You are on page 1of 9

LU LAYOUT

8/8/11

12:09 PM

Page 13

ACCEPTED FROM OPEN CALL

MULTIDOMAIN COLLABORATION:
A NEW FRAMEWORK OF WIRELESS NETWORKS
WITH HIGH TRANSMISSION CAPACITY
JIANHUA LU, YIZHEN JIA, TENGFEI XING, AND XIAOMING TAO, TSINGHUA UNIVERSITY

Arrange multiple
transmissions
coordinately

Allocate transmit
power or schedule
transmissions
adaptively

Combine multiple
received signals
coherently

The bottleneck of
wireless communication lies in the contradiction between
limited spectrum
resources and everincreasing user
demand. Multidomain collaboration
has emerged as a
new framework to
solve the problem,
to some extent,
with a great
spectrum efficiency
improvement.

ABSTRACT
The bottleneck of wireless communication lies
in the contradiction between limited spectrum
resources and ever-increasing user demand. Multidomain collaboration (MDC) has emerged as a
new framework to solve the problem, to some
extent, with a great spectrum efficiency improvement. In this article, we introduce the concept of
MDC with the help of some intuitive examples.
Specifically, the underlying rationale of MDC for
increasing transmission capacity is revealed from
the perspective of both multiplexing gain and
power gain. Cases studies of point-to-point,
point-to-multipoint, multipoint-to-point, and
multipoint-to-multipoint transmissions are conducted as preliminary evaluations of the performance limits that may be achieved by MDC.

INTRODUCTION
After more than two decades of rapid development, wireless communication has become a pillar industry of the worlds high tech sector.
However, existing and upcoming systems are still
unable to mitigate the contradiction between
limited spectrum resources and ever-increasing
user demands, which has been regarded as the
bottleneck of wireless communication ever since.
Traditional wireless communication systems
are likely designed mainly based on independent
optimization of utilization of radio resources. As
a result, their capacities are mostly subject to the
law of diminishing returns, meaning that the
capacity increment becomes smaller and smaller
with the increase of resource of some kind [1].
Take a single-input single-output (SISO) with
additive white Gaussian noise (AWGN) channel,
for example. Its capacity is determined by the
Shannon formula, C = Wlog2 [1 + P/(nW)], where
W, P and n are the channel bandwidth, transmit
power, and noise power spectrum density, respectively. As shown in Fig. 1, the capacity will reach
to a limit as the bandwidth goes to infinity, or
lim C = P / n log 2 e.

IEEE Wireless Communications August 2011

Even though the capacity may increase without limitation on the growth of the transmit
power, the rate of increase decreases gradually
and eventually approaches zero. Another example may be given by cellular networks. Frequency reuse is supposed to provide more available
spectrum to each cell, while intercell interference becomes more serious in turn; thus, the system capacity becomes saturated. In fact, in most
practical systems, the relationship between a
resource input and the capacity holds almost the
same as the one depicted in Fig. 1.
In order to break through the bottleneck and
improve system capacity significantly, various
joint optimizations over multiple domains of
resources were proposed. For example, Foschini
[2] developed a novel signal architecture across
space and time domains, known as V-BLAST, to
realize a large capacity gain. Rhee and Cioffi [3]
studied a joint subcarrier and power allocation
problem in multiuser orthogonal frequency-division multiplexing (OFDM) systems to maximize
the sum capacity. Cai, Shen, and Mark [4] investigated a similar problem with the consideration
of heterogeneous traffic. However, rather than
just focusing on specific models and detailed
algorithms, how and how much in general joint
optimization over multiple domains yields capacity gain should be studied so that a guide light
may illuminate the roadmap of future wireless
system development.
This article proposes a new framework of
wireless networks with combined utilization of
multiple domains, named multidomain collaboration (MDC). MDC extends the traditional
concept of joint optimization to a more general
case, by exploring and taking advantage of positive interaction among multiple domains of radio
resources, such as the time, frequency, space,
and power domains. Within this framework, we
try to reveal the underlying rationale of collaborative use of multiple domains for increasing system capacity, while obtaining the fundamental
bounds of improved capacity gains in more general system models. It is worth noting that the
FLoWS project [5], a five-year Defense
Advanced Research Projects Agency (DARPA)

1536-1284/11/$25.00 2011 IEEE

13

8/8/11

12:09 PM

Page 14

Capacity (bit/s)

LU LAYOUT

Bandwidth (Hz)

Figure 1. An example of diminishing returns in wireless communication. Using


a fixed transmit power, the capacity of an AWGN channel will reach a limit
as the bandwidth goes to infinity.

