radio resource management in cdma- based cognitive...
TRANSCRIPT
47
Chapter 3
Radio Resource Management in CDMA-
Based Cognitive Networks
3.1 Introduction
This chapter discusses Radio Resource Management (RRM) issues in Cognitive Radio
Networks (CRN) [11]. The main intent of the RRM is to efficiently utilize the
available network resources for providing satisfactory quality of service ( OQ S ) to the
mobile users.
The resource allocations in both the primary and the secondary network is
jointly considered, and then study of the optimum transmission power and rate
allocations for supporting best effort traffic in the CRN is done. Design of Interference
Model for Scheduling Algorithm in Cognitive Radio Networks (CRNs) is presented. A
method is also presented to compute throughput for optimal scheduling problem with
an objective to achieve proportional fairness (PF) of the long term average
transmission rates among different links, subject to the interference constraints within
the cognitive network and imposed by the primary network.
3.1.1 Cognitive Radio Network (CRN)
It is a form of wireless communication in which a transceiver can intelligently detect
which communication channels are in use and which are not, and instantly move into
vacant channels while avoiding occupied ones. This optimizes the use of available
48
radio-frequency (RF) spectrum while minimizing interference to other users [47]. In its
most basic form, CRN is a hybrid technology involving software defined radio (SDR)
as applied to spread spectrum communications.
Functions of cognitive radio include the ability of a transceiver to determine its
geographic location, identify and authorize its user, encrypt or decrypt signals, sense
neighbouring wireless devices in operation, and adjust output power and modulation
characteristics.
3.1.2 CRN Architecture
In CRN, new ideas have been proposed to provide more flexible and efficient usage of
the spectrum. CRN Architecture is shown in figure. 3.1.
Primary User (or Licensed User) has a license to operate in a certain spectrum
band. Secondary Network (or Unlicensed Network) does not have license to operate in
a desired band. Hence, the spectrum access is allowed only in an opportunistic
manner. Secondary User (or Unlicensed User) has no spectrum license. Therefore
additional functionalities are required to share the licensed spectrum band [49] [50].
Figure 3.1: CRN Architecture
There are two types of cognitive radio networks:
49
•••• Full Cognitive Radio: Every possible parameter observable by a wireless node or
network is taken into account.
•••• Spectrum Sensing Cognitive Radio: Only the radio frequency spectrum is
considered.
3.1.3 Resource Management in Cognitive Radio Networks
Management of Spectrum
The specification of the licensed spectrum is that, when there are no primary users on
the spectrum or the spectrum is under-used then only it can be used by secondary
networks. The secondary network capacity depends on the transmission activity of the
active primary users. The opportunistic transmission behavior of the secondary networks
imposes distinctive challenges for their coexistence with the primary networks and OQ S
provisioning of the carried services.
Scheduling in cognitive radio networks
Sophisticated scheduling schemes are desirable to allocate resource competently and
fairly among the users in a CRN [51] [52]. Compared to the scheduling in traditional
wireless networks, scheduling in a CRN is more complex due to the opportunistic nature
of the networks [53].
Medium Access Control (MAC) protocols for CRNs
Some research efforts are put on designing MAC protocols in CRNs in both industry
standardization and academic research projects. From the standardization point of view,
the current IEEE 802.22 draft is the first worldwide standard related to cognitive radio.
Its MAC employs the superframe structure, where synchronized distributed sensing, fast
sensing using energy detection, and fine sensing using feature detection are used.
At the beginning of every superframe, the BS sends a special preamble and super
frame control header (SCH) through each TV channel (up to 3 contiguously) that can be
used for communication and is guaranteed to meet the current protection requirements.
In addition, IEEE 802.22 is operated in the point-to-multipoint model, which is
comparably easier than the cases without the control of the BS.
50
Admission control for cognitive radio networks
Admission control is important for limiting the traffic load and providing desired OQ S ,
but the complexity inherent in resource management of ad-hoc networks will make the
admission control in cognitive ad-hoc network even harder [55].
Designing of an effective admission control scheme in a CRN is especially
challenging if it is to be implemented with low complexity.
3.2 Optimum Scheduling (Slot-By-Slot) in CRNS
The secondary links can transmit at the same spectrum as the primary links as long as
the interference that the secondary links cause to the primary links is below a pre-
negotiated interference threshold. In this section the effect of setting different
interference thresholds on the transmission rate of the secondary links and how the
secondary transmissions affect the transmissions of the primary links is studied [11].
3.2.1 System Depiction
Primary network is a CDMA-based cellular network in which different frequency
bands are used in the uplink and the downlink. Instead of communicating with the base
station (BS) directly in the primary network, some mobile stations (MSs) near one
another may form an ad hoc CRN and communicate directly with one another (Figure
3.2).
The same spectrum is shared by the secondary transmissions as the uplink of the
cellular network through spectrum underlay, and cause interference to the uplink
transmissions in the primary network. For groundwork study, only single hop
transmissions in the CRN are considered. The CRN can share the same control
channels with the primary network.
The primary BS can monitor the secondary-to-primary interference and centrally
control the CRN. This is not a problem if the primary network has relatively light
traffic load, which is most likely the case when a secondary network is allowed and
interference threshold of secondary networks with spectrum underlay ( )thINT is set to
be reasonably high, then the control channel in the primary network has low traffic
51
load and can be well used for the secondary network as well.
Figure 3.2: Design of primary and secondary networks
In figure 3.2, PU =Primary user and SU= Secondary user.
Since only the uplink is considered for the primary network, the transmitters are
the MSs, and they share the same receiver, which is the BS. The transmission power of a
primary link is denoted as ipP , , for i=1, 2,…Mp, and that of secondary link is denoted
as isP , , for i= 1, 2,… sM , where,
Mp and Ms= the number of primary and secondary links respectively.
The transmission rate of the thi primary link is ipR . and that of a secondary link is
isR . .The maximum transmission power of the transmitter is limited by max,pP for a
primary link and max,sP for a secondary link.
Let,
2 ,p p ijg = the link gain from the transmitter of the thi primary link to the receiver of
the thj primary link,
2 ,p s ijg = the link gain from the transmitter of the thi primary link to the receiver of
52
the thj secondary link.
2 ,s p ijg = the link gain from the transmitter of the thi secondary link to the receiver
of the thj primary link.
2 ,.s s ijg = the link gain from the transmitter of the thi secondary link to the receiver
of the thj secondary link.
Each link has a strict signal-to-interference-plus-noise ratio (SINR) requirement at
the receiver, which should be above *
pγ (minimum SINR for successful decoding of
primary link) for the primary links and *
sγ (minimum SINR for successful decoding
of secondary link) for the secondary links after the signal is despreaded.
