Power Allocation using Geometric Water Filling for
OFDM-based Cognitive Radio Networks
Ajmery Sultana, Lian Zhao, SMIEEE and Xavier Fernando, SMIEEE
Department of Electrical and Computer Engineering, Ryerson University, Toronto, Canada
Abstract—Cognitive radio (CR) is a promising wireless
paradigm that provides efficient spectral usage. Orthogonal
frequency division multiplexing (OFDM) is a potential technology
providing many advanced functionalities in terms of power
and rate control for cognitive radio networks (CRNs). Power
allocation for CRNs is a crucial task for better interference
management. In this paper, a subcarrier assignment scheme
and a novel power allocation algorithm using geometric water
filling is presented for OFDM based CRNs. This algorithm is
optimized such a way to maximize the sum rate of secondary
users by allocating power more efficiently, while constraining the
1) total transmit power, 2) individual subchannel transmit power
as well as 3) individual subcarrier peak power of secondary users,
for a given interference level to the primary users. Numerical
results show that this algorithm provides better utilization of
power resources thus maximizes the sum rate than the existing
algorithms.
Key words: Cognitive radio, power allocation, OFDM, geo-
metric water filling.
I. INTRODUCTION
Spectrum scarcity is becoming a crucial challenge with ever
increasing demand for services in the wireless communication
paradigm. While there is constant push for new spectrum,
Federal Communication Commissions (FCC) observes that
current licensed spectrum is significantly underutilized [1].
Cognitive radio (CR), first coined by Joseph Mitola [2],
has been proposed as a solution for efficiently utilizing radio
resources of the exsiting spectrum. CRs, with its ability
to smartly interact with the surrounding environment, are
amenable to allow the coexistence licensed (primary) users
(PU) and unlicensed (secondary) users (SU) sharing the same
bandwidth opportunistically without causing harmful interfer-
ence to each other.
However, appropriate power, modulation and data rate man-
agement impose unique challenges in opportunistic spectrum
access in cognitive radio networks (CRNs). Popular orthogonal
frequency division multiplexing (OFDM) better manages these
issues and is a potential technology for CRNs too [3]–[5]. The
subcarrier configurability of OFDM allows SUs to flexibly fill
the spectral holes left by PUs without causing unacceptable
interference. Also, the fast Fourier transform (FFT) component
of OFDM systems provides an additional tool for opportunistic
spectrum detection.
Power allocation problem in OFDM based CRNs has been
widely studied under different settings. A grouped power
allocation scheme is presented in [6] where at the first stage,
power is assigned to some groups considering interference
constraints with the PU band. At the second stage, the re-
maining power is allocated to others with water-filling (WF)
algorithm. A low-complexity algorithm using incrementing
or decrementing power is proposed in [7]. This algorithm is
978-1-5090-1701-0/16/$31.00 ©2016 IEEE
based on constraining the total and transmit powers of each
subchannel. Another power allocation algorithm with partial
channel state information is proposed in [8] where the WF
process is run only once, and then, the final power allocation
vector is directly calculated to maximize the SUs throughput.
In order to maximize the downlink capacity of CRNs, a
suboptimal power allocation algorithm based on linear water-
filling (LWF) scheme is proposed in [9] that considers inter-
ference and transmit power constraints. Another suboptimal
power allocation scheme is proposed in [10] that reduces the
computational complexity of determining the water level to
allocate the power for each subcarrier under a peak power
constraint. For weighted sum rate maximization in CRNs, a
power limited multilevel WF algorithm has been proposed
in [11] where subcarrier power constraints were added to the
traditional problem. With the total power constraint and the
power constraint on each subchannel, an iterative partitioned
water-filling (IPWF) algorithm was proposed in [12] to realize
optimal power allocation in OFDM based CRNs. However,