program, is striving for similar objectives in


mobile ad hoc networks. In fact, recently the
endeavor to achieve enormous capacity improvement has become a significant trend in the
research on wireless communication systems.
The material in this article was presented in part
at IEEE ICWITS 2010.
There are several metrics of system capacity
for wireless networks. Considering that transmission capacity is basic, this article focuses on
increasing transmission capacity by means of
MDC. For a wireless link connecting one transmitter and one receiver, transmission capacity
refers to the (ergodic) channel capacity of the
link; for a network containing multiple pairs of
transmitters and receivers, transmission capacity
refers to the sum capacity of all the links. Since
the total available spectrum for a network is
likely fixed, we do not strictly differentiate transmission capacity and spectrum efficiency
throughout the article.
The remainder of this article is organized as
follows. We present an overview of MDC, from
its concept to practice with intuitive examples.
We evaluate the transmission capacity gains
obtained by MDC under four typical network
topologies (i.e., point-to-point, point-to-multipoint, multipoint-to-point, and multipoint-tomultipoint). Promising applications of MDC in
practical systems or their evolutions are also
investigated. Finally, concluding remarks with a
perspective of future work on MDC are given.

OVERVIEW OF MULTIDOMAIN
COLLABORATION
The word collaboration is commonly defined as
the process of coordinating different types of
individuals to fulfill a certain mission. As such,
in this article MDC refers to the process of
exploiting and making use of positive interaction
among multiple domains to significantly improve
the transmission capacity of wireless networks.
The domains include the basic radio resources
utilized for wireless communications (i.e., the

14

time, frequency, space, and power domains).


Below are two intuitive examples of MDC.
Collaboration between time and frequency
domains: Consider a simple case with one server
and three users. Suppose that the available
bandwidth is two units of spectrum, and the
achievable spectrum efficiency for each user is 1
Mb/s per unit of spectrum. All the users would
like to download their own favorite videos stored
in the server, while the download data rate for
each user is assumed to be constant during the
download process, say, 1 Mb/s. The time lengths
of the videos required by users U 1, U 2, and U 3
are two units, one unit, and one unit of time,
respectively. Furthermore, we assume that all
the users expect to enjoy their videos within two
units of time while ignoring the propagation
time. How do we allocate the radio resources
(i.e., available spectrum and time units) among
these three users? If the resources are assigned
to the users in one domain, either time or frequency, it is impossible to meet the requirements of all three users. In other words, neither
time-division multiplexing (TDM) nor frequency-division multiplexing (FDM) can serve these
three users within two units of spectrum and two
units of time (Fig.s 2a and 2b).
However, combining TDM and FDM together (i.e., assigning radio resources jointly in time
and frequency domains), one can easily find
more than one feasible resource allocation meeting all the users demands, as shown in Fig. 2c as
an example. Compared with TDM and FDM,
this joint TDM and FDM scheme may improve
the transmission capacity from 33 to 100 percent, which embodies the collaboration between
the time and frequency domains.
Collaboration among time, space, and power
domains: Assume a scenario where the users
have no waiting time restriction with sufficient
storage capacity such that the download data
rate may be flexible. All users experience different fading channels which are supposed to be
independent of one another (Fig. 3a). For such a
system, traditional scheduling schemes (e.g.,
round robin or polling) arrange transmissions to
all the users in a predetermined order regardless
of channel conditions. As a result, for one user,
a transmission may happen even when the channel quality is very poor during the scheduled
period, thus leading to low transmission capacity.
Inspired by the idea of MDC, however, one may
employ opportunistic scheduling [6] by transmitting, with all available power, to the user of the
best channel condition, most likely resulting in
improved capacity. This capacity improvement
by opportunistic scheduling may be understood
in two aspects. First, for each user, the randomness of its channel fading in the time domain
likely ensures the emergence of good channel
conditions where the channel strength is much
stronger than average. Second, the independence of the users channel fading in the space
domain makes it highly probable that there is
one user with good channel conditions in each
scheduled period.
An intuitive comparison between opportunistic scheduling and round-robin with an order of
U1 U2 U3 U1 is illustrated in Fig. 3b. It
is shown that in the period 1, U2 will have chance