At the primary receivers, there are two interference models for measuring the
interference level at the primary receivers. The first is to monitor the noise and total
interference from all the secondary and the primary transmitters. This measured
interference is then compared with the interference threshold, and the result is used to
regulate the secondary transmissions. In this way, the interference at the primary
receiver caused by the secondary transmissions is not detected separately. This model
does not require a priori knowledge of the RF environment, and consequently does
not need to distinguish the licensed signals from the interference and noise [58].
The requirement of the second model requires that the aggregate signal strength
coming from the secondary transmitters is measured at the receiver of a primary link
and compared with the interference threshold. In this case, interference caused by the
secondary transmitters should be separated from that caused by primary transmitters
in order to calculate the interference level. In the next section, the power and rate
allocation problems are formulated in the primary-secondary scenario based on these
two interference models.
3.2.2 Power Allocation and Optimum Rate
Consider the SINR requirement of each primary and secondary link. For a primary
link, required SINR can be satisfied if,
53
*
,2,
1
,2,1
,2,,
p
jipsjs
N
j
jippjpj
iippipip
gPgP
gPG
S
γ
η
≥
++∑∑=
≠
(3.1)
Where, , .p i p iG W R= , is the processing gain of the thi primary link. In the
denominator on the left-hand side of (3.1), the first term is the interference from all
other primary links, the second term is the interference from all the secondary links, and
η is the background additive white Gaussian noise (AWGN) power and W is the
spreading bandwidth.
The SINR of the thi secondary link can be satisfied if,
, 2 ,
, , 2 , , 2 ,
p
s i s s ii
N
s i s j s s ji p jgp s jij ij i
WP g
R P g P η≠
=
+ + ∑ ∑
(3.2)
Between the brackets in the denominator on LHS of (3.2), the first term is the
interference from all other secondary links, and the second term is the interference from
all the primary links.
The transmission power of the secondary link is also limited by the interference
threshold for the first interference model,
(3.3)
and for second Interference model,
(3.4)
Depending on the above conditions, the optimization problem is formulated. This
maximizes a function (normally convex) of the secondary link transmission rates,
subject to the SINR requirements of the primary and secondary links and the given
interference threshold, where Rs = (Rs, 1, Rs,2, . . ., RsNs ).
Ns
s, j s2p, ji th
j i
P g INT=
≤∑
, 2 , 2 ,1 .
N s
p j p p ji s p ji thj i j s jP g P g IN Tη
≠ =+ + ≤∑ ∑
54
, 2 ,
1 1
, 1,2,.....,Ns
p j p p ji s, j s2p, ji th p
j j
P g P g η INT i M≠ =
+ + ≤ =∑ ∑
Max f (Rs) (3.5)
Such that, interference for primary link k should satisfy
(3.6)
Similarly, interference for secondary link should satisfy,
(3.7)
In this threshold should be satisfied by,
(3.8)
and the power should be satisfied by the following equation,
, ,max,0 1,2,...,p i p pP P i M≤ ≤ =
(3.9)
, ,max,0 1, 2,...,s i s sP P i M≤ ≤ =
(3.10)
In the optimization problem, the constraint is based on the first interference
, 1, 2,....,s.i s2s,ii
s sNp
s,i s, j s2s, ji p, j p2s, jij ij i
WP gγ i M
R P g P g η
∗
≠=
≥ = + + ∑ ∑
1
, 1, 2,.....,p.i p.i p2p,ii
p pNs
p, j p2p, ji s, j s2p, jij ij
G P gγ i M
P g P g η
∗
≠=
≥ =+ +∑ ∑
55
model. If the second interference model is used, the constraint (3.8) should be changed
to (3.4).
Two objective functions based on different rate allocation criteria are considered.
First, a simple equal speed allocation (ESA) is considered. It means that, all
secondary links transmit at the same rate. The ESA formulation provides perfect rate
fairness among all secondary links.
However, links with poor SINR conditions transmit higher power to achieve the
same rate as other links. Therefore, the link with the worst SINR condition limits the
transmission rate of all links in the secondary network. The second rate allocation
criterion is based on the proportional fairness (PF) for the rate allocations. The concept
of PF is a good scheduling objective to balance the fairness among users and the
resource utilization efficiency [59].
3.2.3 Interference Threshold
A higher interference threshold allows higher transmission power from the secondary
links and can potentially increase the transmission rate of the secondary links. On the
other hand, as the secondary links increase their transmission power, they cause more
interference to the primary links. As a result, transmission power of the primary links
should also be increased. The mutual interference effect eventually reaches a balance,
and then neither the primary nor the secondary links can increase the transmission power.
Secondary-to-primary interference is maximized when at least one MS in the primary
network reaches max,pP .
Consider that homogeneous traffic is carried out by the primary links.
Then,
pip GG =, …….for all i=1, 2,…, pM .
Let ,p iINT represent the aggregate noise and interference that the thi primary link
experiences from all other primary links and all secondary transmitters. With perfect
power control, the actual SINR for the primary link at the BS receiver input is equal to
*
pγ and all the primary links have an equal received power rpP , at the BS. In this case,
we have,
56
*
, ,p p r p i pG P INT γ= or *
. ,p i p p r pINT G P γ=
As the transmission power is limited by max,pP , we have,
max,
,2
,
, p
iipp
rp
ip Pg
PP ≤= (3.11)
It can be written as,
,max 2 ,
, *
p p p p ii
p i
p
G p gINT
γ≤ (3.12)
if, i=1, 2 ... pM ,
Let,
,max 2 ,
,max , *min min
p p p p ii
p p ii i
p
G P gINT INT
γ= = (3.13)
Then ,maxpINT is the upper bound of a meaningful interference threshold in the
sense that if ,maxth pINT INT≤ , the transmission power of the second links is limited by
thINT to satisfy the SINR requirements of the primary links. When ,maxth pINT INT≤ ,
transmission power of the secondary links may not be limited by thINT , since the SINR
requirement at the primary network limits the transmission power of the secondary
links. When ,maxth pINT INT≤ , the SINR of the primary links can be protected (within the
maximum transmission power limit of the MSs) as long as the actual secondary-to-
primary interference is below the threshold. Since 2 ,p p iig are random variables, below we
find the distribution of ,maxpINT and then its mean value. Given for any value y > 0,
{ } ,max 2 ,
,max *Pr Pr min
p p p p ii
pi
p
G P gINT y y
γ
≤ = ≤
(3.14)
*
2 ,
,max
Pr minp
p p iii
p p
yg
G P
γ = ≤ (3.15)
57
*
2 ,
,max
1 Prp
p p ii
i p p
yg
G P
γ = − > ∏ (3.16)
*
2 .
