Spectrum Awareness in Cognitive Radio Systems based on Spectrum Sensing

1
Spectrum Awareness in Cognitive Radio Systems
based on Spectrum Sensing

• Miguel López-Benítez
• Department of Electrical Engineering and
  Electronics
• University of Liverpool, United Kingdom

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014
2
Introduction

• Dynamic Spectrum Access (DSA) / Cognitive Radio
  (CR)
• Opportunistic spectrum access paradigm
• Behaviour and performance depends on primary
  spectrum activity
• Knowledge of spectrum activity statistics ?
  spectrum decisions
• Prediction of future spectrum occupancy patterns
• Selection of channel / band of operation
• Spectrum / radio resource management decisions
• Obtaining spectrum information (research topics)
• 1) Spectrum sensing algorithms
• 2) Estimation of spectrum activity statistics
• 3) Modelling of spectrum occupancy patterns

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 2
3
Spectrum awareness

• Beacon signals
• ? Perfect information
• ? Requires agreement primary-secondary
• ? Changes in legacy radio systems economical /
  technical problems
• Databases
• ? Perfect / accurate information
• ? Relies on an external system (technical,
  administrative legal problems)
• ? Need for geolocation in DSA/CR terminals (cost,
  location accuracy, etc.)
• ? Database updating rate ? not suitable for
  dynamic bands
• Spectrum sensing
• ? Does not rely on an external system
• ? No changes needed to legacy (primary) system
  simple and inexpensive
• ? Suitable for dynamic spectrum bands
• ? Inaccurate information (spectrum sensing errors)

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 3
4
Spectrum sensing algorithms

• Wide range of spectrum sensing algorithms
• Trade-offs detection performance, complexity,
  computational cost, applicability.
• Applicability depends on available information
• Detailed knowledge ? Matched filter
• Certain features ? Feature detector
  (ciclostationarity, pilots, others)
• Correlated signal (oversampling, multiple
  antennas) ? Covariance detector
• No prior information ? Energy detector
• Ideal sensor
• Simple (low complexity and low computational
  cost)
• General applicability (ability to detect any
  signal format)
• High detection performance (high detection prob.,
  low false alarm prob.)

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 4
5
Spectrum sensing algorithms

• Common research trend
• Maximise detection performance,
• At the expense of higher complexity /
  computational cost, limited applicability.
• Alternative approach
• Improve detection performance.
• Without sacrificing complexity / computational
  cost and applicability.
• Why? (motivation)
• Meeting detection performance requirements by one
  single terminal may be unfeasible.
• Cooperative/collaborative sensing and
  network-aided approaches relax requirements.
• Even if feasible, too complex and expensive.
• Inexpensive terminals are key to the success of a
  new radio/mobile technology.
• How? (approach)
• Variations of the energy detection principle
  simple, low complexity, applicability

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 5
6
Spectrum sensing algorithms

• Example Improved Energy Detection (IED)
  algorithm
• Combination of spectrum sensing events based on
  energy detection.
• M. López-Benítez, F. Casadevall, Improved
  energy detection spectrum sensing for cognitive
  radio, IET Communications, Special Issue on
  Cognitive Communications, vol. 6, no. 8, pp.
  785-796, May 2012.

Better performance
Same/similar complexity
Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 6
7
Estimation of activity statistics

• Relevant for spectrum and radio resource
  decision-making processes
• Spectrum activity statistics can be estimated
  from sensing observations
• Sensing observations ? infer period durations ?
  compute activity statistics
• Minimum period duration, mean/variance,
  underlying distribution, etc.
• Practical limitations (e.g., spectrum sensing is
  imperfect)

Perfect Spectrum Sensing (PSS) (e.g., high SNR)
Imperfect Spectrum Sensing (ISS) (e.g., low SNR)
Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 7
8
Estimation of activity statistics

• Relevant aspects
• Activity statistics
• Duration of idle/busy periods (minimum, average,
  variance, distribution, etc.)
• Other more sophisticated metrics (channel
  load/duty cycle, etc.)
• Practical limitations
• Imperfect sensing performance
• Finite sensing period
• Limited number of observations
• Aspects to be analysed
• What are the activity statistics actually
  estimated by DSA/CR terminals under real
  conditions?
• What is the difference (error) between the
  estimated and the real spectrum activity
  statistics?
• What can be done to minimise the estimation
  error?

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 8
9
Modelling of spectrum occupancy

• Spectrum activity statistics can be used to
  parametrise spectrum occupancy models
• Application of spectrum occupancy models
• Analytical studies
• Simulation tools
• Design/dimensioning of DSA/CR networks
• Design of new DSA/CR techniques
• Spectrum measurements
• Models based on real spectrum data ? Realism and
  accuracy
• Challenge harmonisation of methodology
  (equipment, field measurements, data
  post-processing, etc.)

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 9
10
Modelling of spectrum occupancy

• Relevant parameters to be modelled
• Time-dimension parameters
• Channel load (duty cycle)
• Period durations (minimum, average, variance,
  distribution)
• Correlation properties of period durations
• Frequency-dimension parameters
• Statistical distribution of duty cycle
• Clustering of duty cycle
• Number of free channels at any time
• Space-dimension parameters
• Perceived spectrum occupancy level

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 10
11
Conclusions

• Opportunistic nature of DSA/CR systems
• Knowledge on spectrum occupancy is useful
• Spectrum awareness
• Beacon signals
• Data bases
• Spectrum sensing
• Spectrum awareness based on spectrum sensing
• 1) Spectrum sensing algorithms
• 2) Estimation of channel activity statistics
• 3) Modelling of spectrum occupancy

Workshop on Wireless Networks University of
Liverpool, United Kingdom, 25 June 2014 Slide 11
12
for your attention !
Email M.Lopez-Benitez_at_liverpool.ac.uk Website w
ww.lopezbenitez.es