Sunteți pe pagina 1din 9

4806 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO.

9, SEPTEMBER 2009
A Cooperative MAC Protocol with Virtual-Antenna
Array Support in a Multi-AP WLAN System
Yao Hua, Student Member, IEEE, Qian Zhang, Senior Member, IEEE, and Zhisheng Niu, Senior Member, IEEE
AbstractIn this paper we propose a cooperative-aware
medium access control (MAC) protocol in the newly emerged
Multiple Access Point (MAP) WLAN system, where each user
can associate with multiple APs. Leveraging the multiple associ-
ation feature, the spectrum efcient virtual-antenna array (VA)
cooperative strategy can be utilized. However, the APs selected
for VA transmission cannot serve packet transmission individ-
ually, which leads to spatial reuse inefciency. To balance the
tradeoff between spatial reuse efciency and cooperative gain, an
interference model that describes interference among cooperative
transmissions should be rst established. In this paper, a virtual
link model is proposed, where each virtual link is composed of
one end user, one combination of VA APs and one data rate to
represent one cooperative transmission. Based on this model,
a multi-cell virtual link scheduling problem is formulated to
achieve optimal system performance. Combined with local clique
searching procedure and physical carrier sensing adaptation
scheme, the dual decomposition of the scheduling problem can
distributively achieve the near optimal performance. Finally a
VA-based cooperative MAC (V-MAC) protocol is proposed to
implement the cooperative scheduling scheme. Simulation result
shows V-MAC signicantly improves the system throughput
meanwhile guarantees the system fairness.
Index TermsCooperative MAC, multi-AP WLAN, multi-cell
scheduling, physical carrier sensing adaptation.
I. INTRODUCTION
N
EXT generation wireless networks will go beyond
the point-to-point or point-to-multipoint paradigms of
classical cellular networks. Cooperative communication has
emerged as a promising approach to increase spectral and
power efciency, network coverage, and reduce outage prob-
ability. There are generally two types of cooperative trans-
mission strategies. The rst one is relay-based cooperative
strategy, i.e., cooperative nodes transmit separately to the user
after receiving the cooperative data from the source node
by overhearing or specic communication with the source
node [1][2][3]. Another one is virtual antenna array (VA)-
based cooperative strategy, i.e., cooperative nodes transmit
simultaneously with the source node in cooperative phase
and constitute a virtual-antenna array [4]. VA-based coop-
erative strategy is more spectrum efcient than relay-based
one because signals from all the VA nodes are superimposed
Manuscript received November 8, 2008; revised March 13, 2009; accepted
April 10, 2009. The associate editor coordinating the review of this paper and
approving it for publication was J. Zhang.
Y. Hua and Z. Niu are with Tsinghua National Laboratory for In-
formation Science and Technology (TNLIST), and the Department of
Electronic Engineering, Tsinghua University, Beijing P.R. China (e-mail:
huay05@mails.tsinghua.edu.cn; niuzhs@tsinghua.edu.cn).
Q. Zhang is with the Department of Computer Science, Hong Kong
University of Science and Technology (e-mail: qianzh@cs.ust.hk).
Digital Object Identier 10.1109/TWC.2009.081505
Internet
MAP
AC
WLAN Wired