in the above studies, they did not consider individual peak
power constraints for the subcarriers. To the best of author
knowledge, none of the papers in the literature yet considers
individual peak power constraints for subcarriers together with
total power constraint and the power constraint on each sub
channel.
The main contributions of this paper are: the power alloca-
tion problem in OFDM-based CRNs is modeled considering
both individual subcarrier peak power constraints and sub-
channel power constraints with total power constraint. In this
work, a more efficient algorithm, named as iterative partitioned
weighted geometric water-filling with individual peak power
constraints is proposed along with subcarrier assignment. The
proposed algorithm is shown to maximize the sum rate by pro-
viding better utilization of power resources than the existing
algorithms. Also power allocation is performed considering
the weight factor of each channel.
The organization of the rest of this paper is as follows:
section II provides the idea behind the geometric water-
filling approach. Section III describes the system model and
formulates the problem along with the proposed solutions.
Simulation results are depicted in section IV and finally section
V concludes the paper.
II. G EOMETRIC WATER - FILLING M ETHOD
The conventional water-filling (CWF) [13] problem can
be described as follows: given 𝑃𝑇 > 0, is the total signal
power (or volume of the water); the allocated power and the
propagation path gain for the 𝑖𝑡ℎ channel are given as 𝑃𝑖 and
ℎ𝑖 respectively where, 𝑖 = 1...𝑁 ; and 𝑁 is the total number
of subcarriers. Also, if the weighted coefficients 𝑤𝑖 > 0, ∀𝑖,
then without loss of generality, {ℎ𝑖 .𝑤𝑖 }𝑁
𝑖=1 will be positive
and monotonically decreasing, then the optimization problem
can be written as:
𝑁
∑ A. Cognitive Radio System and Transmission Power Con-
max{𝑃𝑖 }𝑁
𝑖=1
𝑤𝑖 log(1 + ℎ𝑖 𝑃𝑖 )
𝑖=1
subject to: 𝐶1 : 0 ≤ 𝑃𝑖 , ∀𝑖;
𝑁
∑ harmful interference to each other, the SU needs to detect
𝐶2 : 𝑃𝑖 = 𝑃𝑇
𝑖=1
where, constraint C1 is the allocated power to be nonnegative
and constraint C2 is the total power constraint. To find the
solution to problem (1), we usually start from the Karush-
Kuhn-Tucker (KKT) conditions of the problem, as a group
of the optimality conditions. The water level (𝜇) needs to
be
∑chosen
𝑁
to satisfy the power sum constraints with equality
( 𝑖=1 𝑃𝑖 = 𝑃𝑇 ) to find the optimal solution.
In [14] a geometric water filling (GWF) approach is pro-
posed to solve the CWF problem and its weighted form. It has
two advantages, 1) the geometric approach can compute the
exact solution to the CWF, including the weighted case, with
less computation without determining the water level through
solving the non-linear system, 2) machinery of the proposed
geometric approach can overcome the limitations of the CWF
algorithm to include more stringent constraints.
Instead of trying to determine the water level 𝜇, which is a
real nonnegative number, the highest water level step, denoted
(1) is thus referred to as weighted geometric water-filling with
sum and individual peak power constraints (GWFPP) [14].
III. S YSTEM M ODEL AND P ROBLEM F ORMULATION
(1) straint
A typical cognitive radio systems is shown in Fig. 1 where
PUs and SUs share the same bandwidth. In order to avoid
the opportunities when PUs are not utilizing the spectrum.
Higher detection probability without errors provide successful
exploitation of opportunities for transmission. In Fig. 1, a disk
propagation model is considered to illustrate SU transmission.
SU can detect any PU’s activity within the detection region.
However, those PUs that fall outside the detection region (like
PU2 in Fig. 1), are undetectable by the SU. To deal with
this situation, as in [15], PU2 defines a protection region with
radius 𝑃 and needs to maintain a certain interference level
𝜂 within this area. In this scenario, SU’s transmission power
𝑃𝑡𝑥 , subjected to interference constraint, can be written as:
𝛽
𝑃𝑡𝑥 ≤ 𝜂(𝐷 − 𝑃 ) (6)
where D is the distance between the SU transmitter and the
nearest undetectable PU transmitter, 𝛽 is the path attenuation
factor and 𝜂 is the maximum allowable interference level.
Interference Region
SU Tx
SU Rx
Detection Region

