Sei sulla pagina 1di 14

WIRELESS COMMUNICATIONS AND MOBILE COMPUTING

Wirel. Commun. Mob. Comput. (2012)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/wcm.2273

RESEARCH ARTICLE

Interference-aware spectrum handover for


cognitive radio networks
Dianjie Lu1,2 , Xiaoxia Huang2* , Weile Zhang3 and Jianping Fan2
1
School of Information Science and Engineering, Shandong Normal University, Jinan, China
2
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China
3
School of Electronic and Information Engineering, Xian Jiaotong University, Xian, China

ABSTRACT
Cognitive radio (CR) is a promising technique for future wireless networks, which significantly improves spectrum
utilization. In CR networks, when the primary users (PUs) appear, the secondary users (SUs) have to switch to other
available channels to avoid the interference to PUs. However, in the multi-SU scenario, it is still a challenging problem
to make an optimal decision on spectrum handover because of the the accumulated interference constraint of PUs
and SUs. In this paper, we propose an interference-aware spectrum handover scheme that aims to maximize the CR
network capacity and minimize the spectrum handover overhead by coordinating SUs’ handover decision optimally in the
PU–SU coexisted CR networks. On the basis of the interference temperature model, the spectrum handover problem is
formulated as a constrained optimization problem, which is in general a non-deterministic polynomial-time hard problem.
To address the problem in a feasible way, we design a heuristic algorithm by using the technique of Branch and Bound.
Finally, we combine our spectrum handover scheme with power control and give a convenient solution in a single-SU
scenario. Experimental results show that our algorithm can improve the network performance efficiently. Copyright © 2012
John Wiley & Sons, Ltd.
KEYWORDS
interference temperature; spectrum handover; power control

*Correspondence
Xiaoxia Huang, 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, China.
E-mail: xx.huang@siat.ac.cn

1. INTRODUCTION may fail to detect any available channel and have to stop
its transmission until an available channel emerges. Then,
Cognitive radio (CR) is an effective method to alleviate the situation becomes the same as the previous case. There-
the frequency scarcity problem [1]. While primary users fore, for SUs, either staying or changing channel requires
(PUs) have the priority to access the licensed spectrum, extra overhead, causing significantly degraded capacity.
secondary users (SUs) can opportunistically use spectrum Indeed, SUs can coexist with PUs on the same licensed
when no harmful effects to the PUs are introduced [2,3]. channels. FCC [4] suggested interference temperature
Intuitively, when PUs appear, SUs have two options, model (ITM) to regulate the SUs’ behavior. According
staying in the original channel or switching to another to this model, the SUs are allowed to coexist with the
channel. In the former case, SUs have to suspend their PUs when the SUs’ aggregated interference to the PUs is
transmission and waste time to wait until the PUs complete below a certain level. In other words, SUs can appropri-
their transmission. Similarly, in the latter case, the SUs ately coordinate their spectrum handover decision to fully
evacuate the occupied spectrum and move to some other exploit all the spectrum opportunities, while the interfer-
available ones. This is the so-called spectrum handover. ence to the PUs is low. In this paper, to maximize the
However, frequent spectrum handovers will cause sig- utilization of the spectrum opportunity and the network
nificant performance degradation to the SUs because of capacity, we propose an interference-aware spectrum
the large handover overhead such as the establishment of handover scheme based on ITM. We extend the ITM to
the wireless connections and re-establishment of the TCP a multi-SU spectrum sharing cognitive network and for-
links. If the number of licensed bands is limited, the SU mulate the decision problem of spectrum handover as a

Copyright © 2012 John Wiley & Sons, Ltd.


Interference-aware spectrum handover D. Lu et al.

