Sei sulla pagina 1di 16

Spectrum sensing in Cognitive

Radio
Sensing the primary transmitter or primary receiver

(Libin Jiang, Assane Gueye)


EE224B, Spring 06

Outline
PT-sensing problem with shadowing
Non-Talk-Zone

Cooperation as an improvement
Examples of protocols
PR-sensing
Comparison

PT-sensing: Radius of Non-TalkZone


Received power with only path loss

Non-Talk-Zone

Received power with shadowing

Required Po for same threshold

Thus new Non-Talk-Zone


(as if we were not shadowed)
CR cannot distinguish this case from
shadowing

Radius of the NTZ

Cooperation reduces the radius of


the NTZ
I am shadowed but my friend might
be not: can she helps me?

By cooperating, I can detect the presence


of the PT even if I am shadowed
Request-NACK protocol

NTZ radius is now:

Cooperation gain as a function of


the number of cooperators

NTZ radius (*R0 (radius without s hadow))

4
m u=2
m u=4
m u=6
m u=8
m u=10
m u=12
m u=14

3 .5

2 .5

1 .5

0 .5

10

20

30

40
50
60
Nu m b e r o f C R

70

80

90

100

Cooperation gain as a function of variance


of shadowing & number of cooperators?

30

25

Radius of NTZ

20

15

10

0
0

10
2

8
4

6
6

4
8

2
10

0
S hadowing varianc e

Number of CRs

Limitations on the cooperation


Gain!
More CRs implies more
collision in accessing the
control channel
(probability of
transmission is 0.1)

Probability of collision

0.5

Also it takes too long to


hop over all the available
channels. CR might
choose to sense m out of
M channels

10

20

30

40
50
60
Numbe r of ne ighbors

70

80

90

R
m
N eff = N total * 1 *
R2 M

100

Cooperation protocols for PTsensing


Request-NACK protocol
A CR which wants to send data sends a Request for channel m.
Each neighbor which hears the Request replies with a NACK (after a
small interval delta) if it regards the channel as busy
NACKs may collide, but the requesting CR detects energy after delta
The CR does not send data if it hears NACK(s)
Not necessary to send Request each time, if the availability of the
10
channel changes slowly
N=6
N=8
N=10

collision
reduce the effective
number of cooperators

N=number of neighbors of each CR.


Each CR sends Request w.p. r, each
CR detects the PT w.p. 0.6

Average effective # of cooperators

Too many NACK(s)


8

0
-4

-3
-2
-1
log10(r): r is prob of sending a Request

Cooperation protocols for PT-sensing


Random Back-off
When a neighbor receives a Request packet from a CR, it sets
a random back-off time from a window [1, W], and sends a
NACK after backing off,
All neighbors keep quiet after the confirmation packet (sent
after the CR successfully receives the first NACK)
2.6

6.5

Ave ra ge num be r of NACKs s e nt

The (average) s lot when the firs t NACK received


6

2.4

5.5

2.2

4.5

1.8

1.6

3.5

1.4

3
10

15

20
number of neighbors

25

30

10

15

20
num be r of ne ighbors

W=32, detection probability of each neighbor is 0.6

25

30

Cooperation protocols for PTsensing


Random Back-off (cont)
1.2

x 10

-3

probability of receiving no NACK


during the time-window

1
0.8
0.6
0.4
0.2
0
10

15

20
number of neighbors

25

30

W=32, detection probability of each neighbor is 0.6

Cooperation protocols for PTsensing


Broadcast protocol
In each time slot, each CR
independently broadcasts, with
probability r, a packet containing
the channel usage information it
has collected.
It also broadcasts the packet
when there are changes of the
availability information.
Each CR combines the
information
in the packets it has received from
its neighbors.

May produce much


communication
Collision
battery life time for CRs

1
prob of receiving a collided packet

0.8

N=5
N=10
N=15
N=20

0.6

0.4

0.2

0
-4

-3
-2
-1
log10(r): r is prob of broadcasting a packet

Exposed terminal problem remains


in PT-sensing
Even with cooperation, Exposed-terminal
problem still exists due to receiver
uncertainty
rmax
r2
PR1

rmax
PR1

PR2
d

PR2

R2

PT

PT

Rmax

Rmax

PR3

PT-sensing

PR3

PR-sensing

PR-sensing
Assume that PR participates to the protocol (e.g.
via sensor nodes [Kanan-Ben])
Can use the Request-NACK protocol or the Randombackoff protocol as in PT-sensing
Except that, upon receiving PR-Request from a CR, PR
responses and CRs keep quiet

The sensor at each PR can also use the Broadcast


protocol, to indicating which channel the PR is using

Spectrum utilization can be further improved by


using slow start algorithm

Comparison of PT-sensing & PRsensing


PT-sensing PR-sensing
Need cooperation of
other CRs due to
shadowing?

Yes

No, but
cooperation helps

Need sensors?

No

Yes

Need control channel? Yes

Yes

Hidden terminal or
Exposed terminal
problems?

Yes

No

Spectrum Utilization

Lower

Higher

Requirement of
detection capability

Higher

Lower

Thank you!

Potrebbero piacerti anche