Link
A
1
A
2
A
1
(VA node)
A
4
(VA node)
u
1
u
2
u
3
u
4
Fig. 1. Multiple-AP architecture.
in the cooperative phase, which remarkably enhances the
transmission robustness. However, stringent synchronization
requirement among VA nodes constrains its utilization.
Cooperative transmission strategies have also been utilized
into wireless local area network (WLAN) system [1][2].
Recently, with the extensive usage of WLAN devices, WLAN
networks demonstrate a trend of high density deployment.
Statistic data in [5] showed that in typical indoor environment,
some users can sense over fty access points (APs). In high
density environment, the average distances between adjacent
APs shrink such that each end user can nd closer APs
with better channel condition for cooperative transmission.
In this paper, we are particularly interested in the scenario
where APs are densely deployed. In our previous work, a
multi-AP (MAP) WLAN architecture is proposed to take
advantage of the high density feature [6]. As shown in Fig.1,
in MAP architecture, all the APs are connected via wired
line to a centralized AP coordinator (AC). By wired line
communication APs can exchange channel state information
(CSI) among the system. Unlike traditional WLAN, each
user can associate with multiple APs in MAP WLAN. Then
users can opportunistically select one associated AP with best
channel quality for transmission and the system performance is
enhanced due to spatial diversity gain. However, only simple
AP selection scheme is adopted in [6], left more advanced
cooperative strategies for future system improvement.
It is observed that VA-based cooperation strategy perfectly
suits for MAP architecture, especially for the downlink trans-
1536-1276/09$25.00 c 2009 IEEE
HUA et al.: A COOPERATIVE MAC PROTOCOL WITH VIRTUAL-ANTENNA ARRAY SUPPORT IN A MULTI-AP WLAN SYSTEM 4807
Phase 1 Phase 1
VA-based cooperation with MAP
Phase 2
Packet 1
Phase 1
Relay
Packet 1
Source
time
Relay-based cooperation
Packet 2
Packet 2
Phase 1 Phase 2
Virtual
Antenna
Packet 1
Source
VA-based cooperation without MAP
Phase 1 Phase 2
Packet 1
Packet 2
Packet 2
Phase 1
Phase 2
Packet 1
Source
Phase 1
Packet 1
Packet 2
Packet 2
Phase 1
Packet 1 Packet 2
Virtual
Antenna
Packet 3
Packet 3
Packet 4
Packet 4
Fig. 2. Virtual-antenna array cooperation v.s. relay cooperation.
mission. First, in MAP architecture, when downlink VA-based
cooperation is considered, each VA node (i.e., AP) can obtain
the cooperative data via wired line communication and the
wireless resource of VA nodes for cooperative data acquisition
is saved, as Fig.2 shows. Consequently the spectrum efciency
is signicantly enhanced. Second, AC can coordinate the
synchronization among VA APs, which cannot be realized in
traditional WLAN system.
Although MAP architecture combined with VA-based co-
operative strategy exploits higher spatial diversity gain, co-
operative transmission also brings the following two new
challenges. First, cooperative transmission may decrease the
spatial reuse efciency because the spatial diversity gain is
obtained at the cost of resource consumption of cooperative
APs. Second, co-channel interference is increased by VA-
based cooperative transmission, especially in high density
environment. Comparing direct transmission in traditional
WLAN, VA-based cooperative transmission brings extra inter-
ference to neighboring cells since more APs transmit packets
simultaneously.
To solve the aforementioned two problems caused by VA-
based cooperative transmission, we should rst establish an
interference model to demonstrate the interference among
cooperative transmissions. In [7], the node exclusive inter-
ference model is utilized, i.e., only links that have a node
in common interfere with each other. However, the wireless
broadcast nature is overlooked. In [8], the cooperation graph
model is proposed. But the combinations of all cooperative
transmissions should be listed, which leads to unacceptable
complexity. In [9], authors modeled the interference among
cooperative transmissions as a unit disk graph. However, the
impact of data rate selection on the interference relationship is
not exploited. Data rate selection and VA AP selection are the
two main factors to determine the interference relationship
among cooperative transmissions. The higher data rate is
adopted, the more liable the link is to be interfered and
vice versa. VA AP selection also affects the interference
relationship. For a given user, when more VA APs are selected
for VA transmission, the received signal strength is higher
and the interference range shrinks correspondingly. In this
paper, to jointly consider the two factors, a novel virtual link
model is proposed where each virtual link associates with
one end user, one combination of VA APs and one data
rate. Essentially a virtual link constitutes a Multiple Input
Single Output (MISO) channel with transmitter side CSI. Once
the transmission power is xed, the interference relationship
between any two virtual links is determined, based on which
a collision graph is established.
Based on the collision graph interference model, a co-
operative virtual link scheduling problem is formulated to
optimize the overall spatial reuse efciency. The scheduling
result achieves higher spatial reuse efciency by determining
1) when to serve which user; 2) which VA APs are involved
in the cooperative transmission; 3) which data rate is used.
Meanwhile, the severe co-channel interference problem in high
density environment is alleviated by scheduling interfering
links with orthogonal time slots.
To guarantee the system scalability, duality decomposition
is utilized to solve the original scheduling optimization prob-
lem such that the scheduling result is computed by each
user individually. Combined with distributed clique searching
procedure as well as physical carrier sensing (PCS) adaptation
scheme, a distributed contention-based MAC protocol V-MAC
is proposed to implement the VA-based cooperation into MAP
WLAN. Simulation result shows that the V-MAC protocol
signicantly increases the system throughput of MAP WLAN
while the fairness is maintained.
The rest of this paper is organized as follows: we rst
formulate the cooperative virtual link scheduling problem in
Section II. In Section III the distributed solution of the optimal
scheduling scheme is discussed. A distributed MAC protocol
based on modication of IEEE 802.11 standard is proposed in
Section IV to implement the scheduling scheme. Followed is
simulation results in Section V. Finally, Section VI concludes
the paper.
II. THEORETICAL MULTI-CELL VIRTUAL LINK
SCHEDULING OPTIMIZATION
In this section the virtual link based scheduling problem is
formulated to evaluate the performance of a multi-cell MAP
WLAN system.
A. System Model
Consider a multi-cell WLAN system with APs and
users randomly distributed in a certain area. The user set is
denoted as and the AP set is denoted as . MAP architecture
is adopted as introduced in Section I. Each user can associate
with at most APs, which is also the number of candidate VA
APs for each user. All the APs are connected to the centralized
coordinator AC by wired line, by help of which cooperative
data packets and CSI can be exchanged among APs.
The downlink transmission is considered. We assume that
each device, either AP or user, has only one antenna, but
several APs can transmit simultaneously for one user by VA-
based cooperative transmission. All the APs have identical
transmission power and work under the same channel.
IEEE 802.11a physical standard is adopted [10], where one
of eight data rates in the rate set can be used as long as the
corresponding required Signal to Interference and Noise Ratio
(SINR) is satised. Saturated trafc is assumed for each end
4808 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009
user. In fact, since the scheduling scheme is updated within
short periods (in the scale of milliseconds), for the unsaturated
scenario, only the active users are scheduled and each active
user is considered to have saturate trafc during the period.
Therefore the scheme can be extended to unsaturated case.
The i.i.d. block fading channel is assumed, i.e., the channel
gain is static during each channel coherence time

and
varies in different

[11].

is the instantaneous channel


gain from AP

to user

, which obeys normalized Rayleigh


distribution (0, 1).
0
is the addictive white Gaussian noise
power. The received signal strength at user

from AP

is
denoted as [11]:

(1)
where is a channel gain constant depending on antenna
height and signal frequency.

is the distance between

and

. is the path loss coefcient, ranging from 2 (free


space) to 4 (full scattering indoor environment).
The set of virtual links in the system is denoted as . Any
virtual link in is denoted as

, where stands for the


th user of the system, stands for the th combination of
candidate VA APs and stands for the th data rate

is dened as the set of candidate VA AP combinations for user


(totally 2

1 combinations) and

is the th element of

.
Therefore the received effective signal strength is

is available if the following expression satises:

0
>

, ,

, (2)
where

is the required SINR threshold for

[19].
For any available virtual link

, since the effective receiv-


ing power and the required SINR are known, the maximum
tolerable interference power is:

0
(3)
Then the following interference model is established. Con-
sider any two virtual links

and

in . We dene two
virtual links interfere with each other as long as the aggregated
interference power of one virtual link exceeds the tolerable
interference of the other one. Also, two virtual links cannot
transmit simultaneously if they have some common APs. In
all,

and

interfere with each other if any of the three


conditions satises:

=
(4)
Then we examine each virtual link pair whether they interfere
with each other by Eq.(4). After that a non-directed contention
graph = (, ) is established to describe the interference
relationship among VA-based cooperative transmissions in
MAP WLAN system, where each vertex represents one
virtual link

and the edge between vertex and indicates


the virtual link

and

interfere with each other.