constrained optimization problem. We show that such a the overhead of storage and computation. In addition, it is
problem is, in general, non-deterministic polynomial-time also necessary to resolve the issue that the pre-selected
hard and that obtaining the optimal solution is computa- target channel may no longer be available when the trans-
tionally infeasible. A heuristic algorithm based on the tech- mission begins. Mohamed et al. [15] proposed a scheme
nique of Branch and Bound is given to address this problem to reduce the number of spectrum handover by using unli-
in a workable way. Finally, we introduce power control censed channels as backup channels. However, if the pool
into the proposed scheme and give a convenient solution of unlicensed channels is limited, the scheme degenerates
in a single-SU scenario. In the single-SU CR network, to the traditional one. L. Giupponi et al. [16] proposed a
only the interference of SUs-to-PU needs to be consid- fuzzy-based approach that can flexibly make decisions on
ered. The problem can be simply solved by calculating spectrum handover in the context characterized by uncer-
the transmission range not interfering with PU. Simu- tain, incomplete, and heterogeneous spectrum usage infor-
lation results show that compared with two alternative mation. The main weakness of the fuzzy-based approach is
approaches, namely a random algorithm and a conserva- the dependability of their decisions on the way how mem-
tive algorithm, the network capacity can be increased by bership functions and fuzzy inference rules are set. Feng
up to 31% and 74%, respectively. et al. [17] extended the spectrum handover problem to min-
The rest of the paper is organized as follows. Section 2 imize the total latency for multi-hop spectrum handover
reviews the related literature on the spectrum access while the network remains connected. In [17], the order
and handover. Section 3 presents the system model. We of spectrum handover for multiple links is coordinated to
formulate the spectrum handover problem as a constrained maintain the network connectivity. With the coordination
optimization model in Section 4. Section 5 extend our pro- result, rerouting is triggered before the spectrum handover
posed model based on power control mechanism. Section 6 really happens in order to protect the communication from
shows our simulation scenario and discusses the simulation interruption. However, in these works, the interference of
results. Finally, we conclude the paper in Section 7. SUs-to-SUs has not been considered, which will degrade
the channel capacity because there is no coordination
among the SUs. Different to the prior works, we consider
2. RELATED WORK both SUs-to-PU interference and SUs-to-SUs interference
to optimize the CR network performance. In this paper, an
The prior works on spectrum access in the CR network interference-aware spectrum handover scheme is proposed
can be divided into two categories: the overlay access to maximize the utilization of the spectrum opportunity and
and the underlay access. In the overlay way, SUs use the the network capacity.
licensed channel opportunistically. Zheng et al. [5] pro-
pose a color-sensitive graph approach which character-
izes the interference between the PUs and the SUs using 3. SYSTEM MODEL
a binary model. In [6], the authors present a proactive
sensing model to discover as many spectrum opportuni- We consider a CR network consisting of Mp licensed
ties as possible in advance, while minimizing the average channels and Ns SUs. Each licensed channel can be
time taken to detect a vacant frequency band. However, denoted by C h.fc ; B/ where fc is the central frequency
in their works, the coexistence of PUs and SUs on the and B is the bandwidth. In this scenario, we assume
same channel is not considered. In the underlay way, an N SUs share a common licensed channel where N 2
SU can operate on the same frequency with PUs, provided f1; 2; : : : ; Ns g. However, SUs’ transmission will cause
the interference temperature at each licensed receiver does harmful interference to a PU when the PU appears on the
not exceed a threshold. Clancy [7] formalizes this ITM licensed channel. As we can see from Figure 1, PU trans-
to maximize both the capacity and the spectral efficiency mitter (PT) is sending data to the PU receiver (PR). For
in a specific RF environment. In [8–11], power control the SU transmitter, the outer circle (dashed) represents the
rules are adopted to reduce the interference to PUs. Game interference range, and the inner circle (solid) represents
theory is generally used to achieve the tradeoff between the the transmission range. In Figure 1, PR is in the interfer-
spectrum utilization and interference [12,13]. ence range of the SU transmitter, where the emissions from
As a critical issue in CR networks, spectrum handover the SU transmitter might cause unbearable interference
plays an important role in spectrum access. Recent which exceeds the interference temperature limit to PR.
works [14–16] have explored different spectrum handover Because PUs are licensed users, they have the priority to
schemes to balance the switching overhead and the access the licensed bands whenever and wherever needed.
spectrum opportunities. In [14], the authors proposed SUs opportunistically send their data by using the avail-
a proactive-decision spectrum handover algorithm. This able licensed bands. For example, as shown in Figure 2(a),
algorithm can reduce the time-consuming wideband sens- P U 1 and P U 2 hold licensed bands C h1 and C h2, respec-
ing through finding the available channels before PUs tively, and S U 1 and S U 2 coexist on the common licensed
appear. Nevertheless, this algorithm should keep the band C h1 simultaneously. We assume that all the SUs are
long-term observation outcomes and make the channel equipped with reconfigurable transceiver that can tune in
sequences ready before data transmission, which increases any frequency band. Usually, there are two cases in the

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

Receiver

SU PR PT
Interference Signal
PU
Transmitter
Noise

Figure 1. Interference between primary user (PU) and secondary user (SU). PR, PU receiver; PT, PU transmitter.

PU1
SU1
Ch1 PU2

Ch1
SU2

(a)

PU1
SU1
Ch2 PU2

Ch1
SU2

(b)

Figure 2. Coexistence of primary user (PUs) and secondary user (SUs). (a) SUs’ transmission without PU: both PU 1 and PU 2 do
not come; SU 1 and SU 2 coexist on the same channel Ch1. (b) SUs’ transmission with PU: when PU 1 comes and PU 2 does not,
SU 1 switches to channel Ch2; SU 2 continues its transmission on channel Ch1.

conventional spectrum handover schemes. In the ‘best’ But this will increase the spectrum sensing and switching
case, when a PU appears, all SUs switch to other idle cost and decrease the capacity of current channel. More-
channels without causing harmful interference to the PR. over, the performance becomes even worse if the number