,max
1 1 Pr
PM
p
p p ii
p p
yg
G P
γ = − − ≤ (3.17)
In the above derivation it is assumed that all 2 ,p p iig ’s are independent and
identically distributed. The mean value of ,maxPINT , can be found as,
{ }( ),max ,max
0
1p pE INT pr INT y dy
∞
= − ≤ ∫ (3.18)
When finding
*
2 ,
,max
Pr ,p
p p ii
p p
yg
G P
γ ≤ we consider that the transmitted
signals suffer from both path loss and log-normal fading. Then 2 ,
Xe
p p ii ig Ad
βα −−= , where
A is the received power at a reference distance,i
d is the distance (normalized to the
reference distance) between the transmitter and the receiver of the thi link, α is the path
loss exponent, 1010log=β , and X is a normally distributed random variable with zero
mean and standard deviation of σ .
The distribution of 2 ,p p iig can be found as,
(3.19)
When all MSs are uniformly distributed in a circular area of radius D, the
*
2 ,
,max
*
,max
*
,max
Pr
Pr
Pr ln ln
p
p p ii
p p
pX
i
p p
p
i
p p
yg
G P
yAd e
G P
yd X
AG P
α β
γ
γ
γα β
−
≤
= ≤
= − − ≤
58
probability density function ( pdf ) of d is ( )Zfd = 22 Dz for Dx ≤≤0 .
Then,
( ) ( )*
, ,max
*
2 ,
,max
2
0 1
Pr
0,
p
p p
p
p p ii
p p
D
x
yln z ln
AG p
yg
G P
fd z dz N dx
γα
β β
γ
σ∞
− −
≤
∫ ∫ (3.20)
Where ( )2,x
N µ σ is the pdf of a normal distribution with mean of µ and standard
deviation ofσ .
3.3 Packet Transmission Scheduling and Joint
Admission Control in CRNs
In this section resource management for supporting traffic with strict quality of service
requirements is studied in a CRN with spectrum underlay. Both admission control and
packet transmission scheduling are considered.
When the traffic load at the primary network is relatively stationary, the capacity
change in the CRN is relatively stationary. Using admission control can effectively
ensure the amount of admitted traffic below the CRN system capacity over a long term..
The actual QoS of the admitted traffic is achieved through packet-by-packet
scheduling. Both streaming traffic and non-real-time (nrt) traffic are considered. The
non-real-time data traffic necessitates an average transmission throughput. By
opportunistically taking advantage of the random link and interference conditions during
the scheduling process, very low session outage probability can be achieved for the
streaming traffic, and the nrt traffic can receive the required throughput.
3.3.1 System Depiction
Consider the identical CRN networks. Two types of traffic are served, non-real-time
(nrt) data traffic and streaming traffic.
59
Let,
Z= a set of the nrt data connections.
S= a set of the streaming connections.
If each link carries only one connection, Z and S can also be used to represent the
set of links carrying the nrt data and that of the links carrying the streaming traffic,
respectively. Each nrt traffic source generates data at a random rate and the data link
requires an average transmission throughput ireqR , .
In case of the streaming traffic, all data can be pre-stored in a buffer at the
transmitter, or the source at the transmitter generates packets after the connection is
established and stores the packets in a buffer before they are transmitted. After a link is
established, the data stored in a buffer at the receiver is replayed at a constant rate ireqR , .
The instantaneous transmission rate of link ( )tRi i, , varies with time due to
random resource availability, the time varying channel and interference conditions.
There is a buffer with an infinite size at each receiver for each nrt link and therefore the
instantaneous transmission rate can be as high as the channel conditions and available
resources allow. For the steaming traffic, the receiver of the thi link has a buffer with a
maximum size of ireqi RB ,max, .
The receiver starts playing the data as soon as its data buffer becomes full for the
first time. Replayed data is removed from the buffer. Session outage occurs when the
receiver buffer does not have sufficient data to keep the required replaying rate.
Intuitively, larger iBmax, results in fewer session outages since the session can tolerate up
to iBmax, consecutive link outages ( )( )0=tRi without a session outage.
If iBmax, is larger, it introduces longer delay to start playing the data at the
receiver. The CR nodes of each link measure the link gain, ( )tg ii which includes both
the path loss and log-normally distributed shadowing. Information of the transmission
power is included in a data message, and the receiver finds the link gain based on the
received signal strength. The aggregate interference that all primary transmissions cause
to the thi secondary link, ( )2 ,p s iINT t , is measured by the CR receiver at the beginning
of each time slot when there is no transmission activity in the CRN. It is assumed that
60
the time slots in the CRN and the primary network are perfectly synchronized, so that
( )2 ,p s iINT t remains constant during one time slot.
The long term average of ( )2 ,p s iINT t is denoted as� 2 ,p s iINT . The link gain
between the transmitter of the thi secondary link to the primary BS, ( )tg pis2 , is measured
at the secondary transmitter based on the received signal strength of the pilot signal
from the BS. The short term link gain and interference values are used for packet
transmission scheduling, and the path loss and � 2 ,p s iINT values are used for admission
control.
3.3.2 Admission Control
Admission control must pledge that all connections receive their required SQO
throughout their lifetime. Therefore, it must consider the long-term channel and
interference conditions, not only the conditions when the connection request is received.
In addition, the admission control should consider theQoS , not only for the link
requesting for admission, but also for all existing links.
It is considered that the position information of all the stations in the CRN is
available at the controller and use jig~ to represent the path loss between the transmitter
of link j and the receiver of link i.
.
The path loss between the transmitters of link i and the BS, denoted as ipsg ,2~ , can
also be obtained based on distance between the station and the BS. The admission
control decisions based on such static information can tolerate random changes in the
incoming traffic, measurement errors for the link gains and MS locations.
Assume there are N−1 existing links in the CRN. When a new connection request
arrives with the required rate newreqR , , the request is sent to the controller, which assumes
that the new request is the th
N link in the system and solves the following optimization
problem.
61
NR~
max (3.21)
such that,
(3.22)
If the new request is a streaming connection, then
2 ,1
N
s p i thiPg INT
=≤∑ �� , (3.23)
if the new request is nrt connection, then
2 , ,1
N
i s p i th dataiPg INT
=≤∑ �� , (3.24)
and,
11
~,, −≤≤= NiRR ireqi (3.25)
NiPPi ≤≤≤≤ 1
~0 max, (3.26)
Where the unknowns are NR~
and iP~
for i = 1....., N, iP~
is the estimated transmission
power for the transmitter of link i, andthdatath INTINT ≤
,. After solving the problem,
if newreqN RR ,
~≥ , the new request is accepted as the thN existing link, and newreqNreq RR ,, = .
