Академический Документы
Профессиональный Документы
Культура Документы
i
(1)
where U
i
[ C
N
R
N
R
and V
i
[ C
N
T
N
T
are unitary matrices
and L
i
[ C
N
R
N
T
is a rectangular matrix whose diagonal
elements are non-negative real numbers.
The diagonal elements l
1
l
2
l
n
min
are the
ordered singular values of the matrix H
i
, where
n
min
= min(N
R
, N
T
).
Fig. 2 shows how the MIMOOFDM channel
decomposes into independent sub-channels with SVD and
pre-processing and post-processing matrices. These sub-
channels are characterised by the channel gains, which are
the singular values of the MIMO channel matrix on each
sub-carrier.
By multiplexing independent data onto these independent
channels, the capacity of ith sub-carrier can be obtained by
C
i
=
n
min
i=1
log
2
1 +
p
ij
l
2
ij
N
0
_ _
bit/s/Hz (2)
where N
0
is the additive white Gaussian noise (AWGN)
variance and p
ij
denotes the transmit power at the jth
antennas of the ith sub-carrier. In an adaptive structure, to
achieve maximum capacity, power of the transmitted
symbol on each of the parallel sub-channels is allocated
optimally.
2.1 Interference limitation on PUs bands
As the consequence of having two coexisting networks, the
signals from secondary base station, which are intended for
its own serviced users, might interfere the reception at the
PUs receivers.
Assume that the transmit signal on the jth antenna of the
ith sub-carrier is a rectangular non-return-to-zero signal. The
power spectral density of this signal can be modelled as [23]
F
SS
( f ) = P
ij
sin p f T
S
pf T
S
_ _
2
(3)
where P
ij
is the total transmit power on the jth antenna of the
ith sub-carrier and T
S
is the symbol duration. The resulting
interference power spilling into the PU band is given by
I
ij
(d
i
, P
ij
) = |g
l
ij
|
2
P
ij
T
S
_
d
il
+B
1
/2
d
il
B
1
/2
sin pf T
S
pf T
S
_ _
2
df (4)
where g
l
ij
denotes the channel gain from the CR base station
to the lth PU sub-carrier for the jth antenna of the ith sub-
carrier. d
il
represents the frequency distance between the
ith sub-carrier of the CR user band and the lth PUs sub-
carrier and B
l
represents each sub-carriers bandwidth.
Figure 1 Coexistence of primary and secondary networks in
CR
Figure 2 Converting the MIMO channel into parallel
channels by SVD
364 IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-com.2010.0262
www.ietdl.org
In addition, the coexistence of PUs and SUs may cause
interference induced by the signals from primary base
station, which are destined to PUs, onto SUs frequency
bands. Nevertheless, we do not consider PUs transmitters
in our analysis as we are concerned about the power
loading at CR transmitters and interference from CR
transmitters to PUs receivers only.
It is obvious from (4) that caused interference on PUs
bands depends on power of CR sub-carriers, in addition to,
distances between CR sub-carriers and primary bands.
A considerable amount of work has been concentrated on
power allocation for downlink MIMOOFDM-based
systems, previously. However, in CR networks, because of
mutual interferences between CR and primary networks,
these proposed power allocation methods are not suitable.
Thus, it is necessary that new and effective power allocation
algorithms be studied in order to maximise the capacity of
CR networks in which not only channel state condition but
also total interference limitation is considered.
3 Optimal power allocation
As we discussed in the previous section, our purpose is to
introduce a new power allocation algorithm that maximises
the total capacity of CR systems provided caused
interferences into the PUs bands do not exceed from a
certain level. This problem can be dened as an
optimisation problem as follows
max C =
N
i=1
n
min
j=1
log
2
1 +
p
ij
l
2
ij
N
0
_ _
(5)
subject to
L
l =1
N
i=1
n
min
j=1
I
l
ij
(d
i,l
, p
i, j
) I th
p
i,j
0
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
In (5), C denotes the total capacity of a CR user and
Ith indicates the maximum tolerable interference by
sub-carriers of PUs introduced by SUs transmission. The
number of these sub-carriers is L and the interference
is obtained by (4). Other variables have the same denition
as (4).
