Cognitive Radio at UCSD

1
Cognitive Radio at UCSD
Faculty Pamela Cosman, Larry Milstein, Larry
Larson, Bhaskar Rao, Don Kimball, Elias Masry,
Rene Cruz Funding sources Center for Wireless
Communications, National Science Foundation, Army
Research Office

• Waveform and receiver design multicarrier (MC-DS
  or MC OFDM) can handle non-contiguous and
  variable spectrum by just not transmitting some
  of the subcarriers
• Compression techniques scalable video coding,
  multiple description coding
• Spectral sniffing Channel sensing
• Interference suppression/avoidance
• MIMO for multiplexing or diversity, depending on
  channel
• Machine Learning
• Economic models allocation of spectrum among
  cognitive radio users

2
Cog Radio in Military Context

• Ad hoc mobile networks and multi-carrier DS CDMA
• Detection of which spectral bands are occupied by
  primary users
• Cooperation among cognitive radio users to
  improve detection

Exchanged information available subcarrier
set, channel estimate, and noise power estimate
of each available subcarrier.
30 users in a 200x200 m2 area, 64 subcarriers,
maximum QAM constellation is 128, BER10-4, PN
0.4
As Doppler spread gets large, channel estimates
degrade, hence throughput degrades.
PI Probability of a subcarrier occupied by
primary user PN Probability of NBI overlaying a
subcarrier of a user.
3
UWB WiMAX Detection

• UWB 3.1 GHz 10GHz
• WiMAX 2.5 2.7, 3.4 3.6, 5.25 5.85 GHz
• Two Connections AP - MT
• 1. Uplink Present only sporadically
• 2. Downlink Present almost continuously

? Goal Detect WiMAX MT when downlink only
1. Small portion of WiMAX LO signal leaks to the
antenna and radiates 2. UWB device detects
unmodulated WiMAX LO leakage, determines presence
of nearby WiMAX MT device
4
Spectrum sensing machine learning

• Spectrum sensing and estimation
• Simple Fourier-based methods to compute power at
  output of BPF have drawbacks (limited ability to
  suppress out of band interference, dont exploit
  structure in signal)
• Data Adaptive Spectral Analysis, Feature
  Extraction and Pattern Recognition
• Superimposed Pilot based estimation
• MIMO systems
• Spectrum sensing complicated by directional
  transmissions
• Facilitated by multiple receive antennas
• Machine Learning for Cognitive Radios
• Adapt both physical and network layer
• Device learns from its past usage history (e.g.,
  channel characteristics, spectral usage, traffic
  conditions)
• Learns from external sources such as access
  points or through device cooperation

5
Allocation of bandwidth to 5 video users based on
distortion-rate characteristics
Note we assume that you DO have to PAY for
scavenged white space
User demand f(current cost, current video
content, rate of expenditure)
6
Fine-Grain Scalable Video Coding

• Base Layer (BL) arrives, ?tolerable quality video
• As progressively more enhancement layer (EL)
  arrives, quality gets progressively better
• Base Layer sent on dedicated spectrum (primary
  user)
• Enhancement Layer is sent opportunistically in
  white space (cognitive radio user)