Otherwise, the user’s request is rejected. Different interference thresholds are set
depending on the new request is streaming or nrt traffic. When ,th data thINT INT= , the nrt
data requests have the same priority as the streaming traffic in admission control; when
datathINT , is smaller, fewer nrt data requests can be accepted, leaving more resources for
potential streaming traffic requests.
�( )2 ,
i ii
N
p s ii i iij i
WP gγ ,1 i N
R P g INT η
∗
≠
≥ ≤ ≤+ +∑� �
� � �
62
Outage performance for the streaming traffic is not explicitly considered in the
admission control. However, the random changes of link gains and interference
conditions provide the packet transmission scheduling with opportunities to achieve low
outage probability.
When the primary user’s activity is comparatively stationary, it is probable for the
CR stations to measure the experienced interference levels from the primary network
and predict the future resource availability with high certainty. Therefore, the QoS of
admitted connections can be well guaranteed for a long term. On the other hand, if there
is any abrupt change in the primary network transmission activity, QoS in the CRN may
be violated and some admitted connections may have to be dropped. For example, when
a new connection is generated at the primary network, especially the transmitter is close
to the receiver of an existing secondary link, the primary-to-secondary interference
experienced by the secondary receiver can increase significantly, both instantaneously
and over a long term.
The controller periodically revaluates the resource availability at the CRN based on
the updated interference levels. It checks the feasibility of all constraints in the above
optimization problem. If any of the constraints cannot be satisfied, a certain criterion is
used to remove one or more connections from the system until all the constraints can be
satisfied for the remaining connections.
Three criteria for connection dropping are considered,
• The most recently admitted connection is dropped first.
• The connection with the highest 2 ,p s iINT is dropped first.
• The connection with the smallest estimated weight is dropped first. For the
weight- based dropping criterion, an estimated weight ( iω� ) is defined for each
connection as follows
�*
2 ,2 , ( )
iii
p s is p i i
gW
g X INTω
γ η=
+
��
� (3.27)
63
In designing these dropping criteria, we do not differentiate links with different type
of the traffic, but this can be an alternative in practical systems.
3.4 Packet Transmission Scheduling
If scheduling process is considered, the first goal is to reduce the session outage for the
streaming links and guarantee the average throughput for the nrt links. This will be
presented in subsections 3.4.1 and 3.4.2. The scheduling should then try to utilize the
available resources in a fair and efficient way.
We propose two scheduling methods, an optimum scheduling in subsection 3.4.3
and a weight-based scheduling in subsection 3.4.4. The former is an ideal method
assuming global information about the link and interference conditions is available, and
the latter is a heuristic scheme based on measured link and interference conditions at the
CR nodes.
3.4.1 Session Outage Probability Reduction for Streaming Traffic
For the streaming traffic, the session outage should be reduced. The minimum
transmission rate at time slot t to keep the session running for the entire duration of the
time slot is given by,
( )( )
,min,
1s i
req ii t
s
T B tR R
T
− −= (3.28)
Where, ( )1−tBi indicates how much time the buffered data at the receiver can
keep the session running without a session outage, if the link will not have a chance to
transmit after time slot t−1. Because of the constant replay rate, the transmitter has full
knowledge about ( )1−tBi . S is a set of streaming links and S’ is defined as,
S’ = {i|Bi (t−1) < Ts, i ∈ S},
Where, sT is the duration of one time slot. Links in S’ should have a higher priority
to transmit at time slot t in order to avoid immediate session outage. Alternatively, the
limited buffer size at the receiver limits the maximum transmission rate. The maximum
transmission rate for link i at time slot t is given by
64
max,
, ,
( 1)( )
i i s
buf i req i
s
B B t TR t R
T
− − += (3.29)
3.4.2. To achieve average throughput for nrt traffic
For the nrt links, the objective is to maximize the resource utilization while satisfying the
average throughput requirements. In the interim, it is noticed, that in order to maximize
the resource utilization, the average throughput can be achieved over a longer period of
time for links with poorer channel and interference conditions.
To have some control over the time scale over which the average throughput is
achieved for the nrt traffic, we define a counter, ( )i tξ , for each nrt link. The initial value
of ( )i tξ is set zero and it is updated at the end of every time slot. It is increased by one if
link i does not transmit at a time slot, and decreased by ,
( )i
req i
R t
R
if the link transmits at
rate ( )iR t .
Define Z ’= {iǀ ξi (t − 1) >ξmax,i∈ Z ’}.
Links inZ ’ have a higher priority to transmit in the next time slot. Instinctively, a
smaller ξmax helps to achieve the average throughput over a shorter period of time, but
may force the links to transmit more frequently than necessary, and this can easily result
in that some links transmit when their transmission conditions are poor.
3.4.3 Optimum scheduling
Let S′′ be a set of the streaming links with Bi (t − 1) ≥ Ts, and Z
′′ be a set of the nrt
links with ξi(t − 1) ≤ ξmax. The objective of the scheduling for links in S ′′ and Z
′′ is to
improve the resource utilization, while keeping fairness among the links. Links with
good transmission conditions should transmit at higher rates in order to efficiently utilize
the available resources.
Meanwhile, other links with poorer transmission conditions should also have
chances to transmit in order to avoid very low rate over a long period of time. The idea of
65
packet scheduling with long term PF can achieve this objective. In this work the
scheduling schemes are reconsidered by jointly considering the streaming traffic and the
nrt traffic.
Let )(, tR iopt and )(. tP iopt , respectively, be the transmission rate and power for link i.
The optimum scheduling scheme is formulated as follows:
,
1
( )max
( )
N opt i
ii
R t
X t=∑ (3.30)
such that, transmission rate and power for link i should satisfy,
( ) ( )
( ) ( ) ( ) ( )
, *
, , 2 ,
1
, 1,2,....,opt i ii
N
opt i opt i ii p s i
j
WP t g ti N
R t P t g t INT t
γ
η≠
≥ =
+ + ∑
(3.31)
transmission rate for link i should satisfy following conditions.