B. Multi-cell Scheduling Optimization Problem
As mentioned in Section I, multi-cell scheduling scheme
is a suitable candidate to solve the two problems caused by
VA-based cooperation, i.e., spatial reuse inefciency and inter-
cell interference. First, the scheduler decides which candidate
VA APs are selected for cooperative transmission according
to joint channel and interference conditions. The more non-
interfering cooperative transmissions are scheduled simultane-
ously, the better the spatial reuse efciency can be. Second,
the scheduler allocates orthogonal time slots for the links
interfering with each other, including both direct links and
cooperative links, so that the co-channel interference caused
by cooperative transmissions is eliminated.
Virtual link scheduling should determine how much time
ratio is allocated to serve each user, with which combination
of VA APs and using which data rate to optimize the system
performance. Since the multi-cell MAP WLAN system using
VA-based cooperation has been modeled as a non-directed
contention graph, the multi-cell virtual link scheduling opti-
mization problem is formulated based on the contention graph.
The scheduling ratio of each virtual link is denoted as

.
For any user , by aggregating the transmission rate of
all the virtual links containing

, the throughput of user is


derived:

, (5)
We aim at obtaining the scheduling ratio for each virtual
link to optimize the overall throughput. However, users at
the edge of the cells may always have bad channel quality
and spare their transmission opportunities for the users near
the APs. Therefore the system fairness cannot be guaranteed
by a throughput-oriented optimization result. A system utility
function is introduced as follows to guarantee the weighted
proportional fairness for each end user [12]:
U() =

log(

) (6)
where

is the weight to the user

considering the impor-


tance of the user. In the following work, we consider equal
importance for each user and

=1 for .
Now the conditions to guarantee a feasible scheduling result
are exploited. A scheduling result is feasible only if at any time
slot, no interfering virtual links are scheduled simultaneously
meanwhile the scheduling time ratio for each virtual link is
fullled. To interpret the feasible condition into the contention
graph model, the maximum clique concept is introduced [13].
One maximum clique is a set of virtual links such that no
two links in the set can be active simultaneously meanwhile
no more link can be added in the set to satisfy the same
condition. We denote the set of maximum cliques in G by
= {(1), (2), ...()}, where is the number of
maximum cliques in the system. Based on graph theory, if
a contention graph is perfect, the sufcient condition for a
feasible scheduling scheme is that for each maximum clique,
the aggregated scheduling ratio of all the links belonging to
this clique is less than one [14]. However, the contention graph
of a practical network cannot be guaranteed to be perfect. In
[15] authors pointed out that if the overall scheduling ratio
HUA et al.: A COOPERATIVE MAC PROTOCOL WITH VIRTUAL-ANTENNA ARRAY SUPPORT IN A MULTI-AP WLAN SYSTEM 4809
of links belonging to any maximum clique is less than 2/3,
there must exist one feasible scheduling, no matter whether the
contention graph is perfect or not. Therefore, the scheduling
feasibility constraint is written as:

()


2
3
, = 1, 2, (7)
Finally, the following optimal scheduling problem is ob-
tained:
max

log

..

()


2
3
= 1, 2

0
(8)
Up to now a lower bound of the system performance
is achieved by solving the optimization problem. Since the
problem is convex, it can be tackled by traditional convex
optimization solution such as interior-point method [16].
III. DISTRIBUTED IMPLEMENTATION OF VIRTUAL LINK
SCHEDULING SCHEME
The solution of the proposed optimization problem in Eq.(8)
requires very high computational complexity. To support
real time scheduling, the distributed solution of the optimal
scheduling problem is required. Therefore the primal problem
of Eq.(8) is decomposed by dual decomposition so that each
user can individually calculate its own scheduling result. Since
the optimization problem satises the Slaters condition, dual
decomposition can achieve the solution of the primal problem
without duality gap [16]. However, the complexity of nding
all maximum cliques is proved NP-hard [13]. A local clique
searching procedure is needed to nd cliques locally with low
complexity to approximate maximum cliques. Nevertheless,
the local clique searching procedure induces interference due
to imperfect scheduling. A distributed adaptive PCS adaptation
scheme is also proposed to alleviate the extra interference.
A. Duality Decomposition
The Lagrangean relaxation of the primal problem of Eq.(8)
is rst introduced. Let (, ) be the Lagrangean of the primal
problem with respect to the rst constraint. The dual problem
is described as follows:
maxmize
0
(, ) =

log

=1

(
2
3


()

)
(9)
where is the clique cost vector. Obviously, the dual problem
can be decomposed into optimization subproblems for each
user. () is denoted as the set of maximum cliques containing
virtual link . Then we optimize the following subproblem for
each user

:
maxmize
0

(, ) = log



:
(

()

,
(10)
Now the problem can be solved iteratively. At rst the
clique cost is assigned arbitrary non-negative values. Then
the subproblems of Eq.(10) are solved by classic convex
optimization algorithm and the temporal optimal scheduling
ratio in the iteration round is achieved. Substituting it to
the dual problem of Eq.(9), the objective function () is
obtained, which is continuous and twice differentiable. Thus
the clique cost is optimized iteratively by the gradient
projection method. We dene () as the scheduling ratio in
the th round. In ( +1)th iteration round, the clique cost

is updated by:

( +1) = max

0 ,

() +(
2
3

()

())

(11)
where is the step size. Then the updated is disseminated
to each user and Eq.(10) is recalculated for the next iteration.
The iteration terminates if the gap between the result of two
successive iteration rounds is less than a threshold. It is proved
that as long as is small enough, the iteration can converge
to the optimal value of problem Eq.(9) [16]. Thus the optimal
value for the dual problem is achieved in the distributed
manner.
However, there are 2

VA AP combinations, data rate


selections for each user and consequently as many as 2

virtual links, which make the iteration difcult to converge


with acceptable delay. To conquer the problem, we restrict
the maximal number of VA APs for each user to some
small number. In the following protocol design, the small
value of is utilized. Simulation results show that signicant
performance gain still can be achieved.
B. Local Clique Searching Procedure
Even through the original problem can be decomposed by
Eq.(10) to reduce the system complexity, the procedure to nd
all the maximum cliques in the system is still required, which
is extremely time consuming. Here a simplied distributed
clique searching procedure is proposed to nd the local cliques
to approximate maximum cliques. The procedure is as follows:
for each user

and the th combination of VA set,

mea-
sures the strength of all the beacon packets that can be decoded
successfully and calculates which virtual links can interfere
with

by Eq.(4). Data rate

is selected as the maximal


achievable data rate for the th combination of VA AP set so
that the system resource is allocated conservatively to avoid
co-channel interference. The set of observed interfering virtual
links plus itself constitute a local clique. By doing the same
procedure for each user and each combination of VA APs,
at most 2

local cliques (one local clique may include the


other one and the total number of local cliques can be less than
2

) are obtained, based on which the approximate optimal


scheduling ratio

is calculated.
However, local clique searching procedure leads to imper-
fect scheduling, i.e., interfering links may transmit simulta-
neously, which results in throughput penalty. The reason is
that the interference range of one link is often larger than
the communication range. Therefore interference signals from
the APs out of the communication range cannot be identied
by end users and the interference cannot be avoided by
the scheduling scheme based on local cliques. In the next
subsection, a PCS adaptation scheme is proposed to alleviate
such interference and the system performance is closer to the
perfect scheduling scheme.
4810 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009
C. Interference Mitigation of Imperfect Scheduling by Physi-
cal Carrier Sensing Tuning
The PCS adaptation scheme was widely used to combat the
interference problem in traditional WLAN system [17]. The
philosophy of the scheme is as follows: if the interference
level is high, the PCS threshold should be decreased to defer
its transmission more conservatively to avoid the interference
and vice versa. However, as the PCS threshold decreases, the
spatial reuse efciency is diminished because larger number of
APs are enclosed in the PCS range to share the same spectrum
resource. Therefore, all the PCS adaptation schemes tried to
nd the optimal tradeoff between the two factors.
Now the PCS adaptation is introduced into VA-based co-
operative WLAN system. The PCS threshold
cs
should be
tuned for each AP, but the transmission decision should be
determined by the VA AP set cooperatively. For any virtual
link, if any AP in its VA AP set senses the interfering power
above its own PCS threshold, the virtual link is deferred.
It is testied in [17] that packet loss rate (

) is a good
indication for PCS adaptation. When the summed

of
virtual links belonging to any AP is high, it implies that
the level of interference of the AP suffers is high and the
PCS threshold should be reduced so that the AP is more
sensitive to interference to prevent potential packet loss. If

is low, higher PCS threshold is preferred for the AP since


the transmission is less likely to be interfered by neighboring
transmissions. The PCS adaptation procedure is described as
follows:
The Procedure of Proposed PCS Adaptation
1: Measure
()

for each AP
2: for = 1 do
3: if
()


max
then
4:
()
cs
= max(
min
,
()
cs

)
5: else if
()


min
then
6:
()
cs
= min(
max
,
()
cs
+

)
7: end if
8: end for
where

are the packet loss indication thresholds


and

is the PCS power tuning granularity.


max
,
min
are the upper and lower bound of the PCS power. Simulation
results show that, by appropriately setting the parameters,
the distributed PCS adaptation can quickly converge to stable
operation point. Combined with dual decomposition and local
clique searching procedures, near optimal system performance
is achieved distributively.
IV. V-MAC: A COOPERATIVE MAC PROTOCOL TO
SUPPORT VIRTUAL-ANTENNA ARRAY COOPERATION
Since a distributed near optimal VA-based cooperative
scheduling scheme have been proposed, in this section we
focus on designing a VA-based cooperative MAC protocol
(V-MAC) to incorporate the proposed scheduling scheme in
MAP WLAN system. To keep the scalability of the proposed
system, the protocol is designed based on the modication of
the standard IEEE 802.11 protocol [10].
Frame
Control
Adress1 CRC
Virtual
link ID
Destination
Address
Adress2 Adress3
2 4 6 6 6 6 4
Association reply frame
Frame
Control
RSSI
AP
address
Destination
Address
Ocects 6 6 6 4
2
CRC
Association request frame
Frame
Control
Transmission
ID
Start time CRC
Virtual
link ID
Duration
Frame
body
2 2 4 2 1 4
1000
Synchronization frame
Fig. 3. Control frame format.
The proposed MAC protocol is divided into the following
four stages: association process, cooperation decision, coop-
eration preparation and cooperative transmission. Four stages
are introduced in parallel in the rest of the section. The MAC
process should be operated for each

so that the channel


state variation can be kept up with. The contention graph
should also be updated for each