by 𝑛∗ , is introduced in [14] to find the solutions for power Transmission Region

PU Tx

allocation. Let 𝑃𝑡 (𝑛) denotes the water volume above step 𝑛 PU Rx

or zero, whichever is greater and the value of 𝑃𝑡 (𝑛) can be SU

found by [14]:{ [ 𝑛−1 ( ) ]}+
∑ 1 1
𝑃𝑡 (𝑛) = 𝑃𝑇 − − 𝑤𝑖 PU1

𝑖=1
ℎ𝑛 ℎ𝑖
D

{ [ 𝑖−1 ]}+ (2)

∑
= 𝑃𝑇 − 𝛿𝑛,𝑖 , for 𝑛 = 1, ..., 𝑁 PU2
P
𝑖=1
Protection Region
where ℎ1𝑖 is the “step depth”of the 𝑖𝑡ℎ stair and 𝛿𝑛,𝑖 is the
“step depth difference”of the 𝑛𝑡ℎ and the 𝑖𝑡ℎ stairs. Due to the
definition of 𝑃𝑡 (𝑛) being the power (water volume) above step Fig. 1: Various regions in a cognitive radio system.
𝑛, it cannot be a negative number. Therefore we use {.}+ in (2) B. OFDM-based Cognitive Radio System
to assign 0 to 𝑃𝑡 (𝑛) if the result inside the bracket is negative. OFDM is a potential technology in terms of modulation and
The corresponding geometric meaning is that the 𝑛∗𝑡ℎ level is power control. It is also highly flexible due to reconfigurable
above water. According
{ to [14], the explicit solution to (1) is: subcarrier structure to fit in CRNs for efficient utilization
𝑃𝑛 ∗ 1 1
[𝑤 + ( ℎ𝑛∗ − ℎ𝑖 )]𝑤𝑖 1 ≤ 𝑖 ≤ 𝑛∗ of spectrum opportunities. Fig. 2 depicts a typical spectrum
𝑃𝑖 = 𝑛∗ (3)
0, 𝑛∗ < 𝑖 ≤ 𝑁, structure in OFDM-based CRNs. There are 𝑀 subchannels
licensed to 𝑀 PU systems that can be potentially used by the
where the water level{step 𝑛∗ is given as } SU based on opportunity detection. There are 𝑁 subcarriers
∗
𝑛 = max 𝑛∣𝑃𝑡 (𝑛) > 0, 1 ≤ 𝑛 ≤ 𝑁 (4) that are distributed among the 𝑀 subchannels. For example,
let the 𝑗 𝑡ℎ subchannel has total 𝐿𝑗 subcarriers that can
be utilized by the SU when PU is absent. For successful
and the power level for this step is transmission, the SU first needs to test any PU transmitter
𝑤𝑛∗
𝑃𝑛∗ = ∑𝑛∗ 𝑃𝑡 (𝑛∗ ) (5) in the desired subchannel. If found, the sum power of all the
𝑖=1 𝑤 𝑖 subcarriers in that subchannel will be set to zero until the PU
When the constraint C1 in (1) is extended to 0 ≤ 𝑃𝑖 ≤ 𝑆𝑖 , i.e., transmission ends. If not, the SU can utilize this subchannel
additional individual peak power constraints, then the problem with the interference constraint described in (6). Let 𝐺𝑗 is the
 Subchannel 1 Subchannel 2 Subchannel j Subchannel M
its. Considering constraints C2 and C3, in [12] the authors
proposed Iterative Partitioned Water-Filling (IPWF) algorithm
……... ……... to obtain the optimal power allocation vector. Constraint C1
