels in the network. Sec-

ondly, we propose a selection criteria energy level bal-

ancing algorithm and implement it into the three dif-

ferent RAs selection schemes and evaluate its effect on

system reliability. The rest of this paper is organized as

follows. Section 2 introduces the effect of battery

drainage of RAs in battery unaware network using our

MATLAB built simulator. Section 3 discusses the bat-

tery aware RA selection optimization schemes. Section

4 introduces the simulation model used. Section 5

evaluates the performance of the proposed schemes and

discusses the simulation results, and Section 6 con-

cludes the paper.

S. M. ELRABIEI ET AL.649

2. Battery Drain of RAs in Battery Unaware

Network

To achieve a reliable downlink transmission for MBMS

traffic and to extend the coverage outside the trans-

mission area, three energy efficient re-multicasting

techniques were proposed in [5] for properly selecting

RAs in a two phase cooperative transmission model. At

phase I, the BS multicasts data to all SSs at high

transmission rate R1, where only subscribers in a good

channel state (SSGCS), e.g., as defined by their CINR

threshold, can successfully receive the data, and the

remaining group of subscribers in a bad channel state

(SSsBCS) fail to receive the data. BS preselects some of

SSGCS to be RAs using one of the selection algorithms in

Elrabiei et al. [5]. Upon receiving signals from BS, each

RA decodes the received signals and then forwards them

to SSsBCS at a proper rate R2 in phase II. By exploiting the

channel state information (CSI) and the location based

service (LBS), a Nearest-Neighbor Discovery Protocol

(NNP), and two other optimized versions of it, based on

RA’s transmission range and instantaneous CSI, were

simulated and studied in the context of WiMAX single

frequency networks. These protocols are described brief-

ly below as follow:

2.1. NNP RA Selection Scheme

The proposed relay agent selection protocol is based on

geographical positioning of users and on the instanta-

neous channel condition of the SSs. Since we have as-

sumed that BS knows the location of each SS, the as-

signed RA is chosen to be the nearest located SSGCS

(neighbor) to the SSBCS of interest, independently of

other SSsBCS. The NNP can be executed periodically

according to the mobility of the users and how often

they change their locations.

2.2. Transmission Radius RA Selection Scheme

The proposed NNP selects a RA for each SSBCS inde-

pendently of other RAs selection, by finding the nearest

SSGCS to the SSBCS from the same MGroup. In worst case

scenario, there is independent RA for each SSBCS, (i.e.,

number of RAs equal number of SSsBCS). Therefore, to

reduce the number of RAs selected, a scheme optimiza-

tion is proposed. If there are two or more SSsBCS, which

are in close proximity to each other, the BS can select

one RA for all of them if they are located within the

RA’s transmission radius. In other words, BS chooses

the RA with Transmission Radius arg min

di



the SSsBCS. For instance, if we consider da1, da2, are the

distances between RAa and SS1BCS, SS2BCS respectively,

and db1, db2, are the distances between RAb and SS1BCS,

SS2BCS respectively, then we select the RA which has:

min(da1 + da2, db1 + db2). The objective of the optimiza-

tion is to minimize the number of RAs as much as possi-

ble, by selecting the SSsGCS that can support re-mul- ti-

casting to the largest number of SSsBCS within its cover-

age transmission range, and not necessary be the nearest

SSGCS neighbor to each SSBCS. The template is used to

format your paper and style the text. All margins, column

widths, line spaces, and text fonts are prescribed; please

do not alter them. You may note peculiarities. For

example, the head margin in this template measures

proportionately more than is customary. This measure-

ment and others are deliberate, using specifications that

anticipate your paper as one part of the entire journals,

and not as an independent document. Please do not re-

vise any of the current designations.

2.3. SS-SS Inter-Link CSI RA Selection Algrithm

The proposed Transmission Radius RA selection algo-

rithm proposed in 2.2 section above reduces further the

number of RAs involved in the re-multicasting process

but at the expense of algorithm reliability. This is be-

cause some of the SSs previously enjoying their own

nearby RA are now located at the edge of transmission

range of the new RA. Adding that to the varying channel

state, it would result into less reliable delivery. In order

to compensate for that and to further reduce the number

of RAs involved, each SS in the cell is assumed to report

also its instantaneous channel quality between itself and

its neighboring SSs (i.e., SS-SS inter-link CSI) that can

overhear in the cell. Such SS-SS inter-link CSI aggrega-

tion can help the BS to select the proper SSsGCS, that have

maximum number of SSsBCS within in its transmis-

sion range and have good inter-link CSI, to support the

re-multicasting. Such technique requires further modify-

cations to the Wimax Channel Quality Indicator Channel

(CQICH) but the implementation is out of the scope of

this paper.