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

of licensed channels is limited. In the ‘worst’ case, all SUs where Boltzmann’s constant k is 1:38  1023 J/K.
remain in the channel. SUs should suspend their transmis- PI .fc ; B/ is the interference power over the channel with
sion in order to not interfere with the PU until it leaves. central frequency fc and bandwidth B.
A compromise solution is to allow part of SUs to switch
to other channels, while the other part of SUs continue to
use the current channel. Thus the coordination among the 3.2. Path loss
SUs on the same channel is necessary to decide whether to
switch or not while keeping the accumulated interference We assume that the path loss is characterized by [19]
below the interference temperature limit.
In our scheme, firstly, each SU transmitter measures f .r/ D 10n log.r/ (3)
the interference temperature on the channel and exchanges
it with other SUs using the same channel. Then, all
where r represents the physical distance between the
the SUs will coordinate to make an optimal spectrum
sender and the receiver. Let n be the path loss exponent,
handover decision. Finally, on the basis of the decision,
which is usually a constant in the range of 2–4.
some SU transmitters will continue their transmission and
the others switch to other available channels. As shown in
Figure 2(b), when P U 1 appears on C h1, S U 1 switches to 3.3. Channel capacity
channel C h2, whereas S U 2 continues its transmission on
channel C h1. For the definition of channel capacity C , Shannon–Hartley
theorem states
3.1. Interference temperature  
S
C D B  log2 1 C (4)
The concept of interference temperature is similar to that No
of noise temperature. As illustrated in Figure 3, for a given
area, FCC establishes an interference temperature limit where B is the bandwidth of the licensed channel, S is the
TL .fc ; B/. This value of TL .fc ; B/ would be the maximum received signal power at the SU receiver. No is the noise
amount of tolerable interference for a given central and interference power over the band.
frequency fc and bandwidth B in a particular location.
In this paper, we use the following formula to define the
interference temperature [18].
4. THE SPECTRUM
HANDOVER MODEL
TN .fc ; B/ C TI .fc ; B/  TL .fc ; B/ (1) During the handover operation, the communication of the
SU has to be interrupted, and the network connectivity
where TN .fc ; B/ is the original noise floor, TI .fc ; B/ is will be seriously affected. From this point of view, large
the interference temperature of the channel with central latency can be introduced. Also, packets may be lost. As
frequency fc and bandwidth B. TI .fc ; B/ can be given as a solution, SUs can coordinate to continue their transmis-
sion or switch to other available channels. However, it also
raises new challenges, such as how to make the optimal
PI .fc ; B/
TI .fc ; B/ D (2) decision of spectrum handover considering both the
kB SU–PU interference and the SU–SU interference. In this
section, we firstly give an extended interference tempera-
ture model (EITM), which extend the ITM to a multi-SU
spectrum sharing cognitive network. In addition, we pro-
pose a spectrum handover optimization model to maximize
the channel capacity based on EITM.

4.1. Extended interference


temperature model

We assume that the PT is far from the PR and SUs. When


PU pairs appear, the transmission from PT leads to higher
interference, which makes it difficult for SUs to continue
transmission. The accumulated interference caused by SUs
also should be reduced to guarantee the PR’s signal to
interference-plus-noise ratio (SINR). So, SUs should make
Figure 3. Interference temperature. a decision of whether switch or not.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

SR ST PR PT
1
2
3

Transmission
K Interference

Figure 4. Interference model in multi-user cognitive radio network. PR, primary user (PU) receiver; PT, PU transmitter; SR, secondary
user (SU) receiver; ST, SU transmitter.

As shown in Figure 4, there are K .K  N / SUs coexist switching overhead. First, we give the channel capacity for
with a PU transmission pair on the same licensed band. The S Ui , which can be described by the following formula [18]
dashed lines represent the interference, and the solid lines
0 1
represent the transmission. We can see that all the K SU
transmitters cause accumulated interference to the PR. At B i C
B Li i Psu_tx C
the same time, each SU transmitter will interfere with any Ci D B log2 B
B1 C
C
C
K
P
other SU receivers. There are two interference constraints @ j A
kBTN .fc ; B/ C Lj i Psu_tx
in EITM. Firstly, SUs must limit the accumulated inter- j D1;j ¤i
ference introduced in the PU band below a certain level. (7)
Secondly, SUs should coordinate to initiate the non-
interfering transmission. Mathematically, these can be where Li i represents multiplicative attenuation due to
modeled as follows. fading and path loss between the i th SU transmitter and
As mentioned in Section 3, SUs-to-PU interference i th SU receiver. And then, we introduce the maximum
should not exceed the interference temperature limit TPU capacity function in the next part.
at a PR. For the PR, we must satisfy
4.2. Spectrum handover
K optimization model
P i
Mi Psu_tx
iD1 In this subsection, we formulate the spectrum handover
TN .fc ; B/ C  TPU .fc ; B/ (5)
kB decision as an optimization problem to maximize the CR
network capacity based on EITM mentioned earlier.
where Mi is a fractional value between 0 and 1, Let xi denote the decision of S Ui . If S Ui decides to
representing a multiplicative attenuation due to fading and switch to another channel, we set xi D 0, otherwise,
path loss between the i th SU transmitter and the licensed xi D 1. We further introduce x D Œx1 ; x2 ; : : : ; xN 
i
receiver. Psu_tx is the i th SU’s transmit power. to represent the decision of SUs. Then, the maximum
Moreover, in our model, SUs-to-SUs interference also capacity function is given by
j
should not exceed an interference temperature TSU for
the j th SU receiver. Then, the following inequality could arg max F.x/ (8)
x
be obtained.
where
K
P i
Lij Psu_tx N
X
iD1;i¤j j F.x/ D xi B  log2 .1 C si /
TN .fc ; B/ C  TSU .fc ; B/ 8j 2 N
kB iD1
(6)
and
where Lij is similar to Mi , except that it represents
multiplicative path loss between the i th SU transmitter and i
Li i Psu_tx
j th SU receiver. si D
N
P j
Our goal is to maximize the capacity of CR network kBTN .fc ; B/ C xj  Lj i Psu_tx
on the licensed channel while minimizing the spectrum j D1;j ¤i