IEEE Wireless Communications August 2011

Page 15

to receive data with opportunistic scheduling


since its channel condition is the best, but U1
will download data with round robin according
to the predetermined order. Because U2s channel strength (approximate 5dB) is much higher
than U 1 s (approximate 15dB), much larger
data rate may be achieved by opportunistic
scheduling in this period. This conclusion also
holds for other periods, implying a significant
capacity gain by the use of the opportunistic
scheduling, which may refers to a collaboration
among time, space and power domains.
The above examples may help to illustrate
the concept and advantage of MDC. The rationale for capacity improvement by MDC may be
revealed from the Shannon formula, which
expresses mathematically the dependence of the
capacity of an AWGN channel on the bandwidth
and the signal-to-noise ratio (SNR). Further
considering a network as a virtual giant link, one
may declare that the transmission capacity of the
network is determined by the effective bandwidth,
W E and effective SNR, E, and W E = K MW 0, E
= KP 0 = KPP0/N0, with W0, P0 being the given
spectrum bandwidth and total transmit power of
the network, respectively, and N 0 the noise
power. K M and K P are defined as multiplexing
gain and power gain, respectively, which depend
on the way of utilization of resources. Since W0,
P0, and N0 are fixed, one may improve the transmission capacity by increasing K M and/or K P ,
which is de facto the goal of MDC. Specifically,
note that in the first example presented above,
the joint TDM and FDM may increase the multiplexing gain up to two fold compared with the
TDM. While in the second one, the power gain
of opportunistic scheduling is nearly 100 times
(i.e., 20 dB) that of round-robin.
From the perspective of multiplexing gain
and power gain, a conceptual diagram of MDC
is depicted in Fig. 4. MDC actually serves as a
methodology of utilization of resources, according to which the transmitters (labeled T1, T2, )
and receivers (labeled R1, R2, ) make collaborative use of radio resources in time, frequency,
space, and power domains so as to increase the
multiplexing gain and/or power gain, and eventually the transmission capacity gain. Specifically,
at the transmitter side, the transmission of multiple data streams may be arranged coordinately
in time, frequency, or space domains, with proper power planning if needed, so that the assigned
spectrum may be reused as aggressively as possible without interference or under a tolerable
interference level. As such, a multiplexing gain is
obtained. Second, the transmit power or transmission opportunity may be adaptively allocated
according to the fluctuating channel characteristic. As a result, the signal-to-noise ratio (SNR)
of the received signal is increased and a power
gain is achieved. Third, at the receiver side, the
multiple copies of received signals possibly available in time, frequency, and space domains may
be combined coherently, which also yields a
power gain. Both the multiplexing gain and
power gain may be translated into transmission
capacity gain.
Below, we study some typical cases of wireless
network topologies, analyzing and evaluating the
capacity gains of MDC with achievable limits.

IEEE Wireless Communications August 2011

Time (unit of time)

12:09 PM

Basic
resource
block

0
0

1
Frequency (unit of spectrum)

(a)

Time (unit of time)

8/8/11

Basic
resource
block

0
0

Frequency (unit of spectrum)


(b)
2

Time (unit of time)

LU LAYOUT

Basic
resource
block
RB3

RB4

RB1

RB2

0
0

Frequency (unit of spectrum)

(c)

Figure 2. An example of collaboration between time and frequency domains.


Three users (U1, U2 and U3) apply for 1 unit of spectrum each, and a continuous period of 2 units, 1 unit, and 1 unit of time, respectively, while the available
resources include 2 units of spectrum and 2 units of time. Thereby, neither a)
TDM, which requires 4 units of time, nor b) FDM, which requires 3 units of
spectrum, can serve these three users. However, scheme c) with joint TDM and
FDM, for example, may assign the basic resource blocks of RB1, RB2, RB3,
and RB4 to U1, U1, U2, and U3, respectively, serving all users.

15

8/8/11

12:09 PM

Page 16

U1

Channel strength (dB)

-5

U2

-15

5
Channel strength (dB)

LU LAYOUT

U3
1

2
Schedule period

U2

U3

U1

U1

U2

U3

-5

-15

2
Schedule period

(a)

(b)

Figure 3. An example of collaboration among time, space and power domains. A server schedules transmissions to three users (U1, U2,
and U3), while two scheduling methods, the opportunity scheduling and round robin, are employed for comparison: a) the channel
fading of the three users, which is supposed to be independent of each other; b) Scheduling results of the opportunistic scheduling (solid
line) and round robin (dashed line), which shows an enormous advantage of opportunistic scheduling.An example of collaboration
among time, space and power domains. A server schedules transmissions to three users (U1, U2 and U3), while two scheduling methods, the opportunity scheduling and round robin, are employed for comparison: a) The channel fading of the three users, which is supposed to be independent of each other; and b) Scheduling results of the opportunistic scheduling (solid line) and round robin (dashed
line), which shows an enormous advantage of opportunistic scheduling.

CASES STUDIES OF MDC WITH


PERFORMANCE EVALUATION
Consider a general wireless network composed
of M t transmitters, labeled T 1, T 2, , T Mt, and
M r receivers, labeled R 1 , R 2 , , R M r . Each of
these nodes may be equipped with multiple
antennas. The channel coefficient from each
transmit antenna to each receive antenna is
assumed to be a zero mean circularly symmetric
complex Gaussian random variable with unit
variance. We then focus on a symmetric case
where the path loss from each transmitter to
each receiver is assumed to be the same and
normalized to one. Each of the receive antennas
suffers from AWGN. Following the notation
from earlier, the total transmit power and noise
power are denoted P 0 and N 0 , respectively.
Moreover, the available spectrum bandwidth is
normalized to one so that transmission capacity is
equal to spectrum efficiency in magnitude.
We now study the performance of MDC under
four cases with typical network topologies: pointto-point, point-to-multipoint, multipoint-to-point,
and multipoint-to-multipoint transmissions.
Specifically, in the first three cases, the emphasis
is put on collaboration between the space and
power domains, while in the last one, the time
and frequency domains are also involved.