The RAs are selected so that channel condition

between them and all SSsBCS has to be in a good state.

Using SS-SS instantaneous Inter-link CSI, R2 in phase II

is determined based on computing the sum of the instan-

taneous received power from all corresponding RAs

using the large-scale and small-scale fading channel

model, by selecting the Adaptive Modulation and Coding

(AMC) level that support the worst CSI between SSsBCS

and the RAs. Further details can be found in [5].

The above presented schemes were found to consid-

erably reduce the amount of energy consumed [5], provi-

ding a lower cost coverage solution with no dereliction in

achieved throughput for all multicast group members. In

this study, we have assumed throughout our proposed

S. M. ELRABIEI ET AL.

650

work that in each frame, the BS selects the proper RAs

for re-multicasting depending on one of the selection

techniques discussed above and that the RA’s battery

capacity reduces due to re-multicasting process only. But

no prior formulation was made to estimate the energy

consumption levels or to predict the network energy map.

There have been several attempts to address the above

issue mentioned especially in the context of wireless

sensor networks, where the main source of energy is bat-

teries. Nevertheless, tackling the process of estimating

the energy levels at all SSs in the network required in a

quantitative closed formula is, somewhat, cumbersome.

However, we decided to use qualitative simulation to

highlight the effect of our proposed re-multicasting tech-

niques or energy consumption levels. The notations used

throughout the paper are listed in Table 1.

3. Proposed Battery-Aware RA Selection

Scheme

The power consumption and the duration of how long the

SS should work as a RA can be adjusted dynamically by

the BS. The accurate knowledge of the available energy

levels in each SS in the network (i.e., energy map of the

whole network) is important information for the BS to

make its selection. This can be implemented either by

allowing the SS to report its battery energy level as extra

information sent to the BS, or estimated by default as the

BS runs the RA selection algorithms and has a priori

knowledge of the RAs selection frequency rate. We have

resorted to the first assumption in our simulation study.

All SSs are involved in aggregating the CSI map of the

network to the BS. Therefore, it is similarly possible to

Table 1. Notations definitions (true unless stated otherwise).

Notation Description

RA Relay agent

SSBCS Subscriber station with bad channel state

SSGCS Subscriber station with good channel state

SSi,j The j-th group member in MGroupi

The rate of the BS in Phase I for MGroupi

The rate of each cooperative transmitter in Phase II for

MGroupi

T The transmission time of Phase I for MGroupi

T The transmission time of Phase II for MGroupi

 Service probability for MGroupi

,ij

E The received signal power for SSi,j in Phase I

p The power

di Distance between SSi and BS

bn The lower boundaries of signal to noise ratio (SNR) for

MCS level n

N0 The noise power

map the network mobile SSs battery draining levels at

least in a coarse grained fashion. Selected RAs can be

considered based on their CSI, geographic locations in

the cell, and their battery draining levels. SSsGCS that

have low battery levels can be spared from the down-link

phase II re-multicasting process until recharged. SSs with

stationery position and infinite power supply can act as

RAs similarly to stationary RSs proposed in 16j standard,

although they don’t have to re-multicast continuously on

every frame. Similar schemes are used extensively in

routing network traffic over wireless links. Therefore we

proposed the following battery aware RA selection algo-

rithm described in the following pseudo-code:

Pseudo Code for Battery-Aware RA Selection Algorithm

chedule the appropriate MGroupi using algorithm described in [6]

ocate all the SSsBCS

ocate all the SSsGCS

or each frame, collect the battery capacity levels in all SSsGCS

Compute the average battery available in all mobile SSsGCS.

liminate all SSGCS that have battery capacity less than the computed

average battery capacity (e.g., average battery capacity of the cell is

computed by taking the average of battery capacity counter values of

all mobile SSsGCS involved in the RA selection process only) from RAs

elections

elect the RAs from the new set of S

GCS based on one of the proposed

A selection algorithms described in [5]

etermine T1, R1, T2, R2 described in [5]

t phase

: BS transmits with R1 at T1

t phase I

: All RAs retransmit with R2 at T2

ecremen

the capacity counter value of the battery (drain_level) for

each RA selected based on the formula drain_level = T2/(T1 + T2)

Once battery capacity levels of all SSsGCS are aggre-

gated to the BS, the algorithm simply excludes certain

SSsGCS, which have battery energy levels below the av-

erage battery capacity of all mobile SSsGCS in the cell,

from the RA’s selection process. Such approach relieves

these SSsGCS from draining their batteries and averages

out the RA selection process, thus, prolonging the net-

work lifetime and, consequently, improving network

reliability in general.