, ( ) 0, 1, 2... ,opt iR t i N> = (3.32)
(3.33)
(3.34)
(3.35)
(3.36)
(3.37)
, ,( ) ( ),o p t i b u f iR t R t i S≤ ∈
'
, m in ,( ) ( ) ,o p t i i
R t R t i S≥ ∈
, m ax,0 ( ) 1, 2... ,opt iP t P i N≤ ≤ =
, 2 , max 2 ,' '( ) ( ) ( ),
opt i s p i j s p ii D j SP t g t INT P g t
∈ ∈≤ −∑ ∑
' , 2 ,( ) ( )opt i s p i thi SP t g t IN T
∈≤∑
66
(3.38)
( ) ( ) ( )'' , 2 , max 2 ,'opt i s p i j s p ii D j S DP t g t INT p g t
∈ ∈ ∪≤ −∑ ∑ (3.39)
If ( )tX i represents a smoothed average rate in the past. It is defined as,
( ) ( ) ( ) ( ),1 1ii c c opt iX t X t R tα α+ = + − (3.40)
Where, 0 1cα< < , for special case, when t=0
The constraints (3.32)-(3.34) are for the transmission rates, and the constraints
(3.36)-(3.39) give different priorities of the links by setting different interference
thresholds. The optimization problem is solved first using constraints (3.31)-(3.34) and
(3.36) for all the links in S.’
After having the results, it is then solved using constraints
(3.31)-(3.36) and (3.36) for all the links in S.”
After the scheduling is done for all the streaming links, it is solved using
constraints (3.31)-(3.34) and (3.38) for all the links in Z’’, and finally solved using
constraints (3.31)-(3.34) and (3.39) for all the links in Z’’. In this manner, the streaming
links in S’ have the highest priority, followed by the streaming links in S
’’, and the links in
Z’’
have the lowest priority. If there is no feasible solution, link k is temporarily removed,
where, ( )1maxarg ',
−=∈
tBk iSii,and the optimization problem is solved again. This is
repeated until a feasible solution is obtained. The optimization problem is nonlinear and
nonconvex [54].
3.4.4 Weight-based scheduling
To design a scheduling scheme that can be implemented in a practical system, following
steps are taken.
', 2 , max 2 ,"( ) ( ) ( ),opt i s p i j s p ii S j S D
P t g t INT P g t∈ ∈ ∪
≤ −∑ ∑
67
Define ( )tLi as the priority level of link i at time slot t. A larger value of ( )tLi
indicates a lower transmission priority for the link. A link with ( )tLi = 1 has the highest
priority to transmit, and a link with ( )tLi = ∞ cannot transmit at time slot t.
' ' ' ' '' '' '' ''
__________________________________________________________________________
lg
__________________________________________________________________________
1: , , ,S d S d
A orithm
Define A S A A S and A= = = =Z Z�
( )
( )
( ) { }
( )
( ) { }
,
,
'
'
'
'
''
. 0,
2 :
3 : arg min 1
4 : 1, , \
5 :
6 :
7 : arg min 1
8 : 1, , \
9 :
10 :
S
d
i
S
kk A
i s s
S
kk A
i s s
S
Initialize C and
L t for all link
While A is not empty do
i B t
C C L t C and A A i
end while
While A is not empty do
i t
C C L t C and A A i
end while
While A i
ξ
∈
∈
=
= ∞
= −
= + = =
= −
= + = =
( )
( ) { }
( )
( ) { }
''
''
'' ''
'' ''
11: arg min
12 : 1, , \
13 :
14 :
15 : arg max
16 : 1, , \
17 :
_______________________________________________
s
d
kk A
i s s
d
kk A
i d d
s not empty do
i w t
C C L t C and A A i
endwhile
while A is not empty do
i w t
C C L t C and A A i
end while
∈
∈
=
= + = =
=
= + = =
____________________________
Algorithm is used to find the priority levels. The streaming links with
( ) si TtB <−1 have the highest priority, and a higher priority is given to the link with
smaller ( )1−tBi . After that, a higher priority is given to the nrt links with ( ) max1 ξξ >−t
and a higher priority is given to the link with larger ( )1−tξ . For each of the other links,
the streaming links have a higher priority than the nrt links. The weight ( )twi for link i at
time slot t is given by,
68
(3.41)
This weight in (3.41) is calculated based on the link and interference conditions at
time slot t−1. Each CR transmitter calculates the weight at the end of time slot t − 1 and
passes it to the controller. At the end of each time slot, the controller collects
information from the CR nodes and calculates the transmission priority of each link for
the next time slot using algorithm 1. For the streaming links with ( ) si TtB <−1 and the
nrt links with ( ) max1 ξξ >−ti , the CR transmitters pass the respective ( )1−tBi and
( )1−tiξ values to the controller at the end of the time slot t − 1.
For other links, the transmitters can pass their calculated ( )twi values to the
controller, which performs algorithm and informs the CR transmitters of their priority
levels through the control channel before the start of time slot t. After performing
Algorithm, transmission power for the links will be found.
Define,
( ) ( )( ), 2 ,, , ( )j i
th rem th j s p jall j L t L tINT INT P t g t
<= − ∑ (3.42)
As the amount of the interference that the BS can still tolerate after the transmissions
of all the CR links with higher priority. The maximum transmission power of a nrt link i
is then given by
( )( )
=tg
INTPtP
ips
remth
i
,2
,
max,min (3.43)
2 ,2 ,
( 1)( ) .
( 1)( ) ( 1)
iii
s p ii p s i
g tWw t
g tX t IN T tγ η∗
−=
−− +
69
So as to find the updated max,remINT for each link, the lower priority link should
transmit slightly later. It can be achieved by backoff process. The transmitter of link i
sets its backoff timer to be ( )δtLi , where δ is equal to the maximum round trip time
required for a signal to propagate between an MS in the CRN and the controller. At the
time when the transmitter of link i completes the backoff, the controller measures the
interference level, updates remthINT , , and broadcasts it in the CR network. Upon
completing the backoff, the CR transmitter calculates its transmission power based on
the current remthINT , . Lower priority links cannot transmit if remthINT , = 0 when they
complete the backoff.
It must be noted that the total amount of time for backoff, denoted as hT , is limited.
Therefore, the number of transmitting links in a given time slot is limited to δhT , and a
link with ( ) δhi TtL > cannot transmit in the current time slot. Instead of having the
MSs perform the backoff and calculate the transmission power, the CR transmitters can
also pass values of ( )1,2 −tg ips to the controller, which calculates the transmission
power for each CR transmitter and passes the values back to the transmitters at the
beginning of the time slot t.
In either way, the controller plays an important role because the transmission power
at the CRN is limited by the interference level that is measured at the controller. Instead
of using the backoff process to select the links that are allowed to transmit, an
optimization problem can be formulated and find the optimum δhT links. Still this
requires the controller to collect information about the mutual interference among the
links, which is practically difficult and therefore not considered here.