POINT-TO-POINT
This case includes only one transmitter and one
receiver (i.e., Mt = Mr = 1). In a single-antenna
case, it has been shown above that the capacity
improvement obtained by increasing only the
bandwidth or transmit power is eventually limited due to the law of diminishing returns. By
exploiting the additional signal dimensions
reserved in the space domain, on the other hand,
multiple-antenna technologies may acquire a sig-

16

nificant multiplexing gain. Moreover, the transmit power may be allocated adaptively according
to the channel conditions to make full use of the
signal dimensions, achieving a considerable
power gain. Then both the multiplexing gain and
power gain may together improve the transmission capacity.
Consider a multiple-antenna case where both
the transmitter and receiver are equipped with N
antennas a typical case of multiple-input multiple-output (MIMO). When the channel state
information (CSI) is only available at the receiver side (termed CSIR), the transmit power is
equally allocated across the eigenmodes of the
channel matrix H due to lack of CSI at the transmitter, and thus the transmission capacity, CP2P,
can be expressed as [6]
N rank

CP2P =

i =1

P /N
E log 2 1 + i2 0
,
N 0

where Nrank is the rank of H, i is the ith nonzero singular value of H, and E[] denotes expectation of random variables. At the high SNR
region, if H is sufficiently random (and thus
probably with full rank) and statistically well
conditioned, the capacity is approximately equal
to Nlog 2 [1 + P 0 /(NN 0 )], indicating an N-fold
multiplexing gain over SISO. However, if H is
statistically rank deficient due to the correlation
of the entries, the multiplexing gain decreases
accordingly. At the low SNR region, whether the
entries of H are correlated or not, the capacity is
always approximately equal to NP0/N0log2e, indicating an N-fold power gain over SISO. Anyway,
under CSIR, the capacity of a sufficiently random and well conditioned MIMO channel
increases linearly with N.
If the CSI is also available at the transmitter
side (termed CSIT) via estimation or feedback,

IEEE Wireless Communications August 2011

LU LAYOUT

8/8/11

12:09 PM

Page 17

T1

R1

T2

Transmitters

R2

Receivers

Time
Space

Arrange multiple
transmissions
coordinately

Multiplexing
gain

Allocate transmit
power or schedule
transmissions
adaptively

Capacity gain

Combine multiple
received signals
coherently

Multiplexing
gain

Frequency
Power

Resource domains

Figure 4. A conceptual diagram of MDC. MDC is dedicated to increase multiplexing gain by a coordinate
arrangement of multiple transmissions, and power gain by adaptive power allocation, scheduling, and/or
coherent combination of multiple received signals. Both the multiplexing gain and power gain may translate to a capacity gain.
the transmit power may be allocated in a waterfilling manner across the eigenmodes of H, yielding an additional power gain [6]. At the high
SNR region, this additional power gain produces
a limited capacity improvement, especially when
H is sufficiently random and statistically well
conditioned. At the low SNR region, however,
this additional power gain may yield considerable capacity improvement. Specifically, when
SNR is low enough, the water-filling scheme will
assign all the available power to the strongest
eigenmode corresponding to the largest singular
value, max. By virtue of
lim log 2 (1 + ) = log 2 e,

the transmission capacity approximates


2 P /N log e, where 2
max
0 0
2
max locates in the interval
[N, N2] with high probability [7]. Therefore, with
CSIT, N to N2 times power gain may be obtained
compared to SISO, which may be translated into
the same amount of capacity gain at the low
SNR region. The upper bound N2 is achievable
when the entries of H are fully correlated [8].
Figure 5 illustrates the capacity gains of
MIMO over SISO with N = 8. Both cases the
entries of H are independent identically distributed (labeled iid) or fully correlated
(labeled corr), are considered. At the high
SNR region (Fig. 5a), the transmission capacity
is mainly increased by the multiplexing gain. As
a result, the capacity gain is nearly N for the iid
case and approximately one for the corr case
due to rank deficiency. Meanwhile, CSIT does
little to further increase the capacity. At the low
SNR region (Fig. 5b), the capacity is mainly
increased by the power gain. Specifically, if the
channel coefficients are fully correlated and
known at the transmitter, N2-fold power gain is
achieved, which can be converted into nearly the
same amount of capacity gain.
This case study provides an intuitive interpretation of the term positive interaction among
domains in the concept of MDC. Take the
(asymptotic) N 2 times of capacity gain as an
example. When SNR is very low, the power

IEEE Wireless Communications August 2011

Coherent combination means that the


useful signals are
added in phase, i.e.,
in amplitude, while
the noisy signals are
added in power due
to their independence. Thereby, a
power gain is
obtained.

domain sends this piece of SNR information to


the space domain. One can then increase the
correlation of the transmit and receive antennas
in space domain by, for example, reducing the
spacial distances among the antennas, thereby
creating the condition under which a large capacity gain may be achieved.
Take an example of applying MDC in a practical point-to-point network, such as the downlink transmission in a Third Generation
Partnership Project (3GPP) Universal Mobile
Telecommunications System/Long Term Evolution (UMTS/LTE) network [9], where both the
e-NodeB and the user equipment (UE) can be
equipped with up to four antennas. The e-NodeB
may use either open-loop, such as multiple-codeword-based transmission, or closed-loop spatial
multiplexing technologies, such as precoding, to
achieve times of capacity gain through multiplexing gain at the high SNR region. At the low
SNR region, one of the MDC strategies mentioned above, the antenna correlation alternating
method, is a valid option to get additional power
gain besides existing diversity techniques, such as
space-time coding.