Once battery capacity levels of all SSsGCS are aggre-

gated to the BS, the algorithm simply excludes certain

SSsGCS, which have battery energy levels below the av-

erage battery capacity of all mobile SSsGCS in the cell,

from the RA’s selection process. Such approach relieves

these SSsGCS from draining their batteries and averages

out the RA selection process, thus, prolonging the net-

work lifetime and, consequently, improving network

reliability in general.

We have formulated a general term for the probability

S. M. ELRABIEI ET AL.

651



Table 2. System parameters.

of any randomly picked SSj becoming a RA in the

MGroupi that is updated in the battery-aware network [8],

and can be estimated using the following formula:

PRA(SSij) =·

1

()

ij n

PE b





,12

min, , ,

BCS i

jSSN

Pdddd 

.

P (battery power > threshold) (1)

where PRA(SSij) is the probability of a SSi to be a RA in

MGroupi, i is the Service Probability for MGroupi

given by [6, Equation (6)],







,0ij n

PE bN

j = SSGC



is the prob-

ability that SS has a good CSI (i.e. SS ), the term

,12

BCS i

jSSN

 is the probability that

SSj is nearest neighbor to the SSBCS, and the term P(battery

power > threshold) is probability that the battery capacity

at the SSj is above a threshold value set by the BS to ensure

the ability of re-multicasting. Further details on the

mathematical formulation are presented in [8].







n, ,Pdd d mi

Number of cells 1

Carrier frequency & BW 2.5 GHz, 10 MHz

Cell radius 2.56 Km

The total number of SSs in the system 200

The number of MGroups M 16

The number of group members in each MGroup Ni Varying

BS maximum transmission power 20 W

SS maximum transmission power 3 W

Noise power −99 dBm

Path loss exponent α 4.375

Close-in reference distance d0 100 m

Lognormal Shadowing 8.9 dB std

Coverage ratio C 50%

Transmitter antenna gain Gt 1

Receiver antenna gain Gr 1

System loss L 1

DL/UL sub-frame duration 1.25 ms/1.25 ms

OFDMA symbol duration τ 102.9 μs

Frame duration 5 ms

Time ratio for multicasting 25%

4. Simulation Model Table 3. Modulation and coding schemes for IEEE 802.16.

Level Modulation and

coding bn (dB) rn (Mbps) Surface

ratio %

0 No Tx 0 0 27.0782 %

1 BPSK (1/2) 3 2.52 19.7459 %

2 QPSK (1/2) 6 5.04 12.3036 %

3 QPSK (3/4) 8.5 7.56 11.0675 %

4 16 QAM (1/2) 11.5 10.08 9.1849 %

5 16 QAM (3/4) 15 15.12 6.3544 %

6 64 QAM (2/3) 18.5 20.16 3.3007 %

7 64 QAM (3/4) 21 22.68 10.9648 %

We have used in this work a channel model that

incorporates both large-scale and small-scale fadings that

are represented by propagation path-loss, shadowing, and

random Rayleigh fading channel [9,10], and no mutual

interference is assumed because the SSs use the

orthogonal channel. As for the adaptive modulation and

coding, the combinations of modulation and coding set

(MCS) levels for the IEEE 802.16 air-interface are

shown in Table 3 in the first four columns, where bn

represents the lower boundaries of SNR (receiver

minimum sensitivity level), and rn is the corresponding

discrete peak transmission rate for the state n used in this

work for 10 MHz system. State 0 represents the state that

no transmission is allowed, which occurs when there is

poor channel condition.

5. Performance Evaluation

5.1. Battery Energy-Unaware System

Performance

When analyzing our results it appeared that some mobile

SSs are selected as RAs more often than their counter-

parts especially when these SSs were located at places

with line of sight connection from the BS and/or have

always good channel state. This frequent selection re-

sulted in the depletion of their limited energy sources

faster than the average depletion rate of the network.

