1. Trends in Spectrum Sharing for Future Wireless Networks
Luiz DaSilva
Professor of Telecommunications
EMC Europe 2016
Wroclaw, Poland, 8 September 2016

2. Trinity College Dublin
Founded 1592

3. CONNECT Future Communications and Networks X Media Rich Applications
smart sensors
microelectronic circuits
RF design
energy harvesting strategies
antennas
thermal strategies
optical technologies
PHY layer signal processing
software/cognitive radio platforms
optical architectures
optical/wireless interface
cognitive networking
virtualization techniques
wireless/mobile architectures
network optimization
network performance monitoring
mobile services
cloud services
spectrum management
privacy/security services
service platforms
Audio-visual media processing
M2M/D2D applications
cyberphysical systems
PHY layer monitoring
sensor networks
Rapid Prototyping and Experimentation
Media Rich Applications
Future Communications and Networks X

4. Roadmap for this talk The state of spectrum sharing
Initial thoughts on the current state of spectrum sharing
2. Sharing of radar bands by small cells
Temporal sharing and cognitive beamforming of radar bands, for small cell applications
3. Spectrum and radio access infrastructure sharing between operators
Tradeoffs in licensed spectrum and RAN sharing among MNOs

5. Acknowledgements
This work was primarily funded by the Science Foundation Ireland.
Much of the work is due to Mr. Francisco Paisana and Dr. Jacek Kibiłda.

6. The state of spectrum sharing

7. Dynamic spectrum sharing
Rewind to ca. 1999
For wireless systems to evolve, they need more spetrum
Significant temporal and geographical variations in spectrum utilization
Licensed spectrum is often under-utilized
Why not open spectrum that is left idle for opportunistic use?
Huge efficiency gains, no harm to incumbents
Everyone is happy!

8. Dynamic spectrum sharing
A (hopefully) more sober view
Limited success in opportunistic spectrum sharing
Inertia, uncertain business models, reluctance from both incumbents and new potential users
Some successful trials in TVWS in Africa, Asia
But spectrum sharing remains a key component of, say, 100x improvement in capacity expected in 5G
Both in additional spectrum and in cell densification
Besides, opportunistic use is not the only way to share spectrum
Unlicensed spectrum sharing has been a huge success

9. Sources of resistance to spectrum sharing
FROM INCUMBENTS
Perceived lack of control over secondary use
Assurances about availability
Lack of compensation
Competitive advantage (for commercial services)
FROM NEWCOMERS TO THE BAND
Uncertainty about QoS expectations
Reliability and availability

10. Addressing concerns
Giving incumbents more control over the decision making process
When, where, under what conditions to share
Creating appropriate incentives to share
Monetary (e.g., incentive auctions, contracts between primary and secondary)
Non-monetary
Intermittently available bands used to complement other spectrum
E.g., control channel and reference signaling in licensed spectrum

11. An attempt at classifying spectrum sharing
INCUMBENT / SECONDARY USE
SHARING AMONG EQUALS
Purely opportunistic
Unlicensed: etiquette for coexistence
Licensed shared access
Licensed: multi-lateral agreements
PCAST (3-tier) model

12. Sharing of radar bands by small cells

13. Spectrum sharing in the radar bands
Motivating factors
Under-utilized spectrum in L, S, and C bands
Radars have low duty cycle, and reasonably predictable incumbent usage patterns
1400
960
2700
1710
3500
5000
5850 MHz
2170
t (sec)

14. Spectrum sharing in the radar bands
Use cases
License-exempt radio frequency devices in 5 GHz band
WLANs
LTE-U or Licensed Assisted Access LTE (LAA-LTE) proposed
Small cell deployments in the 3.5 GHz band
In advanced stages of regulation by the FCC in the US
Sharing of DoD spectrum used for radar systems

15. Spatial and temporal sharing
Spatial sharing: determination of exclusion zones
Temporal sharing: dynamic tracking of incumbent use of spectrum
No temporal and spatial WSs
Temporal WSs
Spatial WSs

16. Geolocation database The case for it
Adaptive dimensioning of radar exclusion zones, according to radar parameters and requirements
DB complexity alleviated due to small number of incumbents
Drawbacks
Temporal sharing not possible
Inefficient spatial sharing due to radar mobility and conservative propagation models

17. The exclusion zone issue
The US case

18. Database-aided sensing
Paisana, Miranda, Marchetti, DaSilva [IEEE DySPAN 2014]

19. CR Operation Modes No exploitation of spatial and temporal WSs allowed
CR geographically deep inside an exclusion zone
Emergency or system malfunction scenarios
No exploitation of temporal WSs allowed
Radar exhibits unpredictable scanning patterns
Bistatic radar systems within interference range
Support of the GL-DB is not enough
Exploitation of temporal WSs allowed
Free exploitation of WSs (sensing is not required)
CR outside any PU exclusion zone and PUs are static