This problem can be reformulated into a standard
optimisation problem and solved using standard convex
optimisation techniques [24]. Let us introduce the
non-negative Lagrange multipliers c and m
ij
. Using
KarushKuhnTucker (KKT) conditions, (5) can be
written as
1
((N
0
/l
2
ij
) +p
ij
)
+m
ij
c
L
l =1
I
l
ij
p
ij
= 0 (6a)
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
L
l =1
N
j=1
n
min
j=1
I
l
ij
(d
i,1
, p
i,;j
) I th (6b)
c . 0 (6c)
m
ij
0 (6d)
p
ij
0 (6e)
m
ij
p
ij
= 0 (6f )
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
By eliminating m
ij
, the above equation can be written as
ij
((N
0
/l
2
ij
) +p
ij
)
+cp
ij
L
l =1
K
(l )
ij
= 0 (7a)
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
1
((N
0
/l
2
ij
) +p
ij
)
L
l =1
K
(l )
ij
c (7b)
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
L
l =1
N
j=1
n
min
j=1
I
l
ij
(d
i,1
, p
i,;j
) I th (7c)
p
ij
0 (7d)
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
where K
(l )
ij
= I
l
ij
/p
ij
.
The following cases are considered to solve the problem:
Case 1: If
c ,
1
((N
0
/l
2
ij
))
L
l =1
K
(l )
ij
then (7b) implies that p
ij
. 0.
Since
L
l =1
K
(l )
ij
. 0, (7a) leads to
c =
1
((N
0
/l
2
ij
) +p
ij
)
L
l =1
K
(l )
ij
and
p
ij
=
1
l
L
l =1
K
(l )
ij
N
0
l
2
ij
(8)
IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370 365
doi: 10.1049/iet-com.2010.0262 & The Institution of Engineering and Technology 2011
www.ietdl.org
Case 2: If
c
1
((N
0
/l
2
ij
))
L
l =1
K
(l )
ij
and if p
ij
. 0
then
c .
1
((N
0
/l
2
ij
+p
ij
))
L
l =1
K
(l )
ij
(9a)
Also we rewrite another form of (7a) as
p
ij
L
l =1
K
(l )
ij
c
1
((N
0
/l
2
ij
) +p
ij
)
L
l =1
K
(l )
ij
_ _
= 0 (9b)
Thus, (9a) and (9b) imply that
p
ij
= 0
Therefore the considered cases indeed the optimal power
prole as
p
ij
= max 0,
1
c
L
l =1
K
(l )
ij
_ _
N
0
l
2
ij
_ _ _ _
(10a)
L
l =1
N
i=1
n
min
j=1
K
(l )
ij
max 0,
1
c
L
l =1
K
(l )
ij
N
0
l
2
ij
_ _
= I th
(10b)
It is important to mention that, in (10b) we replace
the inequality condition with equality. This is why that the
both sides of (10b) are positive and the maximum of
the left-hand side achieved when the right-hand side is
maximum. Also, mention that the power allocation
algorithm in (10a) is simply a water-lling method with
power cap of each allocation. The water level of the jth
antenna of the ith sub-carrier is (1/c
L
l =1
K
(l )
ij
) in which
the threshold is weighted by the interference term
L
l =1
K
(l )
ij
of this branch.
Hence, by using the above scheme, the optimal power
allocation policy is obtained that maximises the
transmission capacity of CR users while keeping the
interference introduced to the PUs below the specic
threshold.
4 Suboptimal power allocation
The complexity of the optimal power prole that was
obtained in the previous section may be quite high,
especially when the number of the antennas is high, since
several iterations may be required in nding the value of c
from (10b). Therefore it is highly interested to propose the
sub-optimal power loading prole.