.
A. Association Process
In MAP architecture, association process is operated to real-
ize multiple associations before the cooperative transmission.
At the beginning of each update period, associated process
is operated and the rest of the three are operated repeatedly
after the associate process. There are three functionalities to be
fullled in this stage. The rst is to determine the associated
APs of each user. Each user senses the signal strength of
beacon packets from all the APs periodically and piggyback
the measured channel quality to the AP. Then APs deliver
the CSI to AC by association request frame. The format of
association request frame is shown in Fig.3(a). The frame
contains a eld indicating the probed receiving signal strength
index (RSSI) of the AP from the end user indexed by the
destination address eld. CSI from all the APs is aggregated
to AC, which is responsible for selecting APs with best
channel quality to associate with each end user. Then AC
informs the association result to the APs by association reply
frame, the format of which is shown in Fig.3(b). The maximal
number of VA APs is xed to 3, which is discussed later.
Association reply frame has three address elds indicating the
three associated APs for the virtual link. Address 1 indicates
the AP with strongest signal strength, which is responsible
for acknowledge (ACK) reception and optimal scheduling
calculation of the user. The AP is referred to as agent AP
of the user.
The second task of the association process is to operate local
clique searching procedure described in Section III-B as well
as the PCS adaptation procedure proposed in Section III-C for
the following cooperative transmission.
Lastly, taking use of CSI of each link and the interference
relationship based on local cliques, the theoretical optimal
HUA et al.: A COOPERATIVE MAC PROTOCOL WITH VIRTUAL-ANTENNA ARRAY SUPPORT IN A MULTI-AP WLAN SYSTEM 4811
scheduling result is calculated by the duality decomposition
method in Section III-A. The scheduling result is calculated
iteratively. In each iteration round, AC rst calculates the
updated link cost by Eq.(11) for each maximum clique (local
clique in the practical solution) in the system and broadcasts
the result to each AP. Then, the agent AP of each user
collects all the clique costs related to the user and calculates
the temporal optimal scheduling ratio of the virtual links
belonging to the user by Eq.(10). The calculated scheduling
ratio is sent back to AC for the next round update. If
the difference of scheduling results between two consecutive
rounds is small enough, AC terminates the iteration process.
The signaling formats of exchanging temporal clique cost and
scheduling ratio between AC and agent APs are identical with
the association request frame except that the RSSI eld is
replaced with the updated clique cost or scheduling ratio value,
which are omitted here.
B. Cooperation Decision
When one downlink packet arrives at AC in the current
slot, AC automatically decides which virtual links to transmit
this packet. Since packets on the interfering virtual links
may be requested for transmission simultaneously, all the
virtual links contend for the transmission opportunity to
avoid collisions. The transmission probability for each virtual
link should be equal to the theatrical multi-cell scheduling
radio calculated in the association process stage. However,
the theoretical scheduling ratio cannot be directly fullled.
Therefore a contention mechanism is proposed to decide
which virtual links are scheduled in each slot. We modify the
backoff mechanism of IEEE 802.11 WLAN MAC and assign
a contention window for each victual link to guarantee that
the transmission probability of each virtual link is the same
as its theoretical scheduling ratio. It is proved in [18] that
the scheduling ratio for each virtual link is achieved if the
contention window for each virtual link is set as follows:
CW

= min(

, CW

) (12)
where is a scalar, and CW

is the maximum contention


window size of the IEEE 802.11 protocol. Once a transmission
is initiated, no matter whether it is collided or successful,
all other virtual links that are within the same clique with
the scheduled one should defer its transmission and reset
the backoff windows rather than freeze the backoff windows.
Through such contention mechanism, the average transmission
time ratio allocated to each virtual link is equal to the
optimal scheduling ratio calculated in the association process
phase. Since the number of contending virtue links increases
exponentially with the increase of number of users in the
system. is set proportional to to avoid extensive collisions
as increases.
C. Cooperation Preparation
As cooperation decision has been made, three crucial works
should be done to prepare for the VA transmission, i.e.,
cooperative data dissemination, channel reservation and VA
APs synchronization. All the three functionalities are realized
1 2 3 4 5
6
8
10
12
14
S
y
s
t
e
m

T
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
Maximal Number of VA APs
V-MAC, N=12
Fig. 4. System throughput versus maximal number of VA APs.
by the synchronization frame, which is broadcasted to the
whole network when one virtual link has been scheduled.
The format of synchronization frame is shown in Fig.3(c).
The cooperative data is transmitted to all of the VA APs by
wired line communications, which is included in the frame
body eld. What is more, the start time eld indicates the
exact start time of this cooperative packet so that all the VA
APs can synchronize. Once any AP receives the broadcast
synchronization data, it checks if it belongs to the VA AP set
indicated by the virtual link ID eld and determines if it will
transmit at the regulated slots. Last, all the interfering virtual
links are informed to defer during its transmission, where the
deference duration is indicated by the duration eld.
D. Cooperative transmission
Once a virtual link has been scheduled and the cooperation
preparation has been done, all the VA APs for the scheduled
virtual link start cooperative transmission. After receiving the
broadcasted synchronization signaling, any of the virtual links
will defer its transmission if either of the two cases happens.
First, the virtual link is contained in the same local clique
with the scheduled link; Second, any of the VA APs belonging
to the virtual link can sense the interference signal from the
scheduled link above its own PCS threshold. If the packet
is successfully received, the end user returns a unicast ACK
packet back to its agent AP. Then all the deferred virtual links
resume to idle state.
Since the bandwidth of wired line channel between AC and
APs is sufciently large compared with the wireless channel,
the overhead and delay of AC signaling transmission including
cooperative data dissemination is overlooked.
V. PERFORMANCE EVALUATION
In this section computer simulation is adopted to verify the
effectiveness of the proposed cooperative V-MAC protocol.
A. Simulation Setup
IEEE 802.11a protocol is adopted [10]. There are eight
available physical data rates, ranging from 6 Mbps to 54 Mbps.
Each virtual link attaches to one of the users, one of the
eight data rates and one of the 2

