Transcript Presentation

Performance of Spectrum
Monitoring for Channel
Selection in Cognitive Radio
Prashant Bajpayee
Indian Institute of Technology, Kanpur
Advisor: Dr. Daniel Noneaker
SURE 2005
Outline
 Background/Motivation
 System Model
 Simulation Details
 Observations
 Conclusion
 Future work
Motivation
 Currently most radio-frequency spectrum is assigned
exclusively to “primary” users (e.g. Broadcasting of T.V.
channels)
 Assigned spectrum is often underutilized in many locations
at specific times
 This “unused ”spectrum is lost opportunity for other
potential communication systems
General Solution
 Exploitation of unutilized spectrum on “secondary” user
basis and Cognitive radio is viewed as a novel approach
for efficient utilization
Characteristic of Cognitive Radio
 Intelligent wireless communication system
 Awareness
 Learning
 Adaptivity
 Efficiency
 Reconfigurable
 Software-defined radio
Advantages and Feasibility of CR
 Advantages
 Highly reliable communications
 Efficient utilization of the radio spectrum
 Dynamically reuse available spectrum
 Feasibility
 Advances in DSP, networking, machine learning,
computer software and hardware
 Convergence of digital radio and computer software
My Research Problem
 Development and simulation of a mathematical
model of channel monitoring function of CR
 Simulate and analyze the effect of time-varying
interference in each channel and design trade-offs
that result
System model
 Multiple – channel communication system
Ch 0
f0
Ch 1
f1
….......
Ch L
fL
 Channel monitoring approach used by destination:
 When no data is transmitting, sequentially monitor
channel to determine level of noise in each channel
 When RTS is received, destination should send CTS
specifying channel with least interference.
System Model (contd.)
 Assume DS Spread Spectrum binary antipodal
signaling is used by transmitter so that
N 1


bk
s(t )   (1)  A x( k 1) N i (t  iTc ) 
k 0
 i 0


 During monitoring of traffic channel j its received

signal is r (t )  n j (t ) and n1( t )' n 2 ( t )'    n j ( t )
are independent random process
 Two types of channel:
 Good channel with lesser interference
 Bad channel with more noise interference

Decision statistic and technique
 Monitor channel 1 by forming statistic
 Where
X1 
R  X Y
2
1
( m  n )Tc
 r (t )c(t ) cos( w t   )dt
i
1
mTc
and ,
Y1 
( m  n )Tc
 r (t )c(t ) sin( w t   )dt
i
mTc
and
c(t ) 

 x  (t  iT )
i  
i
c
1
2
1
2
1
Detection technique
 Similarly for all channels we get chi-square random
variable Rj with degree 2n and pick channel with smallest
value of Rj .
 c(t ) cos(1t  1 )

( m  n )Tc
c(t ) sin( 1t  1 )
mTc

c(t ) cos(2 t   2 )
r (t )  n(t )  s (t )

c(t ) sin( 2t   2 )


( m  n )Tc

X1
+
Y1
mTc
( m  n )Tc

 2
X2
 2
( m  n )Tc

mTc
Y2
Decision dev
 2
mTc
 2
R12
+
R22
Generalized model
 CR alters between frequency monitoring and
other tasks
 Each channel has time-varying interference/noise
 Now each channel is represented as two-state
Markov random process
1
S
 k = state of channel 1 at time k, may be 0 or 1

1
0
1

1 
Generalized model (contd.)
 Now each channel is good or bad depending upon its state at
that time
Monitor
f2
Monitor
f1
t=0
n
work
n+n1
n1 =q*n
-----
work
2n+n1
2n+2n1
Monitor
f1
Ln+Ln1
 All Channels change their states after each time interval = n
 New decision statistics are generated after n+n1 time interval
but only for one channel at a time
 Efficiency of system = n1/n+n1
Simulation Details
 Input parameters:
c = ratio of noise variances of good and bad channels
 n = time required to generate each statistic and equal
to number of chi-square random variables which add
up to form a test statistic
 n1 = the time radio spends on other tasks
 T_coh = measure of rate of change of states
 Assume   


    1 / 2 1  1 / e
1 / T _ coh

Variation of n
Variation of no. of channel
Time-varying case
Variation of no. of channels
Pe Vs T-coh
0
10
channel=2
channel=4
channel=6
Error Probability(Pe)
channel=8
-1
10
-2
10
0
100
200
300
400
T-coh
500
600
700
800
Optimum n for different T-coh
Pe Vs n(for diff. T-coh)
0
Error probability(Pe)
10
-1
10
T-coh=100
T-coh=900
T-coh=3600
n1/n=9
-2
10
0
5
10
15
20
25
n
30
35
40
45
50
Optimum n for different efficiency
Pe Vs n(for diff. efficiency)
0
Error Probability(Pe)
10
-1
10
n1/n=1
T-coh = 900
n1/n=9
n1/n=49
-2
10
0
5
10
15
20
25
n
30
35
40
45
50
Conclusion
 On increasing the ratio of noise variances, error decreases
 Increasing values of n provide better estimates for
test statistics
 Increase in efficiency causes increase in error probability
 Increase in no. of channels may increase or decrease system
performance depending on value of T_coh
 Trade-Off

Smaller n means that test statistics are more “current” but
larger n means that each statistics is more statistically
reliable
Future work
 Consider a more general Markov process with different
parameters
 Analyze the effect of mutual coupling amongst
different radios on channel-noise distribution of each
radio
 Examine the performance in terms of bit error
probability
Acknowledgments
 Dr. Daniel Noneaker
 Dr. Harlan Russell
 Dr. Xiao-bang Xu
 NSF
Questions
???