POINT-TO-MULTIPOINT
A point-to-multipoint transmission is composed
of one transmitter and multiple receivers (i.e.,
M t = 1, M r = M > 1). One typical example is
the downlink transmission that one base station
(BS) transmits data to several mobile stations
(MSs). Considering that the radio signal to one
receiver may also be received by other receivers,
one may combine multiple copies of received
signals coherently. Coherent combination means
that the useful signals are added in phase (i.e., in
amplitude), while the noisy signals are added in
power due to their independence. Thereby,
power gain is obtained. The coherent combination may be regarded as collaboration between
the space and power domains, in the sense that
the signal energy dispersed in a broad space is
efficiently collected.
Consider a simple example where the transmitter employs all the available power to trans-

17

LU LAYOUT

8/8/11

10

12:09 PM

Page 18

70

iid(CSIR)
corr(CSIR)
iid(CSIT)
corr(CSIT)

9
8

60

50

Capacity gain

Capacity gain

iid(CSIR)
corr(CSIR)
iid(CSIT)
corr(CSIT)

5
4
3

40
30
20

2
10

1
0

15

25

20

30

-30

-20

-25

SNR (dB)

SNR (dB)

(a)

(b)

-15

Figure 5. Capacity gain results of an 8 8 MIMO channel: a) at the high SNR region, the capacity gain is nearly N, the number of antennas at the transmitter and the receiver, if the channel coefficients are independent identically distributed (iid), and approaching 1 if the
channel coefficients are fully correlated (corr); b) at the low SNR region, the capacity is mainly increased by the power gain. If the
channel coefficients are fully correlated and known at the transmitter, an N2-fold power gain is achieved.
mit to one receiver in one time slot by turns, i.e.,
in a TDM mode. Each of these nodes is assumed
to be equipped with one antenna. The channel
coefficient between the transmitter T 1 and the
receiver R j (j = 1, 2, , M), h j , experiences
Rayleigh fading as stated above. Without coherent combination, the SNR of the received signal
of Rj is P0hj2/N0. However, with the optimal
coherent combination, that is, the maximum
M
ratio combination (MRC), it is increased to j=1
2
P0hj /N0 [10]. Since all the hj (j = 1, 2, , M)
may follow the same distribution, an average
power gain of M times may be achieved for each
link.
At the low SNR region, by virtue of
lim log 2 (1 + ) = log 2 e,

the transmission capacity gain is nearly equal to


the power gain. At the high SNR region, on the
other hand, due to
lim

log 2 (1 + K P )
= 1,
log 2 (1 + )

the capacity gain should be much less than the


power gain.
Numerical results for verification are show in
Fig. 6. In particular, Fig. 6a plots the capacity
gains at typical values of average SNR. An Mfold power gain may be obtained using MRC. At
low SNR region, this power gain is approximately equivalent to the same amount of capacity
gain, while at high SNR region, the capacity gain
is very small. For clearer explanation in Fig. 6b,
at a low SNR, say, 20dB, the capacity gain is
increased linearly as M increases, while at a high
SNR, say, 20dB, the capacity gain gets saturated
quickly.
MDC may be implemented and provide benefits in practical networks with point-to-multipoint topology, such as the downlink of an IEEE
802.16j network [11]. Specifically, the relay stations (RSs) may assist the base station (BS) to

18

transmit signals to users, increasing the system


coverage and yielding additional power gains. If
user cooperation is allowed in the system, users
may perform as relays for each other. Then, by
carrying out some multiplexing scheme (such as
TDD), each user could receive his own signal
with a significant power gain aided by other
cooperative users. Again, note that this power
gain may be translated into the capacity gain linearly at low SNR region.