-1 VA AP combinations.
The default value of is four and

is 100 milliseconds,
4812 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009
0 2 4 6 8 10
0.0
0.2
0.4
U
p
d
a
t
e
d

S
c
h
e
d
u
l
i
n
g

R
a
t
i
o

Iteration round
User 1
User 2
User 3
User 4
User 5
(a) Iteration convergence of Eq.(10).
0.0 0.3 0.6 0.9 1.2 1.5
-20
-18
-16
-14
-12
-10
-8
A
v
e
r
a
g
e

P
C
S

p
o
w
e
r

t
h
r
e
s
h
o
l
d
(
d
B
m
)
Time (s)
V-MAC, N=5
(b) Iteration convergence of PCS adaptation.
Fig. 5. Performance of distributed procedure.
which are typical values for indoor environment [11]. AP
transmission power is xed to 100 mW and the white noise
power is 0.001 mW. The packet size is xed to 1000 Bytes.
is set to be .
Our simulation is based on random topologies in a 100100
meter square area. A varying number of APs and users are
randomly distributed in the area. Each user has saturated trafc
demand. The simulation result for each scenario is averaged
for 10 topologies and each topology with 100 update periods.
The performance of V-MAC protocol is compared with
classic relay-based cooperative MAC protocol CoopMAC [1].
In CoopMAC, each node can relay its overheard packet as
long as it discovers a two-hop relay path with higher data rate
than the direct transmission path. The performance of classic
non-cooperative MAP protocol [6] is also compared, where
each user associates with APs but can only select the one
with best channel gain for transmission.
The convergence of the proposed distributed schemes in
Section III is rst testied to guarantee the scalability of the
proposed protocol. System performances versus AP density
and channel coefcient are given respectively to observe the
performance gain of V-MAC in different wireless environment.
Finally, the fairness issue is discussed.
B. Simulation Result
The maximal number of VA APs should be rst de-
termined, which is left unsolved in the previous sections. A
12-AP 36-user topology is adopted to investigate the system
performance variation with parameter . The result is shown in
Fig.4. It is clearly shown from the gure that when is larger
than three, the tendency of throughput enhancement with the
increase of diminishes. That is mainly because the newly
added associated APs are farther away from the users as
increases and consequently have less probability with good
channel condition to the end user. Therefore, the additional
cooperative gain that can be achieved by including more APs
into VA AP set is trivial. On the other hand, the increase of
leads to higher computational complexity. By comprehensive
consideration, the value of is xed to three in the future
simulation.
6 8 10 12 14 16 18
0
5
10
15
20
25
30
35
S
y
s
t
e
m

T
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
Number of users
Optimal Scheduling
VMAC
CoopMAC
MAP
Fig. 6. Throughout versus user density.
The convergence of the distributed schemes proposed in
Section III is veried. We choose a 5-AP and 5-user scenario
for simplicity and calculate the updated scheduling ratio by
Eq.(11) for each iteration round. All the updated scheduling ra-
tios for each user are summed up to demonstrate the variation
of scheduling ratios for each iteration round. Fig.5(a) clearly
illustrates that by gradient projection method, scheduling ratios
can quickly converge to global optimal values. Then the same
topology is used to observe the convergence of the PCS
adaptation. It is seen in Fig.5(b) that the average per-AP PCS
power threshold can converge to a stable operation point. The
similar results for the convergence of the distributed procedure
are observed in all of the scenarios. Therefore, the optimal
scheduling result is proved to be achieved distributively and
the scheme scalability is guaranteed.
After that, the impact of AP density on the performance of
the proposed V-MAC protocol is investigated. The number of
APs varies from 6 to 18. The number of end users is three
times the number of APs in the following simulation. The
aggregated system throughput results of the three schemes as
well as the theoretical lower bound of the perfect schedul-
ing scheme by Eq.(8) are compared in Fig.6. It is shown
from Fig.6 that the perfect scheduling scheme has signicant
throughput gain over non-cooperative MAP and CoopMAC.
That is because the scheduling scheme can approach the opti-
HUA et al.: A COOPERATIVE MAC PROTOCOL WITH VIRTUAL-ANTENNA ARRAY SUPPORT IN A MULTI-AP WLAN SYSTEM 4813
2.0 2.5 3.0 3.5 4.0
0
5
10
15
20
25
S
y
s
t
e
m

t
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
Channel coefficient
Optimal Scheduling
V-MAC
CooPMAC
MAP
Fig. 7. Aggregated throughout versus .
mal tradeoff between spatial reuse efciency and cooperative
gain meanwhile the majority of interference is eliminated. The
other two schemes only optimize the transmission rate from
the single cell point of view. The performance of V-MAC
protocol achieves signicant performance gain compared with
the other two schemes as well. However, the performance of
V-MAC is lower than the theoretical lower bound. That is
mainly because we substitute maximum cliques with local
cliques, which leads to interference among simultaneously
scheduled links. The PCS adaption can only partially alleviate
the interference caused by local clique based scheduling. It is
also noticed the performance gain increases as the AP density
increases. That is because end users can nd closer APs for
VA transmission and the cooperative gain is enhanced as the
average distances among adjacent APs decrease.
The impact of channel fading on the system throughput is
also observed. The number of AP is xed to 12. The parameter
is varied from 2 to 4, which represents different wireless
environments. It is illustrated in Fig.7 that the channel fading
coefcient has little impact on the system performance of
non-cooperative MAP and CoopMAC. That is because as
increases, although the co-channel interference is alleviated
due to the increase of path loss, the effective signal strength
also diminishes which reduces the available data rate. The two
factors offset each other. In V-MAC and the perfect scheduling
scheme, system performance increases signicantly as channel
fading coefcient increases, which is mainly because the
interference decreases signicantly as the average path loss
grows. On the other hand, cooperative transmission can still
exploit the spatial diversity and opportunistically nd the VA
APs with high channel quality for cooperative transmission.
The result reveals that for different wireless environment, V-
MAC always has signicant performance gain.
Finally the fairness performance is observed. A random 5-
AP 5-user topology is used for simplicity and the through-
put fairness of the three schemes are compared in Fig.8.
It is clearly shown from the gure that the variance of
per-user throughput in V-MAC is much smaller than that
of the other two. That is because in both CoopMAC and
non-cooperative MAP, the throughput is optimized without
fairness consideration. There is a chance that some users are
at the edge of the cell and always have bad channel quality,
1 2 3 4 5
0.0
0.5
1.0
1.5
2.0
P
e
r