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

The interference constraint for SUs-to-PU is 4.3. The solution

The optimization problem (12) is a multi-constraint


AxT  ˛ (9)
nonlinear 0/1 integer programming problem which is in
general non-deterministic polynomial-time hard. In order
i
where A D ŒA1 ; A2 ; : : : ; AN  with Ai D Mi Psu_tx and to solve this problem, we resort to the well-known Least
˛ D .TPU .fc ; B/  TN .fc ; B//kB. Cost Branch and Bound (LCBB) [20] algorithm. The basic
The interference for SUs-to-SUs can be described as idea of LCBB algorithm is built on two concepts. The first
a symmetric matrix where the entries value indicate one is Branch, which mainly focuses on how to select a
whether the interference exists. We denote the interference better extended node E (the live node whose children are
matrix as currently being generated) in the state-space tree so as to
obtain the optimal solution quickly. The second concept is
2 3 Bound. That is, kill the nodes that are unlikely to produce
0 Y2;1  YN ;1 the optimal solution to reduce searching space. In general,
6 Y1;2 0  YN ;2 7
6 7 LCBB is often used to solve a minimization problem. So,
YD6 : :: :: :: 7 (10)
4 :: : : : 5 we covert the objective function to
Y1;N Y2;N  0
arg min .F.x// (13)
x
where
Least Cost Branch and Bound algorithm relies on the

0 if i D j ; least cost function c.nodex / of a node nodex to select
Yij D 1i N the extended node and kill nodes. To simplify it, we
x i xj if i ¤ j ;
define the lower bound and upper bound of the least
cost function as LBB.nodex / and UBB.nodex /, which
We also denote satisfy LBB.nodex /  c.nodex /  UBB.nodex /.
The node with minimum LBB.nodex / will be selected
2 3
L1;1 L2;1  LN ;1 as the extended node. We denote upper as the upper
6 L1;2 L2;2  LN ;2 7 bound of a minimum cost solution. All live nodes with
6 7
LD6 :: :: :: :: 7 LBB.nodex / > upper will be killed. When a new created
4 : : : : 5
live node nodey satisfies the condition UBB.nodey / <
L1;N L2;N  LN ;N upper, upper should be updated by UBB.nodey /.
Figure 5 shows the flowchart of our algorithm. In the ini-
tialization step, the items of x are set to 0. It is convenient to
sort the SUs descendingly by their capacity-to-interference
2 1 3
Psu_tx ratios. We denote E:cp as the total capacity currently
6 2
Psu_tx 7
6 7 achieved and E:level as the level of the node E in the
and PD6 :: 7 state-space tree. In the while loop, if E:level DD N C 1
4 : 5
N
and E:cp < upper, E is an optimal answer node. The
Psu_tx node will be recorded, and upper should be updated by
the current profit E:cp. If E is not the answer node (has
The interference constraint for SUs-to-SUs is not been extended to the last level, i.e., E:level < N C 1),
we will create the left child El (xi D 1) and right child
Er (xi D 0) of E. As for El , if current x satisfies the con-
.Y ı L/ P  ˇ (11) straints (9), then add El to the live node list; otherwise, kill
El . Similarly, for the node Er , we calculate the LBB.Er /
where ˇ D Œˇ1 ; ˇ2 ; : : : ; ˇN T , ˇi is the and add Er to the live node list only if LBB.Er / < upper
 interference
i .f ; B/ and x satisfies the constraint (9) (11). Here, we do not
threshold for the i th SU, and ˇi D TSU c
 compute the lower and upper bound of El because it
TN .fc ; B/ kB. ı denotes the element-wise product.
is not difficult to find that LBB.El / D LBB.E/ and
Thus, the constrained optimization problem can be for-
UBB.El / D UBB.E/. After generating new nodes, we
mulated as
update the value of upper. Finally, we choose a new live
node with minimum LBB to be the next extended node.
maxfF.x/g One of the key problems in our LCBB algorithm is how
s.t. to compute LBB and UBB of a node. Algorithm 1 and 2
AxT  ˛; (12) shows the basic procedure. We denote cw as the current
.Y ı L/ P  ˇ; interference total, wi as the incremental interference and
xi ; xj 2 f0; 1g; 1  i  N ; 1  j  N pi as the throughput benefits when xi is set to 1. k is

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

Initialization
Create a root
node E

The LBB of E Output the


no
is not larger than best solution
upper ANS

yes

E is an yes E is the
answer node best solution no
till now

no yes

Extend E ANS = E
Get LBB and UBB Set the current profit
Kill nodes if necessary of E to upper

Selet a node E with


Update upper
minimum LBB

Figure 5. Flowchart of our Least Cost Branch and Bound algorithm.

