Documente Academic
Documente Profesional
Documente Cultură
&RJQLWLYH
5DGLRV
Shridhar Mubaraq Mishra
Jana van Greunen
May 13th, 2004
Page 1
:LQGRZRI2SSRUWXQLW\
Recent measurements by the
FCC in the US show 70% of the
allocated spectrum is not utilized
Frequency (Hz)
Time (min)
Page 2
:KDWDUH&RJQLWLYH5DGLRV"
Definition: A cognitive radio (CR) is a radio that can change its transmitter
parameters based on interaction with the environment in which is operates [FCC NPRM - 03-322]
Page 3
6\VWHP0RGHO
Centralized system model with Access
Point
CR1
CR system functions:
1. Sensing
2. Reporting
3. Channel Sounding
4. Channel Allocation
AP
Bluetooth
Cordless phone
c
ion
mi ct
na ele
D y cy S
n
ue
eq
Fr
CR2
Primary
users
CR3
Page 4
3UREOHP6WDWHPHQW
N users
K channels (K >> N)
Channel gains for user (i), : [||hi1||2 ||hi2||2 ||hiK||2]
Power allocated for user (i) : [Pi1 Pi2 PiK]
Optional: requested rates : [R1 R2 RN ]
Goal: minimize total power P, more formally:
Page 5
2XU$SSURDFK
Existing channel allocation framework
What is the solution space?
Are there low complexity algorithms for one user per channel?
How do user rates affect minimum power?
Page 6
6ROXWLRQ6SDFH&KDUDFWHUL]DWLRQ
If no rate constraints but sum capacity constraint then:
Dual of problem in lecture!
Optimal solution is no sharing (Tse98)
Page 7
$QH[DPSOH
Users
Channels
C1
U1
C2
Parameters:
N = 2,
K = 4,
Csum = 2 bits/s/Hz,
Rreq1 = Rreq2 = 1 bits/s/Hz
Channel Gain Matrix (H):
C3
U2
C4
Page 8
([DPSOH&RQWLQXHG
Power per Channel
Optimal:
P* = 0.4344 W,
R1 = 1.5832 bits/s/Hz,
R2 = 0.4168 bits/s/Hz,
0. 1 8
0. 1 6
0. 1 4
0. 1 2
0.1
0. 0 8
0. 0 6
0. 0 4
0. 0 2
Optimal sharing:
Ps = 0.4467 W
Power (W)
0. 18
0. 16
User 1
0. 14
0. 12
0.1
0. 08
User 2
0. 06
0. 04
0. 02
0
0. 1 8
0. 1 6
0. 1 4
Optimal no sharing:
Pnos = 0.4534 W
0. 1 2
0.1
0. 0 8
0. 0 6
0. 0 4
0. 0 2
0
1
Channel Number
Page 9
5HGXFLQJ&RPSOH[LW\WKH*UHHG\$SSURDFK
Two step approach for one user per channel
Channel allocation
C1
U1
4
C2
4.5
U1 -> C2
end for
end while
4
C4
U2 -> C4
U1
C3
U2
3
U2
C1
1
3.5
C3
U1 -> C3
U2 -> C1
3RZHU$OORFDWLRQDVD)XQFWLRQRI5HTXHVWHG5DWHV
Optimal power
algorithm is quick &
easy but R* wont
meet requested rates
Estimate performance
for non-optimal with
requested rate R from
optimal algorithm
Question: How does
Rreq R* influence
minimum power?
It is polynomial in ||RR*||
Page 11
Page 12
&KDUDFWHUL]LQJ&KDQQHO*DLQIRU&RJQLWLYH5DGLRV
Spectrum usage in (0, 2.5) GHz
-40
Cell
-45
-50
-55
PCS
TV bands
-60
-65
-70
-75
-80
-85
-90
0
0.5
1.5
Frequency (Hz)
2.5
9
x 10
Page 13
0DQDJLQJ,QWHUIHUHQFHVHHQE\3ULPDU\8VHUV
Two extreme approaches:
Aggressive approach: Allocate the channel even if subset of CRs
see Primary User (PU) => Might cause interference to PU
Conservative Approach: If any CR detects a Primary User, dont
use that channel => CR system might need more power for same
requested rates
Page 14
7UDGHRII5HVXOWV
Power (Ratio)
Threshold
Key:
Interference to PU
Page 15
6XPPDU\
Contributions:
1. Developed a O(K3) complexity algorithm for one user per
channel allocation
2. Developed model to estimate power penalty from R* given
requested rates R
3. Modeled channel gain matrix for CRs consistent with real data
4. Analyzed the effect of various thresholds on power penalty and
interference to PUs
Page 16
)XWXUH:RUN
Future Work (we plan to do before May 24th)
In a real system power assignment per channel is a function of
sensing radios sensitivity
Highly sensitive radios can allocate more power (Pik
sensitivity)
Page 17
&RQFOXVLRQV
Wireless Overload !!
Page 18
&RQFOXVLRQV
Backup slides
Page 19
*UHHG\962SWLPDOZLWK6KDULQJ
Power per Channel
3. 5
N = 6, K = 8
3
Power (W)
2. 5
2
Op t i ma l s h a r i ng
Gr e e d y
1. 5
0. 5
0
1
Matrix Number
Page 20