MULTIPOINT-TO-POINT
A multipoint-to-point transmission is of multiple
transmitters and one receiver, i.e., Mt = M > 1,
M r = 1. One typical example is the uplink of a
cellular system where multiple MSs transmit
their data to one BS. The same idea utilized in
the point-to-multipoint case, i.e., coherent combination, may also be applied here. In particular,
by means of data exchange among the transmitters aforehand, the data to be sent from one
transmitter may be transmitted by multiple
transmitters, with proper delays and phase-shifts,
in order that the multiple transmitted signals are
automatically combined coherently at the receiver. Such a technology is much alike joint beamforming, which serves as an embodiment of
space domain and power domain collaboration
in the sense that the signal energy radiated from
different space locations is focused together.
Accordingly, a power gain may be expected.
Without loss of generality, we assume that
each of the nodes is equipped with one antenna.
The channel coefficient between the transmitter
Ti (i = 1, 2, , M) and the receiver R1 is denoted hi. The maximum average transmit power of
each transmitter is P0/M.
In a traditional scheme, time-division multiple access (TDMA) is used as a benchmark,
where each transmitter occupies only 1/M of the
transmission time. As such, the maximum instantaneous transmit power can be increased up to
P0. Consequently, the instantaneous SNR of the
received signal transmitted from T i is P 0

IEEE Wireless Communications August 2011

LU LAYOUT

8/8/11

12:09 PM

Page 19

M=4
M=8

7
6
Capacity gain

6
Capacity gain

SNR = 20 dB
SNR = 20 dB

5
4

5
4

1
20

1
15

10

10

15

20

Average SNR: P0 /N0 (dB)

(a)

(b)

Figure 6. Capacity gain results for a point-to-multipoint transmission: a) the capacity gain is nearly equal to M, the number of receivers,
at the low SNR region, resulting from an M-fold power gain, while it gradually diminishes to 1 as the SNR increases; b) the capacity
gain increases almost linearly with M at a low SNR value of 20 dB, while at a high SNR of 20 dB, it gets saturated quickly as M
increases.
hi2/N0. When joint collaboration is employed,
all the transmitters exchange the data they want
to transmit first. These data are converted into
one serial stream and transmitted over all M
transmitters simultaneously. In order to realize
coherent combination at R1 and keep the average transmit power at each transmitter no more
than P0/M, the transmitted signal from Ti, xi, is
constructed as
xi = P0 / M shi* / hi ,
where s is the transmitted data symbol in the
serial stream with unit power [10]. Correspondingly, the instantaneous SNR of this virtual link
2
becomes P0(M
i=1hi) /(MN0). Note that

2
M
i =1 hi

M 2,

the average power gain with joint collaboration


may be equal to or less than M.
Analogous to the conclusion for the case of
point-to-multipoint transmission, the capacity
gain is approximate to the power gain under the
low SNR region. As the SNR increases, the
transmission capacity gain diminishes and eventually approaches one with the SNR going to
infinity. The simulation results are similar to
thosee in Fig. 6 and thus are omitted here.
Suppose an application of MDC in the uplink
of an IEEE 802.16j network [11]. When a user
sends data to the BS, other users may perform
as relays and assist its transmission, providing
additional power gain by implementing joint
transmission. This power gain may get the same
amount of capacity gain with low SNR.

MULTIPOINT-TO-MULTIPOINT
Multipoint-to-multipoint transmission may be
viewed as a combination of the above three
topologies. One typical example is a multicell
system where multiple BSs communicate with
multiple MSs. Consider a multipoint-to-multi-

IEEE Wireless Communications August 2011

point network with M transmitter/receiver pairs


(i.e., Mt = Mr = M > 1). Transmitter Ti(i = 1, 2,
, M) tries to send data to receiver R i , which
may be interfered with by Tj(j i).
Due to the coexistence of multiple transmitter/receiver pairs, one of the most challenging
tasks is to mitigate potential interlink interference. To address this issue, traditional methods
based on orthogonal allocation of radio
resources, such as time division and frequency
division, have been proposed. By applying these
simple and effective methods, the whole network
is decomposed into a set of point-to-point links,
while suffering from inefficiency in terms of
transmission capacity. On the other hand,
advanced interference management technologies
may make more aggressive reuse of radio
resources by collaboration among time, frequency, space, and power domains, yielding significant
multiplexing gains and/or power gains, whose
magnitudes depend on the capability of data
exchange among the transmitters or receivers.
First, we consider the case where there is no
data exchange among the transmitters or
receivers. In order to achieve the multiplexing
gain, interference alignment was proposed
recently [12]. The idea is to construct the transmitting signals in such a manner that they
remain distinguishable at their corresponding
receivers while casting overlapping shadows as
interference at other receivers. At each receiver, all the interference from transmitters is
aligned on certain resource blocks (e.g., time
slots, subcarriers, space eigenmodes), saving
other blocks for interference-free transmission.
Therefore, interference alignment may be
implemented in either the time, frequency, or
space domain.
A simple example of interference alignment
in the time domain is presented in [12]. In its
model, the propagation delay is assumed to be
one symbol duration for desired signal paths and
two symbol durations for interfering paths, and
each transmitter only transmits in odd time slots.