Algorithm 1: Algorithm for computing the LBB Algorithm 2: Algorithm for computing the UBB
Input: cp; cw; k; x Input: cp; cw; k; x
Output: LBB Output: UBB
begin begin
b  cp; b  cp;
c  cw; c  cw;
for i k C 1 to N do for i k C 1 to N do
xi  1; xi  1;
if Satisfy(x) then if Satisfy(x) then
if c C wi  m then if c C wi  m then
c  c C wi ; c  c C wi ;
b  b  pi ; b  b  pi ;
else else
b  b  .m  c/pi =wi ; xi  0;
ret urn b;
ret urn b;
else
xi  0;

ret urn b;

can be satisfied and the nested if statement satisfies the


constraint (9). Note parts of the components of x can be
the index of the last item and m is the interference thresh- evaluated by tracing the state of E’s ancestor nodes while
old. Satisfy(x) returns the result whether the constraint (11) the rest of them are set to 0 to relax the constraint.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

5. SPECTRUM HANDOVER WITH According to Equation (3), the distance from the trans-
POWER CONTROL mitter to the receiver can be estimated by the following
equation
In our EITM, the SUs have only two choices: stay or
s
switch. Actually, SUs can avoid the interference to PUs Pmax
n
by appropriately lowering the transmit power. In this part, rD (17)
Psu_rx
we introduce power control into our spectrum handover
scheme. Because each SU is allowed to perform power
control, the problem becomes even more difficult because To guarantee that the SI NR at the SU receiver is greater
of both SUs-to-PU interference constraint and SUs-to-SUs than the minimum SI NR min required for decoding the
interference constraint. For convenience, we only consider signal, Psu_tx should satisfy
the scenario with single SU. In this case, only the interfer-
Psu_tx
ence of a single SU to the PU is specified. To the best of 10 lg  min (18)
our knowledge, power control has not yet been considered r n kBTN .fc ; B/
in the spectrum handover of CR networks.
In the power control-based spectrum handover scheme, When a PU appears in the licensed band, the maximum
the spectrum handover can be reduced by lowering the transmission range of the SU transmitter must not exceed
transmit power for continuing transmission if the receiver s
is still in the transmission range. Thus, the handover over- Psu_tx
r D n
min (19)
head can be reduced, and the capacity can be improved. 10 10 kBTN .fc ; B/
Assume that an SU is using a licensed channel opportunis-
tically; the spectrum handover decision can be obtained by Finally, we compare r  with r. If r   r, the SU
comparing the physical distance from the SU transmitter to continues to use the current channel; otherwise, it switches
the receiver with the maximum transmission distance with- to another channel. Algorithm 3 shows the pseudo code
out interfering PU. If the receiver is still covered after low- of solving the single-SU model. Note that cal_phy_dist./
ering the transmit power, the SU can continue to transmit; calculates the physical distance from the transmitter to
otherwise, it switches to an idle band. the receiver, cal_max_trans_pow./ gets the the maximum
Now, we consider how to calculate the maximum trans- transmit power, cal_max_trans_dist./ calculates the max-
mit power that can keep the interference at the nearby PU imum transmission range corresponding to the maximum
below TPU . From Equations (1) and (2), we can get the power. dec_value gives the decision of SU through the
transmit power Psu_tx that satisfies the ITM comparison function comp_dist./.

Psu_tx  Bk.TPU .fc ; B/  TN .fc ; B//=M (14)


Algorithm 3: Algorithm for the single-SU solution
where M indicates a multiplicative attenuation due to Input: B; Pmax ; Psu_rx ; ˇ
fading and path loss between the SU transmitter and Output: dec_value
the PR. begin
There are two possible cases in our scheme. In case one, r  cal_phy_dist.Pmax ; Psu_rx /;
the PU is inactive, and the SU can transmit with the power Psu_tx  cal_max_trans_pow.B; ˇ/;
Pmax to achieve the best signal strength at the receiver, r   cal_max_trans_dist.Psu_tx /;
where Pmax is the allowed maximum transmit power. In dec_value  comp_dist.r; r  /
the other case, PU is active, so SU should transmit with the
power constrained by Equation (14). Then, we can get the
transmit power of the SU
In the multi-SU scenario, SUs have two choices: switch
8 or lower power. The problem becomes even more difficult
<Pmax PU is inactive; because it is hard to know the exact power under the con-
Psu_tx D Bk.TPU .fc ; B/  TN .fc ; B// straints of SUs-to-PU interference and SUs-to-SUs inter-
: PU is active:
M ference. Heuristically, we divided the spectrum handover
(15) scheme into two steps: first, we calculate physical distance
of each SU pair. Then, we select the largest distance and
From Equation (3), when the SU transmits with power
calculate the power that can guarantee the transmission.
Pmax , the receiving power Psu_rx is
This power will be looked as the power level after con-
trol; second, we continue to consider the switch decision
Pmax (16)
Psu_rx D rn problem which is also a multi-constraint nonlinear 0/1 inte-
ger programming problem similar with the optimization
The secondary sender obtains the receiving power Psu_rx model mentioned in Section 4.2. Differently, the solution
through exchanging the information with the receiver. determines SUs to switch or lower power.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

6. PERFORMANCE EVALUATION generality, we assume that the distance between the SU


transmitters and receivers are uniformly distributed. We
We consider the wireless microphone as a primary sys- also assume all the SUs have the same interference thresh-
tem that operates on the channels with the bandwidth old TSU . For comparison, we evaluate three approaches:
of 200 kHz and the noise floor TN D 300 K. We set conservative, random, and our proposed algorithm. In the
Mi D 16  1012 and Lij varies from 9  1012 to conservative algorithm, all SUs lower the transmit power
16  1012 . Firstly, we evaluate the cumulated capacity when the PUs appear. Because the transmission cannot be
and overhead of the EITM-based spectrum handover with- guaranteed to succeed after the power falls, the network
out considering power control. Then, we provide insight performance will degrade. In the random algorithm, SUs
into the spectrum handover ratio and the effective data rate, switch to other channels randomly under the constraint of
which measure the performance of power control-based not interfering with the PU.
spectrum handover in CR networks. In Figure 6(a), we evaluate the system capacity with
respect to SU interference temperature limit TSU , which
6.1. Spectrum handover without varies from 0:5  104 to 3:0  104 K. We set the PU
power control interference temperature limit TPU D 8  104 K, which
guarantees that all the interference to PU is below the spec-
6.1.1. Cumulated capacity. ified limit. In Figure 6(b), we evaluate the system capacity
Figure 6(a) and (b) shows the cumulated capacity of with respect to the PU interference temperature limit TPU ,
the multi-SU CR network where N D 8. Without loss of which varies from 1:5  104 to 4:5  104 K. We also set