To do this and simplify the problem, we rst divide the (5)
into two sections and by solving these separately; a total sub-
optimal solution is obtained. The rst problem determines
the sub-carriers power and by using the result of the rst
part, the power of each antenna is determined at the
second problem. To do this, we should follow two stages.
Stage 1: sub-carriers power allocation: In this stage, according
to interference constrain, the power allocated to each sub-
carrier is determined by solving the following optimisation
problem
max C =
N
i=1
log
2
(1 +( p
i
l
2
i
/N
0
)) (11)
subject to
L
l =1
N
i=1
I
l
i
(d
i,l
, p
i
) I th
and p
i
0 i = 1, 2, . . . , N
In (4), I
l
i
(d
i,l
, p
i
) = |g
l
i
|
2
P
i
T
S
_
d
il
+B/2
d
il
B/2
( sinpfT
S
/pfT
S
)
2
df ,
where P
i
is the total transmit power, which can be assigned to
the ith sub-carrier according to total interference constrain
and d
il
represents the frequency distance between the ith
sub-carrier of CR user band and the lth PUs. l
i
is dened
as Frobenius norm H
i
F
of the channel matrix of the ith
sub-carrier and |g
l
i
| is dened as g
i,l
F
/( min{N
R
, N
T
). In
fact, the total power of a sub-carrier is assumed as the sum
of all MIMO paths on that sub-carrier, which is a
reasonable assumption.
By these assumptions, the optimisation problem of (11)
reduces to the particular case of the optimisation problem
of (5) and so can be solved in the similar manner. Solving
this problem leads to the following algorithm determining
the maximum power that can be allocated to each sub-carrier
P
i
= max 0,
1
c
L
l =1
K
l
i
N
0
l
2
i
_ _
(12)
where
K
l
i
= T
S
_
d
il
+Bl /2
d
il
Bl /2
sin pf T
S
pf T
S
_ _
2
df (13)
and c is the Lagrange multiplier which can be obtained from
L
l =1
N
i=1
K
l
i
max 0,
1
c
L
l =1
K
l
i
N
0
l
2
i
_ _
= I th (14)
Stage 2: antennas power allocation: In this stage, by maximum
power that has been allocated in stage 1 for each sub-carrier,
the power of each MIMO antennas is allocated in the way
that the capacity of each sub-carrier maximised and as a
result, the total capacity maximised. To do this, the
following optimisation problem should be solved for each
366 IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-com.2010.0262
www.ietdl.org
sub-carrier
max C =
n
min
l =1
log
2
1 +
p
ij
l
2
ij
N
0
_ _
(15)
subject to
j
p
ij
= P
i
In [22], it has been shown that the above problem has the
following water-lling solution
p
ij
= max 0, m
N
0
l
2
ij
_ _
(16)
where m is Lagranges indicator and is valued to meet
j
p
ij
= P
i
constraint.
Consequently, by performing these two stages, power of
each antenna of each sub-carrier is determined, which has
two major traits: rstly, the total system capacity is
maximised, and secondly, the interference is kept in a
tolerable range.
In order to draw a comparison between the complexity of
the optimal power allocation algorithm that was introduced
in the last section and that of the sub-optimal algorithm
suggested by the present section, it should be noted that
both of these algorithms use the water-lling structure but
in different shapes. The water-lling algorithm has in
general a complexity of O(N log N), where N is the
number of branches for which the algorithm is iterated so
that the optimal power of any branch can be nally
determined. In the optimal power allocation introduced by
(10a) and (10b), if the number of secondary sub-carriers is
N and the mean number of n
min
is considered as M, the
water-lling algorithm must be iterated for NM branches
by any time of program implementation. Therefore this
algorithm has a complexity of O(NM log(NM)). By the
rst step of the sub-optimal algorithm, a complexity of
O(N log N) exists because in this step only the power of
sub-carriers is determined. By the second step, water-lling
algorithm is implemented on each sub-carrier in order that
the power of any antenna can be determined. So, this step
has a complexity of NO(M log M). Through considering
the typical values for the number of CR sub-carriers and
the number of antennas of a system, it is clear that the sub-
optimal algorithm has more complexity than the optimal
algorithm, especially when the number of transmitter
antennas has a large value.