In addition, the concept of exclusive region is adopted to prevent strong mutual
interference among the peer links. A link i in S′′ or Z ′′ cannot transmit if it is within the
exclusive region of any link with a higher priority. The actual transmission power for
each link is limited by the maximum transmission power of the transmitter node. For a
streaming link, the transmission power is also dependent on ( )tR ibuf , , which is the
70
maximum rate limited by the receiver buffer space. The required transmission power for
achieving ( )tR ibuf , is estimated as
(3.44)
where mutual interference from peer links is ignored due to the use of exclusive
regions which prevent strong mutual interference among the peer links. By taking into
consideration the interference threshold and the maximum transmission power of the
node, the actual transmission power is given by
(3.45)
It is noted that mutual interference cannot be completely eliminated by
using the exclusive regions. Given the transmission power of each node, the actual
transmission rate is obtained through rate adaptation so as to meet the SINR
requirements. Mathematically, the transmission rate for link i can be found from
(3.46).Assume that, the equality holds if the actual SINR is equal to *γ then, the
scheduling decisions strongly depend on the interference levels at the primary BS as
shown in Equation (3.46).
(3.46)
In a practical system, the measured interference levels may not be accurate. As a
result, the actual interference at the BS caused by the secondary transmissions, denoted
byme
INT , may exceed thINT . When this occurs, the controller notifies all the transmitting
( )( )
( )
= tPtg
INTPtP ibuf
ips
remth
i ,
,2
,
max, ,min
2 ,1 ,
( ) ( )( )
( ) ( ) ( )
i i ii N
j j i p s ij j i
W P t g tR t
P t g t I N T tγ η∗
= ≠
= + + ∑
, 2 ,
,
( ) ( )( ) ,
( )
b u f i p s i
b u f i
i i
R t I N T tP t
W g t
γ η∗ + =
71
MSs to reduce their transmission power to ( ) thi
me
INTP t
INT× . Furthermore, the allowed
interference threshold can be reduced toth
INTβ× , where β is between 0 and 1, in order
to accommodate potential measurement errors. When β is smaller, more measurement
errors can be tolerated in the scheduling.
3.5 Simulation Result and Discussion
3.5.1 Optimum Scheduling (Slot –by- Slot) in CRNs
Table 3.1: Simulation Parameters for Optimum Scheduling
Transmission rate of each primary link PR 64kbps
SINR threshold γp, γ
s 5dB
Path loss exponent α 4
Standard deviation of log-normal fading σ 4dB
Maximum MS transmission power max,pP and max,sP 0.5W
Spread spectrum bandwidth, W 5 MHz
Reference distance 0d 10 m
Reference base station coverageZ 1000m
Ad hoc transmission range of MSs 300m
Simulation parameters are shown in Table 3.1. The results based on the first
interference model are discussed and then these results are compared based on the two
Interference model.
72
10-10
10-9
10-8
10-7
10-6
10-5
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5x 10
5
INTth
(W)
Secondary
lin
k tra
nsm
issio
n rate
(bps)
Secondary link transmission speed: Equal Speed Allotment
Np=5&Ns=10
Np=10&Ns=5
Np=5&Ns=5
Figure 3.3: Secondary link transmission rate: ESA
Table 3.2: Secondary link transmission rate: ESA, for different values of PN and sN
Np=5&Ns=10 Np=10&Ns=5 Np=5&Ns=5
20000 130000 220000
38000 158000 280000
50000 170000 390000
60000 180000 330000
60000 180000 350000
60000 180000 360000
60000 180000 360000
60000 180000 360000
60000 180000 360000
60000 180000 360000
Figures. 3.3 and 3.4 shows that when the interference threshold is below a
certain value, the secondary link transmission rate increases with the interference
threshold for both Equql Speed Allotment(ESA) and Propotional Speed Allotment
73
(PSA).
10-10
10-9
10-8
10-7
10-6
10-5
0
0.5
1
1.5
2
2.5
3
3.5x 10
5
INTth
(W)
Secondary
lin
k tra
nsm
issio
n rate
(bps)
Secondary link transmission speed: Proportional Speed Allotment
Np=10&Ns=5
Np=5&Ns=10
Np=5&Ns=5
Figure 3.4. Secondary link transmission rate: PSA
Addionally increasing the interference threshold does not affect the secondary
transmission rate anymore, since the transmission of the secondary links is limited by the
primary link’s SINR constraint and the mutual interference between primary and
secondary networks.
When the secondary users increase their transmission power to achieve higher
rates, the increased power generates higher interference to the primary networks. To
sustain the SINR, the primary links will increase their power too. Once the maximized
interference limit at the primary receiver is reached, the secondary users cannot further
increase their transmission rate even if the interference threshold is not reached.
It is also seen from both Figures 3.3 and 3.4 that increasing the number of the
primary or secondary links results in lower secondary link rate as more links are
competing for the network resources.
74
Table 3.3. Secondary link transmission rate : PSA
Np=10&Ns=5 Np=5&Ns=10 Np=5&Ns5
5000 6500 135000
10000 13500 180000
15000 18500 265000
17000 20500 320000
17000 20500 325000
17000 20500 325000
17000 20500 325000
17000 20500 325000
17000 20500 325000
17000 20500 325000
17000 20500 325000
5 10 15 200
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
No. of secondary links
Pdiff of prim
ary
lin
ks
Mp=20
Mp=15
Mp=10
Mp=5
Figure 3.5: Power difference ( pdiff ) of primary links: ESA
After comparing two figures, it is observed that PSA can achieve a lot higher
transmission rate for the secondary links than using ESA, since the former can take better
advantage of good channel conditions of individual links (Table 3.3). If secondary link is
not used, transmission power of the primary links is increased due to extra interference
75
from the secondary links. pdiff is defined as the difference of the average transmission
power of the primary links with and without the secondary transmissions.
Figure. 3.5 (Table 3.4) shows that with the increase of the number of secondary
links, the average primary transmission power of the primary links decreases. This is
explained by the dramatic decrease of the secondary link transmission rate as the number
of secondary links increases.
Table 3.4: Power difference of primary links: ESA
MP=20 MP=15 MP=10 MP=5
0.13 0.165 0.2 0.25
0.1 0.12 0.13 0.16
0.05 0.06 0.08 0.11
0.02 0.025 0.06 0.08
Figure 3.6 :Power difference of primary links:PSA
Because of the equal rate allocation, the transmission rate of the secondary links is
5 10 15 200
1
2
3
4
5
6
7
8
9
10
No. of secondary links
Pdiff of prim
ary
lin
ks
power difference of primary links: PSA
Mp=5
Mp=10
Mp=15
Mp=20
76
limited by the link with the worst link condition. ( Table 3.5).
Figure 3.6 shows that, pdiff of the primary links decrease with the number of
secondary links for PSA.With the same amount of pdiff , the CRN based on PSA can
achieve much higher transmission rate than that based on ESA.