We compare the performance for the three systems,

before and after battery-awareness situations, using

extensive simulations with Matlab. In our simulated

system, the network is composed of one BS and a

number of SSs which are randomly and uniformly

distributed over the coverage area of the BS (i.e., a circle

centered at the BS with a radius of ≈ 2.5 km). There are

different MGroups in the network and each group

includes some members which are randomly selected

from the whole SSs set. Channel model described in

above is applied, and the scheduling scheme for

MGroups proposed in [4] is utilized so that the

throughput performance and fair scheduling is not

jeopardized. The main system simulation parameters are

listed in Table 2 and their descriptions are discussed in

[5]. We have repeated the simulation several times with

different random seeds to calculate the average results.

In Figure 1, the histogram illustrates the number of times

a 50 SSs sample were selected as RAs through 600 frames

duration time. Increasing the SS density in the cell, resulted

in reduction of their severed energy consumption levels but

the general trend persisted. Over time, the discharging

process drained their batteries and they were automatically

eliminated from the network population. Thus, reducing the

system reliability in general and bringing the network

lifetime to a lesser reach.

In order to investigate the effect of battery levels on

S. M. ELRABIEI ET AL.

652

Figure 1. Histogram of number of times the SSs are selected as RAs through the 600 frames.

the network reliability, we have set the number of SSs in

the cell to 50 SSs, 10% of them are stationary with

unlimited power supply (e.g., their battery levels are un-

affected by the re-multicasting process), while the other

90% are mobile, slowly cruising around the cell. In order

to highlight the performance of the three different

schemes introduced in [5], we have isolated the effect of

the re-multicasting process on battery state by assuming

an initial value of battery state and each re-multicasting

frame results in a amount of energy consumed, where, T1

and T2 are the first and the second phase of the transmis-

sion burst T assigned for downlink multicast transmis-

sion duration respectively, where T = T1 + T2. All other

energy consumption processes such as UL transmission,

DL unicast transmission, processing, etc., were ignored

in order to isolate the effect of battery drainage on sys-

tem performance. We have repeated the simulation sev-

eral thousand times in order to calculate the average re-

sults. The normalized average capacity per mobile SSGCS

is calculated for the mobile SSGCS only and not for all SSs

in the cell.

The effect of the re-multicasting process at each frame

on the average battery capacity per mobile SSGCS in the

network is plotted in Figure 2. The average battery ca-

pacity was normalized relative to the initial battery ca-

pacity value of the SS capacity. The results were plotted

for the NNP RA scheme, transmission radius RA selec-

tion scheme, and SS-SS inter-link CSI RA selection

scheme. The average normalized battery capacity is de-

creasing with increasing the time elapsed. Obviously,

decreasing the capacity in each re-multicasting process

depends on the duration of the phase II duration (T2).

After 1000 frames the battery capacity per mobile SSGCS

reduces to half of the initial value using the optimized

selection techniques, whereas the NNP the capacity re-

duces only 20% for its initial value. This is due to the

fact that our optimized schemes selected RAs based on

distance and CSI criteria that fit fewer SSsGCS. These

SSsGCS were used more often depleting their energy lev-

els quickly in comparison to other SSs bringing in the

process the network average battery state to lower levels

faster. On the other hand, the NNP algorithm used the

simpler selection criterion by allowing more RAs in the

network to be replaced by other RAs more frequently,

thus resulting in smoother average battery energy con-

sumption throughout the cell.

The same results were confirmed by the average

network reliability performance over time as shown in

Figure 3. The general trend is that the reliability decreased

with increasing the time elapsed, and overtime, some RAs

discharged all their battery capacity, hence reducing the

number of RAs in the network. The optimized schemes are

often deeply affected in comparison to the NNP, because

they have fewer RAs, when the stationary RAs are not

affected at all. After 1000 frames, the reliability of the

optimized schemes decreased to about 40%, while the

NNP reliability reduced only to about 70%.

5.2. Battery Energy-Aware System Performance

In this section, the performance of our proposed scheme

was calculated in terms of relative power consumption

levels and system reliability. Extensive simulations with

Matlab were conducted using the implemented bat-

tery-aware RA selection schemes. The same simulation

set-up was maintained in order to facilitate the “before”

S. M. ELRABIEI ET AL. 653

Figure 2. Average normalized battery capacity per mobile SSGCS versus frame number.