was introduced in [14] where the authors proposed GWFPP
1 2 3…. ……...mj …………..N approach to solve the weighted radio resource allocation
problems, that also provides optimal result. In this paper, we
Fig. 2: Spectrum of a SU in OFDM-based CRNs combined all those constraints to construct an optimal power
interference constraint on the 𝑗 𝑡ℎ subchannel after spectrum allocation scheme.
{
detection, then,
0 𝑃 𝑈𝑗 is detected D. Subcarrier Allocation
𝐺𝑗 ≜ (7)
𝜂𝑗 (𝐷𝑗 − 𝑃𝑗 )𝛽𝑗 𝑃 𝑈𝑗 is not detected
Assume 𝐼𝑚 be the number of subcarriers allocated for
where, 𝜂𝑗 is the maximum allowable interference level for one subchannel, is a variable. Since the utilization of any
𝑃 𝑈𝑗 within the protection region whose radius is 𝑃𝑗 , 𝐷𝑗 subchannel by the SU, is bounded by the power constraint
is the distance between the SUs transmitter and the nearest on the 𝑗th subchannel (𝐺𝑗 ), the transmit power for one
undetectable 𝑃 𝑈𝑗 ’s transmitter and 𝛽𝑗 is the corresponding subchannel on each subcarrier is given by
𝐺𝑗
path attenuation factor. 𝑃𝑚,𝑖𝑚 = , 𝑖𝑚 𝜖 𝐼ˆ𝑚 (11)
𝐼𝑚
C. Problem Formulation
where 𝐼ˆ𝑚 represents the set of subcarriers allocated to one
Consider an OFDM communication system similar to [16]: subchannel and 𝐼𝑚 is the number of subcarriers allocated to
𝑦𝑛 [𝑚] = ℎ𝑛 𝑥𝑛 [𝑚] + 𝑤𝑛 [𝑚], where 𝑛 = 1, ..., 𝑁 (8) one subchannel. The achieveable data rate (using (9)) can be
where, 𝑥𝑛 [𝑚], 𝑦𝑛 [𝑚], 𝑤𝑛 [𝑚] and ℎ𝑛 are the input, output, shown as
∑𝑀 ∑
noise signal and channel gain, respectively. Assume 𝑃𝑇 > 0, 𝐷= 𝑅𝑚,𝑖𝑚 (𝑃𝑚,𝑖𝑚 ), 𝑖𝑚 𝜖 𝐼ˆ𝑚 (12)
as the total power constraint and 𝑃𝑖 is the transmit power 𝑚=1 𝑖𝑚 𝜖𝐼𝑚
of an OFDM block, then the achievable data rate of reliable
communication using the OFDM channel is, For a given set 𝐼ˆ𝑚 for a subchannel, if one more subcarrier
∑𝑁
𝑃𝑖 ∣ℎ𝑖 ∣
2 𝑖𝑚 ∗ is allocated to that subchannel, i. e. 𝐼ˆ𝑚 is replaced by
𝑅(𝑃𝑖 ) = 𝐵 log(1 +
𝑁0
) (9) 𝐼ˆ𝑚 ∪ {𝑖𝑚 ∗ }, the change of achievable rate for that subchannel,
𝑖=1 Δ𝑑𝑚,𝑖𝑚 ∗ (𝐼ˆ𝑚 ), can be given by
where, 𝑁0 is the noise power spectral density and B is the 𝐺𝑗 𝐺𝑗
Δ𝑑𝑚,𝑖𝑚 ∗ (𝐼ˆ𝑚 ) = 𝐷𝑚 ( ) − 𝐷𝑚 ( )
subcarrier spacing (bandwidth). Thus the power allocation has 𝐼𝑚 + 1 𝐼𝑚
to be done to maximize the sum rate in (9). In OFDM based ∑ 𝐺𝑗 𝐺𝑗 (13)
= 𝑑𝑚,𝑖𝑚 ( ) − 𝑑𝑚,𝑖𝑚 ( )
CRNs, the subchannel transmit power constraints impose 𝐼 𝑚 + 1 𝐼 𝑚
further restrictions on the power allocation in addition to the 𝑖𝑚 𝜖𝐼ˆ ∗
𝑚 ∪{𝑖𝑚 }