Table 3.5: Power difference for diff. Values of MP
Figure 3.7- 3.9 shows ,maxpINT versus different system parameters. The actual
average interference at the primary link is also shown for ESA and PSA, respectively
200 400 600 800 1000 1200 1400 1600 1800 200010
-10
10-9
10-8
10-7
10-6
10-5
10-4
Dmax
(m)
Inte
rfere
nce to P
rim
ary
Lin
ks(W
)
Maximum interference vs. cell size
PSA
ESA
INTp,max
Figure 3.7. Highest interference vs. Cell size
It is seen that a primary network with a smaller cell size, larger max,pP or larger
Mp=20 Mp=15 Mp=10 Mp=5
2.5 1.8 1.3 0.8
2.2 1.6 1.2 0.6
2.1 1.5 1.1 0.5
2.0 1.4 1.0 0.4
77
fading deviation may tolerate a larger interference threshold for the same number of
secondary links ( Table 3.6, 3.7 and 3.8).
Table 3.6: maxINT For PSA and ESA related to cell size
Most importantly, the figures show that PSA can cause significantly lower
interference to the primary links than ESA. With ESA, the link with the worst SINR
condition may transmit much higher power than other links. When this link is close to the
primary link receiver (the BS), the primary link transmitters should transmit very high
power to combat the high interference.
Therefore, in ESA, it is more likely that the primary link transmitters reach the
maximum transmission power limit. This also explains that the interference level at the
BS using ESA is almost the same as,maxp
INT . In contrast, PSA does not encourage the
secondary links with poor SINR to transmit as high rate as the links with good SINR, and
therefore the interference level at the primary links is limited by the maximum
transmission power of both the primary and secondary transmitters.
PSA ESA maxINT
1.00E05 2.80E05 3.00E05
2.50E07 3.00E06 3.10E06
1.50E08 3.00E07 3.20E07
3.00E09 7.00E08 9.00E08
8.00E10 1.30E08 1.80E08
78
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.5
1
1.5
2
2.5
3
3.5
4
4.5x 10
-7
Pmax
(W)
Inte
rfere
nce t
o p
rim
ary
lin
ks
Maximum interference vs Pp,max
PSA
ESA
INTp,max
Figure 3.8: Maximum interference vs. max,pP
Table 3.7:,maxp
INT for PSA and ESA related to max,pP
ESA PSA ,maxpINT
1.00E08 1.20E07 1.26E07
1.10E08 1.60E07 1.63E07
1.20E08 2.00E07 2.11E07
1.30E08 2.50E07 2.50E07
1.60E08 3.00E07 2.90E07
1.80E08 3.50E07 3.3507
2.20E08 3.00E07 3.80E07
79
2 3 4 5 6 7 80
0.5
1
1.5
2
2.5
3
3.5
4x 10
-7
sigma(db)
Inte
rfere
nce to P
rim
ary
Lin
ks(W
)
Maximum interference vs sigma
PSA
ESA
INTp,max
Figure 3.9 :Maximum interference vs. Sigma
Table 3.8: ,maxp
INT for PSA and ESA
PSA ESA ,maxpINT
1.00E08 6.00E08 6.60E08
8.00E09 9.50E08 9.50E08
1.10E08 1.30E07 1.28E07
1.90E08 1.60E07 1.62E07
3.00E08 2.00E07 2.10E07
6.00E08 2.70E07 2.75E07
1.00E07 3.70E07 3.80E07
Transmission rate of the secondary links is shown in Fig. 3.10 based on the two
interference models defined in (3.3) and (3.4).
The difference is more obvious when thINT is small and close to the noise power
level. When thINT is relatively large, the two interference models do not make obvious
difference to the secondary transmission rate (Table 3.9).
80
10-10
10-9
0
1
2
3
4
5
6
7
8
9
10x 10
7
INTth
(W)
Secondary
lin
k tra
nsm
issio
n rate
(bps)
Comparision of two interference models
Interference Model I
Interference Model II
Figure 3.10: Comparison of two interference models
Table 3.9: Comparison of interference model 1 and 2.
INTth Interference Model I Interference Model II
1.00E-10 0 0 31000000 50000000
2.00E-10 32000000 31000000 32000000 65000000
3.00E-10 35000000 70000000 38000000 75000000
3.00E-10 51000000 79000000 52000000 80000000
7.00E-10 60000000 86000000 60500000 87000000
1.00E-09 61000000 88000000 61500000 89000000
When thINT is small, using the second interference model achieves much higher
transmission rate than using the first interference model, since in the latter case, noise
and interference from the primary network can dominate the interference threshold,
while in first interference model, the secondary transmissions can take advantage of all
the interference allowed by thINT .
81
As thINT increases, the secondary transmission power increases and eventually
is limited by their mutual interference and SINR constraints, but not by thINT . At this
point, the two interference models result in about the same transmission rate at the
secondary network.
3.5.2. Joint Admission Control and Packet Transmission Scheduling in
CRNs
Simulation parameters are listed in Table 3.10.
Table 3.10: Simulation Parameters of Joint Admission Control
Parameter Value
Cell radius 1000 m
Path loss exponent 4
Standard deviation of log-normal fading 8 dB
Spread spectrum bandwidth ,w 5 MHz
SINR threshold *γ (primary and secondary) 5dB
Background AWGN noise powerη 1010− W
Number of primary links 10 Transmission rate of primary link 32 Kbps
Maximum transmission power of primary MS 0.5 W
Maximum transmission power of secondary MS maxP 0.05 W
Exclusive region size (radius) 100 m
Interference threshold thINT 1210− W
Secondary user transmission range 100 m
Streaming connection arrival rate 100 requests/sec nrt data connection arrival rate 100 requests/sec 100 requests/sec
Average streaming connection duration 60 sec
Average nrt connection duration 120 sec
Data packet arrival rate Time slot duration sT 5ms 2.5 packets/sec
Overhead time per time slot hT 0.5 ms
82
For the CRN, two independent Poisson processes are used to generate the
connection requests for the streaming traffic and nrt data traffic, respectively. The new
connection is randomly (with equal possibility) assigned to one of the transmitter-receiver
pairs that do not have active connections at the time. The duration of each admitted
connection follows an exponential distribution. For each admitted nrt connection, data
packets each with a fixed size are generated using another Poisson process.
Figure 3.11 demonstrates that setting different values for datathINT , can change the
relative amounts of the streaming and nrt traffic admitted into the system. When
datathINT , =0, all the system resource is used for the streaming traffic.
As datathINT , increases, more nrt data connections and fewer streaming
connections can be admitted into the system. [61] (Table 3.11)
.