5 Traditional power loading
schemes in a MIMO-OFDM structure
In this section, two most popular power allocation algorithms
in conventional MIMO-OFDM systems are considered. In
both of these algorithms, the goal is maximising the total
capacity of the system according to total power constrain in
transmitters. In order to make a fair performance
comparison among these algorithms with the new proposed
algorithms, all schemes should maintain a given
interference threshold. Then it would be interesting to
observe which scheme offers a higher transmission rate for
the CR user. Therefore at rst we determine the maximum
power which can be transmitted for a given interference
threshold. The maximum power is derived as follows
P
i
=
I th
N
i=1
L
l =1
K
(l )
i
P
total
=
N
i=1
P
i
(17)
By determining the total power of the transmitter, power
allocation is done by the following methods:
5.1 Water-lling algorithm
This method maximises the total capacity of a CR user under
the total power constrain. The method is again based on
using SVD. Using SVD on each OFDM sub-carrier, the
MIMO-OFDM channel is decomposed into parallel,
independent sub-channels. Therefore in order to obtain the
optimal power prole, the following optimisation problem
should be solved
max C =
N
i=1
n
min
j=1
log
2
1 +
p
ij
l
2
ij
N
0
_ _
subject to
(18)
N
i=1
n
min
j=1
p
i, j
P
total
p
i,j
0
i = 1, 2, . . . , N and j = 1, 2, . . . , n
min
It has been shown that the above problem has the following
water-lling solution [22]
p
i, j
= max 0, m
N
0
l
2
ij
_ _
(19)
where m is Lagranges indicator and can be computed as
N
i=1
n
min
j=1
max 0, m
N
0
l
2
ij
_ _
= P
total
(20)
5.2 Uniform water-lling algorithm
This method is a sub-optimal solution for the problem (18).
In this method, the total transmits power, which is obtained
in (17), is allocated equally among all the sub-carriers. In the
IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370 367
doi: 10.1049/iet-com.2010.0262 & The Institution of Engineering and Technology 2011
www.ietdl.org
next stage, power of each MIMO antenna is allocated in the
way that the capacities of each sub-carrier maximise. In order
to achieve this task, the water-lling algorithm that was
introduced in stage 2 in the previous section is used and
therefore power of each antenna is determined.
6 Simulation results
In this section, the proposed power allocation methods are
evaluated in CR networks by computer simulations,
comparing with previous methods.
The simulations are performed for a MIMO-OFDM-
based CR network under the scenario given in Fig. 1.
Total spectrum band is divided into sub-channels equal to
1 MHz frequency bands and each of these sub-channels is
allocated to an OFDM sub-carrier. It is assumed that the
frequency bands, occupied by the PUs, are known in the
CR base station and active sub-carriers have been
determined similar to Fig. 3. Both the CR base station and
the CR user can have different number of antennas.
Consider that there are 32 sub-carriers; that 16 of them are
used by PUs and the rest of them are used by the CR network
in three equal bands as shown in Fig. 3.We use values for T
S
and B
l
of 1 ms and 1 MHz, respectively. AWGN of variance
10
23
is assumed and channel gains h and g are assumed to be
Rayleigh fading with an average power gain of 1 dB. Each
simulation was run 1000 times and shown that values are
the average of these results.
Fig. 4 displays the total capacity of a CR system against
caused interference on PUs for the optimal, sub-optimal
and water-lling methods. In the MIMOOFDM
structure, four transmitter antennas in the base station and
four antennas in the user receiver (4 4) are considered.