total transmit power constraint. Assuming the allocated signal

In our proposed algorithm, assuming 𝐼ˆ𝑚 has been allocated
power and the propagation path gain for the 𝑖𝑡ℎ subcarrier
2 to one subchannel, whether an unallocated subcarrier 𝑖 can
are given as 𝑃𝑖 and ℎ𝑖 (where ∣ℎ𝑁𝑛0∣ = ℎ𝑖 ) respectively, be allocated to one subchannel, depends on whether its
𝑖 = 1, ..., 𝑁 and, the weights 𝑤𝑖 > 0, ∀𝑖 then the optimal data rate increase Δ𝑑𝑚,𝑖𝑚 (𝐼ˆ𝑚 ) is the maximum among all
power allocation problem can be expressed as: the subchannels. That is, subcarrier i will be allocated to
∑𝑁
𝑘 ∗ = max𝑚,𝑖𝑚 Δ𝑑𝑚,𝑖𝑚 (𝐼ˆ𝑚 ). After subcarrier allocation, in the
max{𝑃𝑖 }𝑁
𝑖=1
𝑤𝑖 log(1 + ℎ𝑖 𝑃𝑖 ) (10)
𝑖=1
subject to: C1: 0 ≤ 𝑃𝑖 ≤ 𝑆𝑖 , ∀𝑖; Algorithm 1: Subcarrier Allocation
𝑁
∑ 1 initialization;
C2: 𝑃𝑖 = 𝑃𝑇 2 𝐼ˆ𝑚 = 𝜙, 𝐼𝑚 = 0, Δ𝑑𝑚,𝑖𝑚 (𝐼𝑚 ) = 𝐷𝑚 (𝐺𝑗 )
𝑖=1 3 for i 𝜖 N do
C3: 𝐹𝑗 ≤ 𝐺𝑗 , ∀𝑗; 4 for m 𝜖 M do
where, 𝑁 is the total number of subcarriers and 𝑗 = 5 compute Δ𝑑𝑚,𝑖𝑚 (𝐼ˆ𝑚 ) using (13)
1, ..., 𝑀 ; and 𝑀 is the total number of subchannels, 𝐹𝑗 =
∑𝑚𝑗+1 −1 6 𝑘 ∗ = max𝑚,𝑖𝑚 Δ𝑑𝑚,𝑖𝑚 (𝐼ˆ𝑚 );
𝑖=𝑚𝑗 𝑃𝑖 , 𝑗𝜖𝐴 is the power allocated to the 𝑗 𝑡ℎ subchannel 7 𝐼ˆ𝑚
∗ = 𝐼ˆ ∪ {𝑖};
𝑚
and 𝑚𝑗 is the index of the first subcarrier and 𝑚𝑗+1 − 1 is 8 𝐼𝑚∗ ∗
= 𝐼𝑚 + 1;
the index of the last subcarrier in the 𝑗 𝑡ℎ subchannel. 9 end
The power allocation problem in (10) is constructed con- 10 end
sidering three constraints. Constraint C1 considers individual
peak power constraints, constraint C2 includes total power
constraint and constraint C3 incorporates individual subchan- second step, power allocation is carried out on its allocated
nel power constraints caused by the PUs’ interference lim- subcarrier 𝐴ˆ𝑚 .
E. Power Allocation Algorithm 2: Iterative Partitioned Weighted Geometric Water-
In this paper, utilizing IPWF [12] and GWFPP [14] ap- filling with Individual Peak Power Constraints (IGPP)
proach, we proposed Iterative Partitioned Weighted Geometric Input: vector { ℎ1𝑖 }, {𝑤𝑖 }, {𝑆𝑖 } for 𝑖 = 1, 2, ..., 𝑁 ,
Water-Filling with Individual Peak Power (IGPP) algorithm, the set 𝐸 = 1, 2, ..., 𝑁 and 𝑃𝑇
that also provides optimal result. For IGPP the first step is to 1 initialization;
divide all the subchannels into two sets, say A and B, and 2 𝐴 = {𝑗∣𝑗 = 1, 2, ..., 𝑀 }
there are 2𝑀 partitions in total. 𝐸 set is a subsequence of the 3 𝐵 = 𝜙, 𝐹 = 𝜙, Λ = 𝜙, 𝑃 ∗ = 𝑃𝑇 ;