(a) Capacity vs. Interference Temperature Limit of SU

(b) Capacity vs. Interference Temperature Limitof PU

Figure 6. Impact of interference temperature limit of secondary user (SU) and primary user (PU).

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

TSU D 4  104 K where only the interference to PU is spectrum handover overhead through optimally handling
needed to be considered. the coexistence of PU and SUs.
Figure 6(a) and (b) show that the capacity rises quickly
as interference temperature threshold increases, although 6.2. Power control-based
a smaller threshold leads to significantly reduced capacity. spectrum handover
It means the higher threshold can ensure that more SUs
resume their transmission. Specifically, from Figure 6(a) 6.2.1. Spectrum handover ratio.
and (b), we can see that, as the interference limit rises, We assume that there is only one SU pair that uses the
the capacity of conservative algorithm increases firstly, same channel with the PU. The arrival process of PUs is
and then remains unchanged. This is because the trans- assumed to be Poisson distributed with rate  [21]. Let
mission cannot be continued because lower power leads pa be the probability that a PU appears during a slot.
to lower SI NR. It can also be seen that the performance Assume that the appearance in each slot is independent.
of all algorithms is comparable firstly. The reason is that In the aggressive approach, the SU must perform spectrum
when the limit is low, only few SU can access the licensed handover once the PU appears. So, we can get the spectrum
band. Obviously, our method is superior to the others handover ratio of the aggressive approach
when the interference temperature limit becomes higher.
Our proposed approach improves the capacity by up to paggr D P f PU is active g D pa (20)
31% and 74% than random and conservative algorithms
in Figure 6(a). Additionally, we observe the difference However, in our proposed approach, spectrum handover
between the two figures. As for all the algorithms, the happens only when the maximum transmission distance
capacity in Figure 6(a) is larger than that in Figure 6(b) without interfering PU is less than the physical distance
at the same point of interference limit. It means that the from the SU transmitter to the receiver (r  < r). Then, the
interference temperature limit of PU has a greater impact ratio of spectrum handover can be expressed as
on that of SU.
p D P fr  < r \ PU is active g D P fr  < rg  paggr (21)
6.1.2. Cumulated overhead.
In Figure 7, we compare the spectrum handover over- In the ‘best’ situation, P fr  < rg D 0, which means
head of the three approaches with respect to the number of that the spectrum handover ratio p D 0; in the ‘worst’
SUs. In the aggressive algorithm, all SUs on the licensed situation, P fr  < rg D 1, which means that the spectrum
channel switch to other available channels as soon as the handover ratio p D paggr . Figure 8 shows the performance
PU appears. Assume that the SU receivers are uniformly gain of the proposed scheme. Assume that the SU receivers
distributed in the coverage of the corresponding SU trans- are uniformly distributed in the coverage of the transmitter,
mitter, and a handover costs 3:5 ms, we can calculate we study the spectrum handover ratio of the conventional
the cumulative overhead of all the handovers. As shown scheme and the scheme with power control. The test is
in Figure 7, the overhead of the aggressive algorithm is repeated for 10 000 times. As shown in Figure 8, the aver-
larger than others because the handovers lead to longer age handover ratio for SUs increases with respect to the
latency. Obviously, our approach can drastically reduce the arrival rate of PU. That means spectrum handover happens

Figure 7. Cumulative overhead versus number of secondary users.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

0.016
Spectrum handover without power control
0.014 Spectrum handover with power control

Spectrum handover ratio


0.012

0.01

0.008

0.006

0.004

0.002

0
0 0.2 0.4 0.6 0.8 1
Appearance prob. of primary user

Figure 8. Spectrum handover ratio versus arrival rate of primary user.

Figure 9. Effective data rate versus arrival rate of primary user.

more frequently when the arrival rate of PU increases. We introduce the spectrum handover ratio p mentioned
However, our proposed scheme can greatly reduce the earlier, and LMP can be determined by
ratio. We can observe from Figure 8 that our approach
n  
X
reduces the percentage of spectrum handover by up to 33% m
when the PU’s arrival rate reaches 1.0. pm D j p j .1  p/mj (23)
j D0

6.2.2. Effective data rate. Denote W as the traffic load of an SU. We define
We introduce the definition of link maintenance prob- effective data rate Re as the average number of bits
ability (LMP) [22] to investigate the effective data rate received per unit time. Then, the Re of the SUs can be
of our scheme. LMP is defined as the probability that an expressed as
SU finishes its transmission within n handover trials. For
simplicity, we partition the transmission time of both PUs W
and SUs into slots. We assume that the transmission of the Re D (24)
TQ
SU lasts for m slots; the LMP can be defined as
where TQ is the average time of an SU’s transmission with
pm , Prfthe number of handover trials is no more traffic load W . In addition, the spectrum handover scheme
than nj an m_slot transmissiong needs to perform the radio sensing and spectrum handover
(22) for each handover trail. Let Ts be the latency of channel