u
s
e
r

t
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
User index
V-MAC
CoopMAC
MAP
Fig. 8. User fairness.
or some links are starved by adjacent transmissions. The
performance of these users are always very low, like user
3 and 4 in this simulation topology. In V-MAC protocol,
since the fairness utility function is optimized, scheduling
decision can guarantee proportional fairness by compensating
resources of users with good channel quality for the ones with
bad channel quality. Take this scenario for example, V-MAC
procedure forces the users 1 and 4 to spare some resource
for the transmission of user 2 and 3 which are with bad
channel quality and the overall system fairness is signicantly
improved.
VI. CONCLUSION
In this paper, a virtual-antenna array (VA) based cooperative
scheme is developed for MAP WLAN system. To solve
both spatial reuse inefciency and co-channel interference
problem caused by VA cooperation, the virtual link concept is
proposed. Then a cooperative virtual link scheduling scheme is
proposed to obtain the theoretical optimal scheduling scheme,
which is implemented into a distributed contention-based
MAC protocol. Simulation results demonstrate that the system
throughput is increased by over 50% compared with the
classic relay-based cooperative MAC protocol and the non-
cooperative MAP protocol in all wireless environment while
the fairness of the system is guaranteed. The dual decompo-
sition, local clique searching procedure and distributed PCS
adaptation scheme provide good system scalability of the
proposed MAC protocol.
ACKNOWLEDGMENT
The research was supported in part by grants from RGC
under the contracts CERG 622508 and N HKUST609/07, by
grant from HKUST under the contract RPC06/07.EG05, by
grant from ITC HK ITP/023/08LP, the grant from Huawei-
HKUST joint lab, the NSFC oversea Young Investigator
grant under Grant No. 60629203, the Foundation of Scien-
tic and Technological Planning Project of Nansha District,
Guangzhou, China, the National Basic Research Program
of China (973 Program: 2007CB310607), NSFC-RGC joint
program(60731160631) and the International ST Cooperation
Program of China (ISCP) (No. 2008DFA12100).
4814 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 9, SEPTEMBER 2009
REFERENCES
[1] P. Liu, Z. Tao, S. Narayanan, T. Korakis, and S. Panwar, CoopMAC: a
cooperative MAC for wireless LANs, IEEE J. Select. Areas Commun.,
vol. 25, no. 2, pp. 340354 Feb. 2007.
[2] H. Zhu and G. Cao, rDCF: a relay-enabled medium access control
protocol for wireless ad hoc networks, in Proc. IEEE INFOCOM, Mar.
2005.
[3] Q. Zhang, J. Jia, and J. Zhang, Cooperative relay to improve diversity
in cognitive radio networks, IEEE Commun. Mag., vol. 47, no. 2, pp.
111117. Feb. 2009.
[4] G. Jakllari, S. V. Krishnamurthy, M. Faloutsos, P. Krishnamurthy, and O.
Ercetin, A framework for distributed spatio-temporal communications in
mobile ad hoc networks, in Proc. IEEE INFOCOM, Apr. 2006.
[5] A. Akella, G. Judd, S. Seshan, and P. Steenkiste, Self-management in
chaotic wireless deployments, in Proc. ACM MobiCom, Aug. 2005.
[6] Y. Zhu, Z. Zhou, B. Tan, Z. Niu, Q. Zhang, and J. Zhu, Leveraging multi-
AP diversity for transmission resilience in wireless networks: architecture
and performance analysis, to appear in IEEE Trans. Wireless Commun.,
2009.
[7] G. Sharma, R. Mazumdar, and N. Shroff, On the complexity of schedul-
ing in wireless networks, in Proc. ACM Mobicom, Aug. 2006.
[8] A. E. Khandani, J. Abounadi, E. Modiano, and L. Zheng, Cooperative
routing in wireless networks, in Proc. Allerton Conf. on Commun.
Control and Computing, 2003.
[9] Y. Zhu and H. Zheng, Understanding the impact of interference on
collaborative relays, IEEE Trans. Mobile Computing, vol. 7, no. 6, pp.
724-736, June 2008.
[10] Wireless LAN medium access control (MAC) and physical layer (PHY)
specications: high-speed physical layer in the 5 GHZ band, IEEE, 1999.
[11] T. Rappaport, Wireless Communications: Principles and Practice, 2nd
ed. Upper Saddle River, NJ: Prentice Hall PTR, 2001.
[12] X. Lin and N. Shroff, The impact of imperfect scheduling on cross-
layer rate control in multihop wireless networks, in Proc. IEEE INFO-
COM, Apr. 2005.
[13] R. Diestel, Graph Theory. Springer, 2005.
[14] M. Kodialam and T. Nandagopal, Characterizing achievable rates in
multi-hop wireless networks: the joint routing and scheduling problem,
in Proc. ACM MobiCom, Aug. 2003.
[15] Z. Fang and B. Bensaou, Fair bandwidth sharing algorithms based on
game theory frameworks for wireless ad-hoc networks, in Proc. IEEE
INFOCOM , Apr. 2004.
[16] S. Boyd and L. Vandenberghe, Convex Optimization. U.K.: Cambridge
Univ. Press, 2004.
[17] Y. Zhu, Q. Zhang, Z. Niu, and J. Zhu, On optimal physical carrier sens-
ing: theoretical analysis and protocol design, in Proc. IEEE INFOCOM,
Apr. 2007.
[18] B. Bensaou and Z. Fang, A fair MAC protocol for IEEE 802.11-based
ad hoc networks: design and implementation, IEEE Trans. Wireless
Commun., vol. 6, no. 8, pp. 2934-2941, Aug. 2007.
[19] X. Yang and N. Vaidya, On physical carrier sensing in wireless ad hoc
networks, in Proc. IEEE INFOCOM, Apr. 2005.
Yao Hua [S09] received the B.S. degree from
Xidian University, Xian, China. He is currently
working toward the Ph.D degree in electronic engi-
neering at the Tsinghua University, Beijing, China.
His research interests include MAC layer design
of wireless systems such as Wireless LAN and
cooperative networks.