sequence 1, 2, ..., 𝑁 and 𝑁 is the total number of subcarriers.
4 𝐶 = {𝑖∣the ith subcarrier belongs to the jth subchannel, 𝑗𝜖𝐴};
The next step would be to perform GWFPP for every 𝑗 𝑡ℎ
5 for j 𝜖 A do
subchannel and calculate 𝑃𝑖 using (3), (4) and (5). Let the set
Λ is defined by the set {𝑖∣𝑃𝑖 > 𝑆𝑖 , 𝑖𝜖𝐸}. If Λ is an empty 6 for i 𝜖 C do
set, then 𝑃𝑖 will be the output, otherwise, 𝑃𝑖 will be set by 7 compute {𝑃𝑖 } using GWF
the individual peak power 𝑆𝑖 . We need to update the set 𝐸 8 Λ = {𝑖∣𝑃𝑖 > 𝑆𝑖 , 𝑖𝜖𝐸}
and the total power 𝑃𝑇 at the end of each iteration. Since 9 if Λ = 𝜙 then
the finite set 𝐸 is getting smaller and smaller until the set 10 output: {𝑃𝑖 };
Λ is empty, algorithm GWFPP carries out at most, 𝑁 loops 11 else
to compute the optimal solution [14]. The next steps are to 12 𝑃𝑖 = 𝑆𝑖 , 𝑖𝜖𝐴;
remove partitions that are represented by set 𝐷 which is a 13 end
set of subchannels that satisfies 𝐹𝑗 < 𝐺𝑗 where 𝑗𝜖𝐴. For 14 𝐸 =𝐸∖Λ
each of the remaining partitions in the remainder set A, the 15 end
geometric water-filling is performed on the subcarriers that ∑𝑚𝑗+1 −1
16 𝐹𝑗 = 𝑖=𝑚 𝑃𝑖 , 𝑗𝜖𝐴
belong to the 𝑗 𝑡ℎ subchannel. Algorithm IGPP carries out the 𝑗
17 𝐷 = {𝑗∣𝐹𝑗 ≤ 𝐺𝑗 , 𝑗𝜖𝐴};
loops to compute the optimal solution until the set A is empty.
18 𝐿 = {𝑖∣the ith subcarrier belongs to the jth subchannel, 𝑗𝜖𝐷}
The last step would be to verify each partition whether they
satisfy 𝐹𝑗 < 𝐺𝑗 where 𝑗𝜖𝐵. According to the paper [12], 19 𝐴 = 𝐴 ∖ 𝐷, 𝐵 = 𝐵 ∪ 𝐷
there is only one available partition and the corresponding 20 𝐶 = 𝐶 ∖ 𝐿, 𝐹 = 𝐹 ∪ 𝐿
power allocation vector is the solution. Based on the property 21 if 𝐴 = 𝜙 then
of IGPP and the strict constraints on the objective function, 22 output: {𝑃𝑖 };
there can be only one power allocation vector which satisfies 23 else ∑
all the conditions. In CWF problem, it is dificult to obtain the 24 𝑃 ∗ = 𝑃 ∗ − 𝑖𝜖𝐿 𝑃𝑖 ;
explicit expression of the optimal power allocation vector. A 25 end
derived algorithm obtained from [12] and [14] forming IGPP, 26 end
is described in Algorithm 2 description.
TABLE I: Simulation Parameters
IV. S IMULATION R ESULTS
Parameters Values
The proposed algorithm IGPP along with the existing meth-
No. of SU and PU 1, 4
ods (IPWF, GWFPP), have been implemented using MAT- Number of subchannels 4
LAB. Randomized fading have been used to generate the out- Number of subcarriers 15
