Fundamental Design Tradeoffs in Cognitive Radio Systems

1
Fundamental Design Tradeoffs in Cognitive Radio
Systems

• Mubaraq Mishra
• Berkeley Wireless Research Center
• University of California at Berkeley
• Joint work with Prof. Anant Sahai, Prof. R. W.
  Brodersen
• Rahul Tandra and Niels Hoven
• BWRC Retreat
• June 05, 2006

2
Cognitive System Design Problem
K systems
Decodability Radius
N radios
17dB SNR
Primary System
Cognitive Radio Systems
Design Issues What is the required Probability
of Detection of the system? How much fading
margin do we need to budget for? How do we set
the Sensing Time?
Design Goal Minimize Interference to Primary
System Metric Probability of Harmful
Interference (PHI)
3
Fading Margin Effect of Multiple Radios
Complementary CDF of Loss from Multipath and
Shadowing
Higher the Cooperative gain, lesser the margin
for fading
Cooperative Gain
4
Cooperative Gain
Cooperative Gains Vs Number of Users
Cooperation enables system to overcome SNR limits
from Noise uncertainty Increasing number of
users reduces fading margins
Distance dependent Path Loss
Required gain to overcome Uncertainty Limit
17db Gain
Multipath Raleigh Fading, Shadowing Log Normal
(3.5dB)
5
Shadowing correlation
Independent Shadowing
Correlated Shadowing
1
Exponential Decay Model
Shadowing Correlation Coefficient
0
Displacement
6
Shadowing correlation leads to reduced gains
Increasing User Densities
Under the exponential shadowing assumption, we
would like to poll users that are far away
200 users, Gain 2dB
30 users, Gain 4dB
7
Real World Shadowing Correlation
Exponential drop
Base Station 1, Omni Antenna
Gradual drop
Base Station 2 Sectored Antenna
Drive Measurements from Qualcomm Frequency 1900
MHz Total displacement 1.5km
Exponential Model
8
Impact of Misbehaving/Adversarial radios
Cooperative gains with ? fraction untrusted users
are bounded by those achievable by a trusted
population of 1/? trusted users. To achieve
these gains, we need M 1 when ? large.
Cooperative Gains Vs Number of Users
Distrust limits Cooperative gains
9
How do we set Sensing Time?

• Case I Primary User absent
• Short Sensing time for fast access to spectrum
• Yet need to meet required PD,system
• Case II Primary User present
• Fast detection of Primary to reduce interference
• Radios need to start sensing new frequency band

10
Setting Sensing Time
Required Probability of Detection per radio
decreases with number of radios Required
Probability of False Alarm per radio decreases
with number of radios spirit
Sensing Time Vs Number of Users
11
Sensing Time Energy Coherent Detector
CCDF of Sensing Times
10 KHz
High TX power
6 MHz
Med TX power
Low TX power
Energy Detector Performs Better
Coherent Detector Performs Better
Need a Combination of Detectors
12
Cognitive System Design Problem Another Look
No Talk Radius
Protected Radius
Decodability Radius
Margin
F(50, 90)
17dB SNR
Primary System
Cognitive Radios
Design Goals Minimize Interference to Primary
System Maximize Cognitive Radio capacity
Metrics Interference (Protection Margin)
Secondary User Power Density Probability of
False Alarm Sensing Overhead
13
Inter System Design
Protection Margin
Number of Radios (N)
Shadowing Correlation
Fading Margin
Secondary Power Density
No Talk Radius
Noise Uncertainty
Uncertainty Margin per Radio
Uncertainty from other Secondary users
Coordination Radius
14
Coordination Radius Versus Power Density
TX Power 80dBm Decay Model r-2 exp(-?
d)
Bluetooth
WiFi
Inter-system coordination radius (m)
Secondary System Density (dBW/m2)
15
Coordination Radius Wall !!
Secondary to Primary decay is slower than Primary
to Secondary decay
Bluetooth
WiFi
Inter-system coordination radius (m)
Secondary System Density (dBW/m2)
16
Coordination Radius with Cooperation
Higher power density Large no talk radius, More
Radios for cooperation Weaker Primary
Signal Uncertainty needs to be
small Coordination radius increased.
Bluetooth
WiFi
Optimal densities
Inter-system coordination radius (m)
Non-coherent Detector
Coherent Detector, 10 pilot power
Secondary System Density (dBW/m2)
17
Conclusions

• Cognitive Radio system design is complex
• Cooperation helps reduce fading margins but
• is effected by shadowing correlation
• and distrust
• Sensing Time is significantly reduced due to
  cooperation
• Can reduce target sensing times when Primary is
  absent
• Detection times are vastly improved when Primary
  is present
• Coherent Energy Detectors outperform each other
  in different regimes -- need both detectors in
  the network
• Coordination Radius is required to manage
  uncertainty
• Coordination radius can also exhibit wall like
  effects
• There exists a power density at which the
  coordination radius is minimized.