As we see in this gure, the optimal power allocation
scheme outperforms other algorithms. In other words, in
this algorithm with a specic capacity, the interference
introduced to PUs bands is lower. The proposed sub-
optimal methods performance is close to the optimal
algorithm and is clearly better than that achieved by the
classical water-lling power loading. It is due to the fact
that the sub-optimal algorithm is directly derived from the
optimality conditions but the classical method did not
consider interference constraint.
Fig. 5, compares the sum capacities of these algorithms
when there are two transmitter antennas in the base station
and two antennas in the user receiver (2 2). Comparing
the results of Figs. 4 and 5 shows that the total capacity of
the system increases by increasing the number of antennas
in all cases. Furthermore, as we expect, the performance of
our sub-optimal algorithm approaches to the optimal
algorithms performance by decreasing the number of
antennas.
It should be noted that for the simulations of the optimal
power allocation algorithm, the number of iterations, which
are necessary for any symbol are dependent on the channel
conditions, the number of sub-carriers, the number of
transceivers antennas and the threshold level of the
tolerable interference onto PUs bands. For example, in the
scenario of Fig. 4, for the threshold level of 0.001,
the number of iterations required for any symbol is in
average about 30 times and for the threshold level of 0.01
it is about 15 times.
Figure 3 Distribution of available spectrum to PUs and SUs
Figure 4 Maximum capacity of a CR user against interference introduced to the PUs band for a 4 4 CR system
368 IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-com.2010.0262
www.ietdl.org
Fig. 6 shows the transmit power of the CR user against the
interference introduced to the PU band for different
algorithms for a 4 4 CR system. This gure
demonstrates that the optimal scheme allows transmission
of higher power than the other schemes for a given
interference threshold. This is because channel gains
between the transmit antennas and PUs are varying and the
allocation schemes proposed in this paper assign more
power to those antennas which produces less instantaneous
interference to PUs band.
7 Conclusion
In this paper, we investigated power loading scheme for CR
networks, employing the MIMOOFDM structure. It has
been shown analytically that the conventional power
loading algorithm cannot be used in CR networks. Then,
we proposed an optimal power allocation policy and new
algorithm that maximise the transmission capacity of CR
users; meanwhile the interference introduced to the PUs
remained below the specic threshold. Furthermore, we
proposed a sub-optimal power loading prole, which has
less complexity for practical implementation. The
simulation results are evaluated for the performance of the
new power loading algorithms and then are compared with
previous ones. It has been shown that the new algorithms
are more efcient and suitable to apply in CR networks.
8 References
[1] HAYKIN S.: Cognitive radio: brain-empowered wireless
communications, IEEE J. Sel. Areas Commun., 2005, 23,
pp. 201220
Figure 5 Maximum capacity of a CR user against interference introduced to the PUs band for a 2 2 CR system
Figure 6 CR transmit power against interference introduced to the PUs band for a 4 4 CR system
IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370 369
doi: 10.1049/iet-com.2010.0262 & The Institution of Engineering and Technology 2011
www.ietdl.org
[2] AKYILDIZ F., LEE W.Y., VURAN M.C., MOHANTY S.: Next
generation/dynamic spectrum access/cognitive radio
wireless networks: a survey, Comput. Netw., 2006, 50,
pp. 21272159
[3] ETKIN R., PAREKH A., TSE D.: Spectrum sharing for
unlicenced bands, IEEE J. Sel. Areas Commun., 2007, 25,
(3), pp. 517528
[4] MENON R., BUEHRER R.M., REED J.H.: Outage probability
based comparison of underlay and overlay spectrum
sharing techniques. IEEE Int. Conf., DYSPAN, November
2005, pp. 101109
[5] BERTHOLD U., JONDRAL F.K.: Guidelines for designing OFDM
overlay systems. IEEE Int. Conf., DYSPAN, November 2005,
pp. 626629
[6] WEISS T.A., JONDRAL F.K.: Spectrum pooling: an
innovative strategy for the enhancement of
spectrum efciency, IEEE Commun. Mag., 2004, 43,
pp. S814
[7] NEE R.V., PRASAD R.: OFDM for wireless multimedia
communication (Artech House, 2000)
[8] BERTHOLD U., JONDRAL F., BRANDES S., SCHNELL M.: OFDM-
based overlay systems: a promising approach for
enhancing spectral efciency [Topics in radio
communications], IEEE Commun. Mag., 2007, 45, (12),
pp. 5258
[9] FARHANG-BOROUJENY B., KEMPTER R.: Multicarrier
communication techniques for spectrum sensing and
communication in cognitive radios, IEEE Commun. Mag.,
2008, 46, pp. 8085
[10] PAULRAJ A.J., GORE D.A., NABAR R.U., BOLCSKEI H.: An overview
of MIMO communications a key to gigabit wireless, Proc.