comes of each algorithms. Here, we assumed that the distance Protection region (𝑃𝑗 ) 1m
between the SU and the PU is known, thus equation (7) has Path attenuation factor 2
been used to calculate the subchannel power constraints (𝐺𝑗 ). Total power (𝑃𝑇 ) 159
According to the subcarrier assignment strategy, described Bandwidth (B) 15KHz
in previous section, we got subcarriers 1, 2, 3, 4 to be in Subchannel power constraints (𝐺𝑗 ) {69, 44, 29, 17}
subchannel 1, subcarriers 5, 6, 7 to be in subchannel 2,
subcarriers 8, 9, 10, 11 to be in subchannel 3 and the last For each subcarrier, we get a different set of water level
set of subcarriers 12, 13, 14, 15 to be in subchannel 4. The which follows the individual peak power constraint but they
simulation parameters are summarized in Table I. are different from the GWFPP algorithm due to the total
Fig. 3 (a) shows power allocation when IPWF method subchannel power constraint. Fig. 3 (d) compares the sum rates
has been applied. Since IPWF works based on partitions of for the IPWF, GWFPP and IGPP algorithms. IGPP algorithm
subchannels, the water levels are different for each subchannel allows the sum rate to be better than IPW and GWFPP, due
due to the total subchannel power constraint for the partitions. to the better utilisation of the power resources.
Fig. 3 (b) shows power allocation when GWFPP algorithm Fig. 4 shows allocated power with corresponding noise
has been applied. This algorithm mainly consider individual power by IGPP when considering different weight factors for
peak power constraints. As a result, for each channel, we get subcarriers. In 𝑋 axis, cumulative summation of weighted
a different set of water level which follows the individual coefficients are presented, as we moved towards right. In Fig.
peak power constraint. Fig. 3 (c) shows power allocation when 4, the width of each bar varies due to the different weight
IGPP algorithm has been applied. Due to the total subchannel factors of the subcarriers.
power constraint and individual peak power constraints, the Fig. 5 compares the allocated power for the IPWF, GWFPP
simulation result varies from the results of IPWF and GWFPP. and IGPP algorithms. For each subchannel, we sorted the
 Allocated Power by IPWF (a) Allocated Power by GWFPP (b)
allocated less power in the subcarrier that has highest noise