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

sensing and spectrum handover. The average transmission REFERENCES


time TQ becomes
n h 
X i 1. Mitola J, III. Cognitive radio: an integrated agent archi-
m
TQ D j p j .1  p/mj  j Ts C Td (25) tecture for software defined radio, PhD Thesis, KTH
j D0
Royal Institute of Technology, Stockholm, Sweden,
2000.
where Td D W =rSU , and rSU is the transmission’s rate of
SU. When n D m, pm D 1, it means that the SU can finish 2. Akyildiz IF, Lee WY, Vuran MC, Mohanty S. Next
its transmission within n handover trials. In this case, the generation/dynamic spectrum access/cognitive radio
transmission time in Equation (25) can be written as wireless networks: a survey. Computer Networks 2006;
50: 2127–2159.
TQ D .mpTs C Td / (26) 3. Haykin S. Cognitive radio: brain-empowered wireless
communications. IEEE Journal on Selected Areas in
In Figure 9, we compare the average effective data rate Communications 2005; 23(4): 201–220.
of the two schemes with respect to . We set W to 1500 bits 4. Federal Communications Commission. Establishment
and rSU to 12  106 bit/s. As expected, a smaller  leads to of interference temperature metric to quantify and
significantly improved effective data rate because the rate
manage interference and to expand available unli-
decreases as  increases. As we can see, the effective data
censed operation in certain fixed mobile and satellite
rate heavily depends on the latency of channel sensing and
spectrum handover. In addition, the effective data rate also frequency bands, ET Docket 03-289, Notice of Inquiry
decreases quickly as the latency of channel sensing and and Proposed Rulemaking, 2003.
spectrum handover grows. As illustrated in Figure 9, the 5. Zheng H, Peng C. Collaboration and fairness in oppor-
effective data rate of our proposed scheme is higher than tunistic spectrum access. In Proceedings of the IEEE
the scheme without power control. When the PU arrival International Conference on Communication, Seoul,
rate is high, our approach improves the effective data rate Korea, Vol. 5, May 2005; 3132–3136.
by up to 30%. In other words, the proposed scheme is more 6. Kim H, Shin KG. Efficient discovery of spectrum
effective when the PU appears frequently. opportunities with MAC-Layer sensing in cognitive
radio networks. IEEE Transactions on Mobile Comput-
7. CONCLUSION ing 2008; 7(5): 533–545.
7. Clancy TC. Formalizing the interference temperature
Spectrum handover is a critical issue in CR networks.
model. Wiley Journal on Wireless Communications
Frequent spectrum handover leads to disruptions of SU’s
and Mobile Computing 2007; 7(9): 1077–1086.
transmission. To address this problem, we propose an
interference-aware spectrum handover model. In this 8. Hoang AT, Liang YC. Downlink channel assignment
model, we explore the coexistence opportunity of SUs and and power control for cognitive radio networks. IEEE
PU to maximize the capacity of the licensed channels and Transactions on Wireless Communication 2008; 7(8):
minimize the spectrum switching overhead considering 3106–3117.
the accumulated interference at PUs and SUs. We formu- 9. Hoang AT, Liang YC, Islam MH. Power control and
late the decision of spectrum handover as a constrained channel allocation in cognitive radio networks with
optimization problem and then derive the solution by primary users’ cooperation. IEEE Transactions on
employing the technique of LCBB. At last, we extend our Mobile Computing 2010; 9(3): 348–360.
proposed model on the basis of power control. Numerical
10. Xiang J, Zhang Y, Skeie T, He J. QoS aware admis-
results show that our algorithm can improve the network
performance significantly. sion and power control for cognitive radio cellular
In the future work, we would like to extend our work to a networks. Wiley Wireless Communications and Mobile
dynamic scenario where the status of PUs (active/inactive) Computing 2009; 9(11): 1520–1531.
and the activity of SUs (arrival/departure) should 11. Almalfouh SM, Stuber GL. Interference-aware power
be considered. allocation in cognitive radio networks with imperfect
spectrum sensing. In Proceedings of the IEEE Confer-
ence on Communications, Cape Town, South Africa,
ACKNOWLEDGEMENTS July 2010; 1–6.
This work was supported by the National High Tech- 12. Lee H, Kwon H, Motskin A, Guibas L. Interference-
nology Research and Development Program of China aware MAC protocol for wireless networks by a
(863 program) under grant No. 2011AA010503 and the game-theoretic approach. In Proceedings of the IEEE
National Natural Science Foundation of China under Infocom’09, Rio de Janeiro, Brazil, April 2009;
grant No. 60903192. 1854–1862.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
D. Lu et al. Interference-aware spectrum handover