Qian Zhang [M00, SM04] (qianzh@cs.ust.hk)
received the BS, MS, and PhD degrees from Wuhan
University, China, in 1994, 1996, and 1999, re-
spectively, all in computer science. She joined the
Hong Kong University of Science and Technology
in September 2005 as an Associate Professor. Before
that, she was at Microsoft Research Asia, Beijing,
China, from July 1999, where she was the research
manager of the Wireless and Networking Group.
She has published more than 200 refereed papers in
international leading journals and key conferences
in the areas of wireless/Internet multimedia networking, wireless communi-
cations and networking, and overlay networking. She is the inventor of about
30 pending patents. Her current research interests are in the areas of wireless
communications and networking, IP networking, multimedia, P2P overlay, and
wireless security. She has also participated many activities in the IETF ROHC
(Robust Header Compression) WG group for TCP/IP header compression.
Dr. Zhang is the associate editor for the IEEE TRANSACTIONS ON WIRE-
LESS COMMUNICATIONS, IEEE TRANSACTIONS ON MULTIMEDIA, IEEE
TRANSACTIONS ON VEHICULAR TECHNOLOGIES (2004-2008), COMPUTER
NETWORKS AND COMPUTER COMMUNICATIONS. She has also served as
guest editor for the IEEE WIRELESS COMMUNICATIONS, IEEE JOURNAL
ON SELECTED AREAS IN COMMUNICATIONS, IEEE COMMUNICATION
MAGAZINES, ACM/SPRINGER JOURNAL OF MOBILE NETWORKS AND AP-
PLICATIONS (MONET), and COMPUTER NETWORKS. She has also involved
in the organization committee for many IEEE and ACM conferences.
Dr. Zhang received TR 100 (MIT Technology Review) worlds top young
innovator award in 2004, the Best Asia Pacic (AP) Young Researcher
Award elected by the IEEE Communication Society in 2004, and the Best
Paper Award by the Multimedia Technical Committee (MMTC) of the IEEE
Communications Society and Best Paper Award in QShine 2006, IEEE
Globecom 2007, and IEEE ICDCS 2008. She received the Oversea Young
Investigator Award from the National Natural Science Foundation of China
(NSFC) in 2006. Dr. Zhang is Chair of the Multimedia Communication
Technical Committee of the IEEE Communications Society. She is also
a member of the Visual Signal Processing and Communication Technical
Committee and the Multimedia System and Application Technical Committee
of the IEEE Circuits and Systems Society.

Zhisheng Niu [M 98, SM 99]
(niuzhs@tsinghua.edu.cn) graduated from Northern
Jiaotong University, Beijing, China, in 1985, and
got his M.E. and D.E. degrees from Toyohashi
University of Technology, Toyohashi, Japan, in
1989 and 1992, respectively. In 1994, he joined
with Tsinghua University, Beijing, China, where
he is now a full professor at the Department of
Electronic Engineering and the deputy dean of the
School of Information Science and Technology
in charge of research activities and international
collaboration. He is also an adjunction professor of Beijing Jiaotong
University (formally Northern Jiaotong University).
From April 1992 to March 1994, he was with Fujitsu Laboratories Ltd.,
Kawasaki, Japan. From October 1995 to February 1996, he was a visiting
research fellow of the Communications Research Laboratory of the Ministry
of Posts and Telecommunications of Japan. From February 1997 to February
1998, he was a visiting senior researcher of Central Research Laboratory,
Hitachi Ltd. He also visited Saga University, Japan, and Polytechnic
University, USA, in 2001 and 2002, respectively.
Prof. Nius current research interests include teletrafc theory and
queueing theory, performance evaluation of broadband multimedia networks,
radio resource management, mobile Internet, wireless ad hoc networks and
wireless sensor networks, and Stratospheric Communication Systems. He
received the PAACS Friendship Award from the Institute of Electronics,
Information, and Communication Engineers (IEICE) of Japan in 1991
and the Best Paper Award from the 13th Asia-Pacic Conference on
Communications (APCC2007).
Dr. Niu is a senior member of the IEEE and the Chinese Institute of
Electronics (CIE) and a fellow of the IEICE, Japan. He has been the TPC
Chair of APCC2004 and the TPC Co-chair of IEEE ICC2008. He is now
the director of IEEE Communication Society Asia-Pacic Board.

S-ar putea să vă placă și