19

LU LAYOUT

8/8/11

12:09 PM

Page 20

Capacity gain

Point-to-point
(Mt = Mr = 1, Na = N)

Multiplexing
gain

Power
gain

N~N2

High
SNR

Low
SNR

N~N2

Point-to-multipoint
(Mt = 1, Mr = M, Na = 1)

Multipoint-to-point
(Mt = M, Mr = 1, Na = 1)

M~M2

M/2 or M

M~M2

Multipoint-to-multipoint
(Mt = Mr = M, Na = 1)

M/2 or M

Table 1. Potential performance gains of MDC with typical network topologies


(Mt: number of transmitters, Mr: number of receivers, Na: number of antennas at each node).
For each receiver, the desired signal is received
in the even time slots and all the interference is
received in the odd time slots. Therefore, each
transmitter/receiver pair equivalently utilizes the
whole spectrum half the time. Compared to the
traditional time-division scheme, an M/2-fold
multiplexing gain is achieved in theory.
Now let us consider the scenario where the
transmitters can exchange data with one another. Each transmitter may be viewed as a transmit
antenna of a virtual giant transmitter, and the
data for each receiver can be jointly precoded
before transmission. From the perspective of
information theory, the network turns into a
broadcast channel whose spatial degrees of freedom is M times of an interference channel [13].
Therefore, an M-fold multiplexing gain and certain power gain could be obtained comparing
with orthogonal strategy by using well-designed
precoding algorithm.
Furthermore, if both the transmitters and
receivers can exchange data with one another,
the network can be viewed as a M M (distributed) MIMO channel. This way, the technologies and conclusions presented in the
point-to-point case are applicable here, meaning
that an M-fold multiplexing gain or a power gain
from M to M2 may be obtained.
Applications of MDC in practical multipoint-tomultipoint communication networks may be found
in the study of future wireless systems. In LTEAdvanced, Inter-Node B Coordination has been
proposed to achieve collaboration among eNodeBs [14], where adjacent e-NodeBs are connected by fibers to exchange data with one another.
When multiple e-NodeBs are communicating with
multiple users, both uplink and downlink multiplexing gain may be achieved by employing
advanced joint signal processing of uplink and
downlink signals. By introducing user cooperation,
an additional power gain may be achieved and a
higher capacity gain can be expected.

SUMMARY OF IMPORTANT RESULTS


The potential multiplexing gains and/or
power gains of MDC with four typical topologies are summarized in Table 1. It is noted

20

that the power gain, compared to the multiplexing gain, is expected to be obtained more
easily or in a larger amount. However, only
at the low SNR region may the power gain
be translated into nearly the same amount of
capacity gain.

FUTURE PERSPECTIVE OF MDC


It has been shown that MDC will be an important technical paradigm to advance transmission
capacities of wireless networks. However, the
exciting capacity gains evaluated above are based
on some conditions which may not be supported
in current systems. For example, in the point-tomultipoint case, maximum radio combination is
implemented on the basis of data exchange
among the receivers; in the multipoint-to-point
and multipoint-to-multipoint cases, the implementation of joint beamforming or network
MIMO has to rely on data exchange among the
transmitters. Such functions are not available in
current cellular systems or wireless local area
networks (WLANs). Moreover, technical
progress in synchronization, demodulation, and
so on are necessary to realize the promised
transmission capacity gains at the low SNR
region in practice. Although relevant research
has been launched from both the theoretical [15]
and standardization perspectives [14], a great
deal of future work is desirable to address these
problems.

CONCLUSIONS
This article introduces the concept of multidomain collaboration and reveals its rationale for
improving transmission capacity of wireless networks from the perspective of both multiplexing
and power gains. The capacity gains achieved
by MDC are evaluated for the cases of pointto-point, point-to-multipoint, multipoint-topoint,
and
multipoint-to-multipoint
transmission. Specifically, the power gain of an
N N MIMO point-to-point link may increase
up to N 2 with fully correlated channel coefficients using several collaboration techniques,
thereby increasing the transmission capacity
considerably at the low SNR region. Moreover,
for a network with M transmitters and M
receivers, a multiplexing gain of M/2 or M or a
power gain with upper bound of M 2 may be
achieved, depending on the capability of data
exchange, both of which may lead to a large
capacity gain. It turns out that the capacity gain
is probably larger at the low SNR region than
that at the high SNR region. This raises some
technical challenges such as synchronization,
demodulation, equalization, and other signal
processing technologies at the extremely low
SNR region. Possible technical solutions to
these challenges require comprehensive study
in the future.

ACKNOWLEDGMENT
The authors would like to thank Mr. Danhua
Zhang, Ms. Hui Deng, Mr. Weiliang Zeng, and
Dr. Mai Xu for their helpful discussion, and the
anonymous reviewers for their constructive comments that helped to improve the presentation
of this article.