Fading & Power Allocation

60

Fading & Power Allocation

60
Fading
Power
Fading
Power
power and allocated more power in the subcarrier that has
40 40
lowest noise power than the other two algorithms. Thus, the
20
20 IGPP allows better utilization of the available power resources
than IPW and GWFPP.
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Subcarriers
Subcarriers
V. C ONCLUSION
Allocated Power by IGPP (c) (d)
60
In this paper, an optimal power allocation strategy that
Fading & Power Allocation

Fading
40
Power

maximizes the sum rate by providing better utilization of

Sum rate (kbps)

40 30

20
power resources in OFDM based CRNs, is investigated. The
20
10
allocation is done according to total power constraint, indi-
0 0 vidual subchannel power constraint, and individual subcarrier
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 IPWF GWFPP IGPP
Subcarriers peak power constraint. Numerical examples are provided to
illustrate the effectiveness of the algorithm. The proposed al-
Fig. 3: Power Allocation comparison for three schemes: (a)
gorithm allows better utilization of available power resources,
IPWF; (b) GWFPP; (c) IGPP; (d) achieved sum rate for three
thus maximizes the sum rate than IPWF and GWFPP. The
schemes. algorithm allocated more power to wireless channels that have
less fading, saving power resources.
R EFERENCES
[1] Federal Communications Commission, “Spectrum policy task force,” pp.
02 – 135, Nov 2002.
[2] J. Mitola, “Cognitive radio: an integrated agent architecture for software
defined radio,” PhD thesis, KTH Royal Institute of Technology, Stock-
holm, Sweden, 2000.
[3] T. Weiss and F. Jondral, “Spectrum pooling: An innovative strategy for
enhancement of spectrum efficiency,” IEEE Comm. Mag., vol. 42, no. 3,
pp. 8 – 14, Mar 2004.
[4] U. Berthold and F. Jondral, “Guidelines for designing OFDM overlay
systems,” in Proc. 1st IEEE Int. Symp. New Frontiers in DySPAN, pp.
626 – 629, Nov 2005.
[5] H. Tang, “Some physical layer issues of wide-band cognitive radio
systems,” in Proc. 1st IEEE Int. Symp. New Frontiers in DySPAN, pp.
151 – 159, Nov 2005.
[6] E. Hosseini and A. Falahati, “Power allocation grouping scheme con-
sidering constraints in two separate stages for OFDM-based cognitive
radio system,” in Proc. IEEE Int. Conf. Electrical Information and
Communication Technology (EICT), pp. 1 – 6, Feb 2014.
[7] Q. Qi, A. Minturn, and Y. Yang, “An efficient water-filling algorithm
for power allocation in OFDM-based cognitive radio systems,” in Proc.
IEEE International Conference on Systems and Informatics (ICSAI), pp.
Fig. 4: Power allocation by IGPP considering weight factors 2069 – 2073, May 2012.
[8] X. Zhou, B. Wu, P.-H. Ho, and X. Ling, “An efficient power allocation
Subchannel-1 Subchannel-2
40 20 algorithm for OFDM based underlay cognitive radio networks,” in Proc.
IEEE Global Telecommunications Conference (GLOBECOM), pp. 1 –
Allocated Power

Allocated Power

IPWF
30
GWFPP 15
5, Dec 2011.
IGPP [9] W. Jian, Y. Longxiang, and L. Xu, “Resource allocation based on
20 IPWF linear waterfilling algorithm in CR systems,” in Proc. IEEE Wireless
10 GWFPP
10 IGPP Communications, Networking and Mobile Computing (WiCOM), pp. 1
– 4, Sept. 2011.
0 5 [10] C.-H. Chen and C.-L. Wang, “An efficient power allocation algorithm for
1 2 3 4 1 2 3
multiuser OFDM-based cognitive radio systems,” in Proc. IEEE Wireless
Subcarriers Subcarriers
Subchannel-3 Subchannel-4
Communications and Networking Conference (WCNC), pp. 1 – 6, April
25 15 2010.
[11] Z. Tang, G. Wei, and Y. Zhu, “Weighted sum rate maximization for
Allocated Power

Allocated Power

20 IPWF IPWF
GWFPP 10 GWFPP OFDM-based cognitive radio systems,” Telecommun Syst, vol. 42, pp.
15 IGPP IGPP 77 – 84, Oct. 2009.
10 [12] P. Wang, M. Zhao, L. Xiao, S. Zhou, and J. Wang, “Power allocation in
5
5
OFDM-based cognitive radio systems,” in Proc. IEEE Global Telecom-
munications Conference (GLOBECOM), pp. 4061 – 4065, Nov 2007.
0
1 2 3 4
0
1 2 3 4
[13] D. P. Palomar and J. R. Fonollosa, “Practical algorithms for a family of
Subcarriers Subcarriers water-filling solutions,” IEEE Transaction on Signal Processing, vol. 53,
no. 2, pp. 686 – 695, Feb 2005.
[14] P. He, L. Zhao, S. Zhou, and Z. Niu, “Water-Filling: A geometric ap-
Fig. 5: Optimal power allocation vs subcarriers for each proach and its application to solve generalized radio resource allocation
subchannel using IPW, GWFPP and IGPP problems,” in IEEE transactions on Wireless Communications, pp. 3637
– 3646, July 2013.
subcarriers according to their noise power. Thus, the subcarrier [15] Q. Zhao and B.M.Sadler, “A survey of dynamic spectrum access: signal
processing, networking, and regulatory policy,” IEEE Signal Processing
that has highest noise power, comes first and the subcarrier that Magazine, vol. 55, no. 5, pp. 2294 – 2309, May 2007.
has lowest noise power, comes last for each subchannel. From [16] D. Tse and P. Viswanath, Fundamentals of Wireless Communications.
Fig. 5, we found that, for each subchannel, the algorithm IGPP Cambridge University Press, May.