Figure 3. Reliability of the proposed schemes versus frame number.

and “after” comparison. The number of RAs is basically

dependent on the number of SSs in the cell, the location

distribution of SSsBCS within the cell and on the wireless

channel conditions. The normalized average battery ca-

pacity per mobile SSGCS in the network at each frame for

all proposed schemes, was computed and plotted in Fig-

ure 4. The implementation of the battery-awareness on

the three RA selection schemes demonstrated more bal-

anced performance over time achieving almost a steady

state.

Similarly, the more balanced performance achieved for all

three RA selection schemes is translated into slower

degradation in system reliability as illustrated in Figure 5.

The battery-aware SS-SS inter-link CSI selection scheme

achieved the best performance in comparison to the other

two schemes, but the overall results are much better than

those shown in Figure 3. This is because the BS always

monitors the battery state of the whole network allowing it

to fine tune the process of RA selection continuously,

hence, prolonging the network life time.

S. M. ELRABIEI ET AL.

654

Figure 4. Average normalized battery capacity per mobile SSGCS versus frame number for the battery aware optimized

scheme.

Figure 5. Reliability of the proposed schemes versus frame number for the battery aware optimized scheme.

6. Conclusions

The relay agent selection balancing algorithm, intro-

duced in this paper for downlink multicasting traffic over

WiMAX channels, demonstrated the effectiveness of

considering the battery energy levels of mobile sub-

scriber stations as an additional RA selection criterion

when supporting re-multicasting over 2-phased down-

link frames. Such added criterion has prolonged the net-

work lifetime by reducing the average node energy con-

sumption levels with emphasis on reliable MBMS traffic

delivery. Amongst the algorithms, the battery-aware

SS-SS inter-link CSI RA selection scheme achieved the

best performance in comparison to the other two NNP

and Transmission-Radius RA selection schemes. Never-

theless, the overall results demonstrated more reliable

S. M. ELRABIEI ET AL.655

delivery over longer periods of time.

7. References

[1] Y. Eisenberg and C. Logan, “Cooperative Communi cations

for Improved Throughput, Range and Covert ness,” Military

Communications Conference, IEEE MILCOM 2008, Con-

vention Center, San Diego, 16-19 November 2008, pp.

1-7.

[2] A. Nosratinia, T. E. Hunter and A. Hedayat, “Cooperative

Communication in Wireless Networks,” IEEE Comm-

unications Magazine, Vol. 42, No. 10, 2004, pp. 74-80.

doi:10.1109/MCOM.2004.1341264

[3] P. Liu, Z. Tao, Z. Lin, E. Erkip and S. Panwar,

“Cooperative Wireless Communications: A Cross-Layer

Approach,” IEEE Wireless Communications, Vol. 13, No.

4, 2006, pp. 84-92. doi:10.1109/MWC.2006.1678169

[4] Q. Zhang, J. C. Jia and J. Zhang, “Cooperative Relay to

Improve Diversity in Cognitive Radio Resource,” IEEE

Communications Magazine, Vol. 47, No. 2, 2009, pp. 111-

117. doi:10.1109/MCOM.2009.4785388

[5] S. M. Elrabiei and M. H. Habaebi, “Energy Efficient

Cooperative Multicasting for MBS WiMAX Traffic,”

21st Annual International Symposium on Personal, Idoor

and Mobile Radio Communications PIMRC’2010,

Istanbul, 26-30 September 2010, pp. 1717-1722.

[6] F. Hou, et al., “A Cooperative Multicast Scheduling

Scheme for Supporting Multimedia Services over IEEE

802.16 Networks,” IEEE Transactions on Wireless Com-

munications, Vol. 8, No. 3, 2009, pp. 1508-1519.

doi:10.1109/TWC.2009.080417

[7] A. Ephremides, “Energy Concerns in Wireless Networks,”

IEEE Wireless Communications, Vol. 9, No. 4, 2002, pp.

48-59. doi:10.1109/MWC.2002.1028877

[8] S. M. Elrabiei and M. H. Habaebi, “Energy Efficient

Cooperative Multicasting for MBS WiMAX Traffic,” 5th

International Symposium on Wireless Pervasive Computing

ISWPC’2010, Modena, 5-7 May 2010, pp. 600-605.

[9] J. G. Andrews, A. Ghosh and R. Muhamed, “Funda-

mentals of WiMAX,” Prentice-Hall, Upper Saddle River,

2007.

[10] Theodore S. Rappaport, “Wireless Communications: Prin

ciples and Practice,” 2nd Edition, Prentice-Hall, Upper

Saddle River, 2002.