IEEE, 2004, 92, (2), pp. 198218
[11] GESBERT D., SHAFI M., SHIU D., SMITH P.J., NAGUIB A.: From
theory to practice: an overview of MIMO spacetime
coded wireless systems, IEEE J. Sel. Areas Commun.,
2003, 21, (3), pp. 281302
[12] BOLCSKEI H., ZURICH E.: MIMO-OFDM wireless systems:
basics, perspectives, and challenges, IEEE Wirel.
Commun., 2006, 13, pp. 3137
[13] ZHANG W., XIA X.G., BEN LETAIEF K.: Space-time/frequency
coding for MIMO-OFDM in next generation broadband
wireless systems, IEEE Wirel. Commun., 2007, 14,
pp. 3243
[14] KONG J., LU Z., LI H.: A power and bit allocation algorithm
of changeable BER for MIMO-OFDM system. IEEE Int.
Symp. on Intelligent Signal Processing and
Communication Systems, 2005, pp. 117120
[15] GAO L., LUO Z., TANG B., LIU Y., GAO J.: A low-complexity
adaptive bit and power allocation algorithm for MIMO-
OFDM systems. IEEE Conf., ISCIT, 2005, pp. 542545
[16] PANDHARIPANDE A.: Adaptive modulation for MIMO-
OFDM systems. IEEE Conf., VTC, 2004, pp. 12661270
[17] WANG P., ZHAO M., XIAO L., ZHOU S., WANG J.: Power allocation
in OFDM-based cognitive radio systems. IEEE Global
Telecommun. Conf. (GLOBECOM), December 2007,
pp. 40614065
[18] BANSAL G., HOSSAIN M.J., BHARGAVA V.K.: Optimal and
suboptimal power allocation schemes for OFDM-based
cognitive radio systems, IEEE Trans. Wirel. Commun.,
2008, 7, (11), pp. 47104718
[19] NGO D.T., TELLAMBURA C., NGUYEN H.H.: Resource allocation
for OFDM-based cognitive radio multicast network. IEEE
Conf., WCNC 2009, pp. 16
[20] HASAN Z., HOSSAIN E., DESPINS C., BHARGAVA V.K.: Power
allocation for cognitive radios based on primary user
activity in an OFDM system. IEEE Global Telecommun.
Conf. (GLOBECOM), December 2008, pp. 16
[21] BANSAL G., HOSSAIN MD.J., KALIGINEEDI P., ET AL.: Some
research issues in cognitive radio networks. IEEE Conf.,
AFRCON 2007, pp. 16
[22] TSE D., VISWANATH P.: Fundamental of wireless
communication (Cambridge University Press, 2005)
[23] WEISS T.A., HILLENBRAND J., KROHN A., JONDRAL F.K.: Matual
interference in OFDM-based spectrum pooling systems.
IEEE Conf., VTC 2004, pp. 18731877
[24] BOYD S., VANDENBERGHE L.: Convex optimization
(Cambridge University Press, 2004)
370 IET Commun., 2011, Vol. 5, Iss. 3, pp. 362370
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-com.2010.0262
www.ietdl.org