0 10 20 30 40 50 60 70 80 900
10
20
30
40
50
60
70
80
INTth,data
/INTth
(%)
Num
ber
of
adm
itte
d c
onnections
Effect of INTth,data
on amount of admitted traffic
streaming traffic,Rreq,i
=100kbps
nrt traffc,Rmin,i
=20kbps
Figure 3.11: Effect of ,th dataINT on amount of admitted traffic
83
Table 3.11: Streaming and nrt traffic
600 700 800 900 1000 1100 1200 1300 1400 15000
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Average Duration (time slots),INTth
,data
/INTth
=30%(K=3)
P(R
av
r, dur/R
min
<1)
Cumlative avrage data throughput
VC=900
VC=600
VC=300
Figure 3.12: Cumulative average data throughput
Figure 3.12 shows performance of the nrt traffic where iavrR , in the vertical axis
is the cumulative average throughput of an admitted nrt connection over the time
interval indicated in the horizontal axis. (Table 3.12)
Streaming
traffic nrt traffic
20 0
18 18
16.5 22
15 28
13 35
11 32
10.5 53
10 63
9 70
8.9 78
84
Table 3.12: Different values of VC
It is shown that when the time interval is sufficiently long (about 1500 time slots
with the given parameters), all connections can receive their required average
throughput.
100 200 300 400 500 600 700 800 9000
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2x 10
7
Virtual counter limit (time slots) INTth,data
/INTth
=30% (K=3)
Tra
nsm
issio
n e
ffic
iency(b
its p
er
slo
t)
Virtual
Distributed
Figure 3.13: Throughput and delay performance of nrt traffic
VC=900 VC=600 VC=300
0.3 0.39 0.51
0.23 0.3 0.37
0.16 0.22 0.26
0.1 0.15 0.19
0.07 0.1 0.125
0.05 0.07 0.09
0.03 0.05 0.065
0.02 0.03 0.035
0.01 0.02 0.035
0 0.01 0.015
85
Figure.3.13 shows performance of the nrt traffic when using the weight-based
scheduling, where iavrR , in the vertical axis is the cumulative average throughput of an
admitted nrt connection over the time interval indicated in the horizontal axis. It is
shown that when the time interval is sufficiently long (about 1500 time slots with the
given parameters), all connections can receive their required average throughput.
Within a short period of time, using smaller maxξ may result in smoother rate
changes since it forces the nrt connections to transmit more frequently. On the other
hand, setting a larger value for maxξ may benefit the system resource utilization. Within
a short period of time, using smaller maxξ may result in smoother rate changes since it
forces the nrt connections to transmit more frequently. The packet transmission delay
and throughput performance of the admitted nrt traffic is evaluated in figure 3.13. The
figure shows that the performance of the optimum scheduling and the weight-based
scheduling is quite close, in terms of both throughput and delay. This shows the
effectiveness of the weight-based scheduling. (Table 3.13)
Table 3.13: Comparison between Virtual and Distributed values
The connection dropping rates are shown in figure 3.14. It shows that as maxξ
increases, the throughput increases, while data transmission delays increases too. This
shows the tradeoff between the resource utilization and the latency (refer Table 3.14). It
is seen that it is possible to achieve very low connection dropping rate.
Virtual Distributed
2000000 2200000
3300000 3500000
3300000 3500000
5300000 5500000
6300000 6300000
7100000 7300000
8000000 8300000
8500000 8900000
9000000 9300000
86
Figure 3.14: Connection dropping rate
The criterion that drops the most recent admitted connection, does not consider the
interference condition and therefore results in the highest dropping rate, and the criterion
based on primary to secondary interference achieves drop rate performance between the
other two criteria. (Table 3.14)
Table 3.14: Connection dropping rate for different values of M
M=5,smallest
Weight
M=15,smallest
weight
M=5,worst
gain
M=15,worst
gain
M=5,recently
accepted
M=5,recently
accepted
0 0 0.0001 0.0005 0.00058 0.0006
0.0001 0.0001 0.0002 0.0008 0.001 0.0012
0.0002 0.0002 0.00029 0.0011 0.00135 0.0019
0.00028 0.00035 0.00039 0.00155 0.00199 0.0025
0.00038 0.00055 0.0007 0.0021 0.0026 0.0032
0.0007 0.00082 0.00098 0.0026 0.0033 0.003
0.00087 0.00115 0.0015 0.0032 0.0031 0.0038
0.0012 0.00138 0.0021 0.0038 0.0038 0.0058
0.0016 0.0019 0.0028 0.0036 0.0058 0.0072
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
1
2
3
4
5
6
7
8x 10
-3
Probability from "off" to "on"
Connection d
ropin
g r
ate
Connection droping rate
M=5,Drop link with smallest weight
M=15,Drop link with smallest weight
M=5,Drop link with worst link gain
M=15,Drop link with worst link gain
M=5,Drop most recently accepted
M=15,Drop most recently accepted
87
3.6. Conclusion
Interference model for scheduling algorithm in CDMA based CNR is designed The
actual average interference at the primary link is also shown for ESA and PSA,
respectively. Results indicate that there is a limit on the interference threshold beyond
which the secondary link transmission rate cannot be increased by increasing the
interference threshold. Also, the increase of the secondary user transmission rate will
consume additional power from the primary user, and the same amount of power increase
from the primary users can support higher rate of the secondary links using proportional
rate allocation, compared to using equal rate allocation among the secondary links.
In contrast, PSA does not encourage the secondary links with poor SINR to
transmit as high rate as the links with good SINR, and therefore the interference level at
the primary links is limited by the maximum transmission power of both the primary and
secondary transmitters.
A throughput computation scheme for real-time data traffic is proposed and
scheduling schemes for each type of the traffic are designed. By using static channel and
interference conditions, the admission control scheme effectively limits amount of the
traffic admitted into the system to be below capacity. The heuristic scheduling scheme
opportunistically takes gain of the random channel and interference conditions and
achieves low and fair session outage probability for the streaming traffic and high
throughput and fair transmission delay for the data traffic, and its performance is close to
the optimum scheduling. In this way, analysis of the optimum power and rate
allocations in a CDMA-based cognitive wireless network with spectrum underlay is done.
Comparison of two interference models is discussed.
Our analysis indicates that there is a limit on the interference threshold beyond
which the secondary link transmission rate cannot be increased by increasing the
interference threshold. Furthermore, the increase of the secondary user transmission rate
will consume extra power from the primary user, and the same amount of power increase
from the primary users can support higher rate of the secondary links using proportional
speed allocation (PSA), compared to equal speed allocation (ESA) among the secondary
links.