13. Pang J, Scutari G, Palomar DP, Facchinei F. Design Xiaoxia Huang, associate professor,
of cognitive radio systems under temperature- received the BE and ME degree in
interference constraints: a variational inequality Electrical Engineering from Huazhong
approach. IEEE Transactions on Signal Processing University of Science and Technol-
2010; 58(6): 3251–3271. ogy, China, in 2000 and 2002, respec-
tively. She received her PhD from the
14. Wang CW, Wang LC. Modeling and analysis for
Electrical and Computer Engineering,
proactive-decision spectrum handoff in cognitive radio
State University of Florida in 2007.
networks. In Proceedings of the IEEE International She is currently Deputy Director of the Center for Real-
Conference on Communications, Dresden, Germany, Time Monitoring and Communications Technologies in
June 2009; 1–6. Shenzhen Institutes of Advanced Technology. Her research
15. Kalil MA, Al-Mahdi H, Mitschele-Thiel A. Analysis interest includes wireless sensor networks, cognitive radio
of Opportunistic Spectrum Access in Cognitive Ad Hoc networks, wireless communications, and mobile comput-
Networks. Springer: Berlin/Heidelberg, 2009. ing. She has served on technical program committee (TPC)
16. Giupponi L, Perez-Neira AI. Fuzzy-based spectrum of IEEE WCNC 2011, ICC 2011, IEEE Globecom 2011,
handoff in cognitive radio networks. In Proceedings QShine 2010, International Conference on Embedded and
of the 3rd International Conference on CrownCom’08, Ubiquitous Computing 2010. She has served as a reviewer
of many prestigious journals and conferences, for instance,
Singapore, May 2008; 1–6.
IEEE Transactions on Mobile Computing, IEEE Trans-
17. Feng W, Cao J, Zhang C, Liu C. Joint optimization
actions on Communications, IEEE Transactions on Wire-
of spectrum handoff scheduling and routing in multi- less Communications, IEEE Transactions on Vehicular
hop multi-radio cognitive networks. In Proceedings Technology, ACM Wireless Networks. She has published
of the IEEE International Conference on ICDCS’09, 20 papers on the top conference and journal of wire-
Montreal, Quebec, Canada, July 2009; 85–92. less networks and communications, with a coauthor of
18. Clancy TC. Dynamic spectrum access using the inter- an academic work. The six patents of wireless net-
ference temperature model. Annals of Telecommunica- works are pending trial. She has served as one of the
tions 2009; 64(7): 573–592. key members of the NSFC project, 863 national major
19. Rappaport TS. Wireless Communications Principles projects, Shenzhen Public projects, national major projects,
and Practice. Prentice-Hall: New Jersey, US, 1998. Guangdong Province major projects and 973 sub-project
and the National Science Foundation (NSF) projects. She
20. Horowitz E, Sahni S, Rajasekaran S. Computer
is one of academic leaders of Shenzhen Key Laboratory of
Algorithms/C++. Silicon Press: New Jersey, US, 2008.
High Performance Data Mining.
21. Kalil MA, Liers F, Volkert T, Mitschele-Thiel A. A
novel opportunistic spectrum sharing scheme for cog- Weile Zhang received the BS degree
nitive ad hoc networks. In Proceedings of the 5th in Information Engineering from Xi’an
Workshop on Mobile Ad-Hoc Networks (WMAN’09), Jiaotong University, Xi’an, China, in
Kassel, Germany, Vol. 17, March 2009. 2006. From October 2010 to October
22. Wang LC, Anderson C. On the performance of spec- 2011, he was a visiting scholar at the
trum handoff for link maintenance in cognitive radio. Department of Computer Science, Uni-
In Proceedings of the 3rd International Symposium on versity of California, Santa Barbara,
Wireless Pervasive Computing, Santorini, Greece, May CA, USA. His research interests include
broadband wireless communications, array signal process-
2008; 670–674.
ing and localization in wireless networks.

Jianping Fan received his BS degree


from Nankai University in 1984, PhD
AUTHORS’ BIOGRAPHIES degree in Institute of Software (IOS),
Chinese Academy of Sciences (CAS)
Dianjie Lu received the PhD degrees in 1990. He worked at Institute of
in 2012 from Shenzhen Institutes of Computing Technology from 1990 to
Advanced Technology (SIAT), Chinese 2006, where he served as deputy chief
Academy of Sciences. He is cur- engineer of national research center for
rently an assistant professor at School intelligent computing systems director of National Engi-
of Information Science and Engi- neering Center on High Performance Computing, vice
neering, Shandong Normal University, Director of institute of computing technology. Since 2006,
China. His current research interests he served as Director of Shenzhen Institute of Advanced
include resource allocation and routing in cognitive Technology, CAS. Dr. Fan’s research interests include high
radio networks. performance computing, Grid computing, and computer

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm
Interference-aware spectrum handover D. Lu et al.

architecture. He took part in and developed Dawning I, nology progress, and the outstanding young scientist of
Dawning 1000, Dawning 3000, Dawning 4000, and other CAS. Dr. Fan has served as editor of journal of com-
series of Dawning supercomputers. He has published puter research and development, joint professor in Nankai
more than 70 papers and 1 book. He also acquired 11 University, council member of china digital library con-
pending or issued patents. Dr. Fan has received many sultant committee, advisor of the 11th five years science
awards, such as outstanding award of CAS science and and technology development plan of ministry of informa-
technology progress, first price of national science and tion industry, and general chair of HPC China 2007 and
technology progress-first price of Beijing science and tech- GCC 2008.

Wirel. Commun. Mob. Comput. (2012) © 2012 John Wiley & Sons, Ltd.
DOI: 10.1002/wcm

Potrebbero piacerti anche