IEEE Wireless Communications August 2011

LU LAYOUT

8/8/11

12:09 PM

Page 21

REFERENCES
[1] K. E. Case and R. C. Fair, Principles of Economics, 5th
ed., Prentice-Hall, 1999.
[2] G. Foschini, Layered Space-Time Architecture for Wireless Communication in a Fading Environment When
Using Multi-Element Antennas, Bell Labs Tech. J., vol.
1, no. 2, 1996, pp. 4159.
[3] W. Rhee and J. Cioffi, Increase in Capacity of Multiuser
OFDM System Using Dynamic Subchannel Allocation,
Proc. VTC 2000-Spring, Tokyo, Japan, vol. 2, 2002, pp.
108589.
[4] J. Cai, X. Shen, and J. Mark, Downlink Resource Management for Packet Transmission in OFDM Wireless
Communication Systems, IEEE Trans. Wireless Commun., vol. 4, no. 4, July 2005, pp. 16881703.
[5] A. Goldsmith et al., Beyond Shannon: The Quest for
Fundamental Performance Limits of Wireless Ad Hoc
Networks, IEEE Commun. Mag., Jan. 2011.
[6] D. Tse and P. Viswanath, Fundamentals of Wireless
Communication, Cambridge Univ. Press, 2005.
[7] A. Zanella, M. Chiani, and M. Win, Performance of
MIMO MRC in correlated Rayleigh Fading Environments, Proc. VTC-Spring 2005, vol. 3, Stockholm, Sweden, May 2005, pp. 163337.
[8] X. Wu and R. Srikant, MIMO Channels in the Low-SNR
Regime: Communication Rate, Error Exponent, and Signal Peakiness, IEEE Trans. Info. Theory, vol. 53, no. 4,
Apr. 2007, pp. 12901309.
[9] 3GPP R1-050766, Multiple Antenna Solutions for EUTRA, Aug. 2005.
[10] A. Goldsmith, Wireless Communications, Cambridge
Univ. Press, 2005.
[11] W. Ni, G. Shen, and S. Jin, Cooperative Relay
Approaches in IEEE 802.16j, IEEE C802.16j-07/258r1,
IEEE 802.16 Broadband Wireless Access Working
Group, Apr. 2007.
[12] V. Cadambe and S. Jafar, Interference Alignment and
Degrees of Freedom of the K-User Interference Channel, IEEE Trans. Info. Theory, vol. 54, no. 8, Aug.
2008, pp. 342541.
[13] J. Zhang et al., Networked MIMO with Clustered Linear Precoding, IEEE Trans. Wireless Commun., vol. 8,
no. 4, Apr. 2009, pp. 191021.
[14] M. Sawahashi et al., Coordinated Multipoint Transmission/Reception Techniques for LTE-Advanced, IEEE
Wireless Commun., vol. 17, no. 3, June 2010, pp.
2634.

IEEE Wireless Communications August 2011

[15] P. Cuff, H. Permuter, and T. Cover, Coordination


Capacity, IEEE Trans. Info. Theory, vol. 56, no. 9, Sept.
2010, pp. 4181206.

BIOGRAPHIES
J I A N H U A L U [SM08] (lujh@wmc.ee.tsinghua.edu.cn)
received B.S.E.E. and M.S.E.E. degrees from Tsinghua University, Beijing, China, in 1986 and 1989, respectively, and
his Ph.D. degree in electrical and electronic engineering
from Hong Kong University of Science and Technology,
Kowloon. Since 1989 he has been with the Department of
Electronic Engineering, Tsinghua University, where he is
currently a professor. His research interests include broadband wireless communication, multimedia signal processing, satellite communication, and wireless networking. He
has published more than 180 technical papers in international journals and conference proceedings. He has been
an active member of professional societies. He was one of
the recipients of best paper awards at the International
Conference on Communications, Circuits and Systems
2002 and ChinaCom 2006, and was awarded the National
Distinguished Young Scholar Fund by the NSF committee
of China in 2006. He has served in numerous IEEE conferences as a member of Technical Program Committees and
served as Lead Chair of the General Symposium of IEEE
ICC 2008, as well as a Program Committee Co-Chair of the
9th IEEE International Conference on Cognitive Informatics. He is currently a chief scientist of the National Basic
Research Program, China. He is a Senior Member of the
IEEE Signal Processing Society.

This will raise some


technical challenges
such as synchronization, demodulation,
equalization and
other signal processing technologies at
extremely low SNR
region. Possible
technical solutions to
these challenges
require a comprehensive study in the
future.

Y IZHEN J IA [S08, M11] (jiayz@wmc.ee.tsinghua.edu.cn)


received his B.E. degree from the Department of Electronic
Engineering, Tsinghua University, Beijing, China, in 2005,
where he is currently working toward a Ph.D. degree. His
research interests include cooperative communications,
wireless resource management, and satellite communications.
TENGFEI XING [S08, M11] (xingtf@wmc.ee.tsinghua.edu.cn)
received his B.E. degree from the Department of Electronic
Engineering, Tsinghua University, Beijing, China, in 2005,
where he is currently working toward a Ph.D. degree. His
research interests include cooperative communications,
delay tolerant networks, and space-time coding.

21

You might also like