20. Degree of database support
Limited or moderate support
Limited Support
Threshold and channel availability check time (CACT)
Minimum and maximum assumed by PW, PRF, and antenna scan period (ASP)
The CR estimates the intercepted radar signals’ parameters through signal processing techniques
Moderate Support
For each radar, the DB provides: threshold, ASP, PW, PRF, fc, bandwidth, and some considerations regarding the radar mobility and frequency agility
Autocorrelation-based detection techniques

21. Degree of database support
Full support
Full Support
The DB also provides: IPM, radar antenna radiation pattern
Matched Filter detection techniques

22. Impact of degree of database support
Scenario: CR’s bandwidth of 20 MHz, TW=11ms
Radar is LFM with PW of 28 µs and PRF of 1000 µs

23. Measurements
Cork Naval Base, Ireland

24. Measurements
Cork Naval Base, Ireland

25. Signal processing for incumbent detection/prediction
Paisana, Marchetti, DaSilva [IEEE TCCN, under review]
More sophisticated signal processing techniques needed in practice, due to scattering, and possible presence of multiple radars

26. Augmenting the DB with beamforming
Paisana, Finn, Marchetti, DaSilva [IEEE DySPAN 2015]
Variable exclusion zones, depending on the interference cancellation capabilities of the SUs No
Yes

27. System model
LTE small cell eNB and UEs
Magnetron radar system

28. Fading model Based on the WINNER channel model
Channel matrix modelled as a superposition of several propagation path matrices of diffuse small scale parameters (AoA, AoD, delay, power, etc.)

29. Summary
Database-aided sensing as a possible extension to the current SAS model
Design of non-coherent radar signal detection and feature estimation, radar de-interleaving, scan tracking for temporal sharing
Proposal and evaluation of cognitive beamforming for more efficient use of spectrum bands
Military radar signal measurements and characterization
Proof-of-concept implementation of radar spectrum sharing using software-defined radio platform

30. Spectrum and Infrastructure Sharing among Operators

31. Sharing economy in mobile networks
Passive sharing
Towers, sites
Active sharing
Antennas, entire base stations, elements of the core network
Mobile Virtual Network Operators
₺Leasing of₺ a network
Over-the-top Service Providers
₺Leasing of₺ multiple networks

32. Studying tradeoffs between spectrum and RAN sharing between MNOs
Kibiłda, Di Francesco, Malandrino, DaSilva [IEEE DySPAN 2015]
No sharing
Infrastructure only
Spectrum only
Infrastructure + Spectrum (Full)

33. Comparing these alternatives
Metrics:
Coverage
Throughput
Dependence on:
Operator network deployment patterns
Sharing agreements in place
Spectrum sharing coordination

34. A resource management question
From mobile operators’ point of view, can infrastructure sharing be traded for spectrum sharing?

35. Operator network deployment patterns
Independently distributed inter-operator infrastructure
Spatially clustered inter-operator infrastructure
Spatially co-located inter-operator infrastructure

36. Spectrum sharing scheme
Channel bonding
Best channel selection

37. Spectrum sharing coordination
Operator 2 can use Operator 1’s spectrum as long as there is no base station of Operator 1 within some radius R

38. Independent deployment by two operators
Poisson point process (PPP) case
For no sharing and full sharing cases, can rely on results derived in the literature
For infrastructure only and spectrum only sharing, we can derive results for coverage and average data rate

39. Analytical expressions for coverage and user rate I
PPP with intensitites λ1 and λ2, with η =λ1/λ2

40. Analytical expressions for coverage and user rate II
PPP with intensitites λ1 and λ2, with η =λ1/λ2

41. Independent PPP is a simplistic assumption
Cluster processes to model multi-operator deployments
Premise: multi-operator RAN deployments exhibit significantly more clustering than single-operator
Investigated goodness of fit of log-Gaussian Cox process (LGCP), Matern cluster process (MCP) and Thomas process (TP)
Deployment data from Ireland, Poland, and the UK

42. Some of our results Kibiłda, Galkin, DaSilva [IEEE TMC 2016]
Combined multi-operator deployments cluster at shorter distances (high demand areas) and repulse at longer
Log-Gauss Cox Process provides the closest match to real data
Results are robust to several cities tested for in Europe

43. Validation We can cross-validate closed-forms and simulations
Channel bonding
Best channel selection

44. Coverage probability PPP LGCP GPP 𝑢→0 aggregation
Spectrum sharing sceheme
selection

45. Coverage probability
The impact of clustering, with Gauss-Poisson process and cluster radius 𝑢
Infrastructure sharing
Spectrum sharing
Full sharing

46. Average user rate
The impact of clustering, with Gauss-Poisson process and cluster radius 𝑢
PPP & GPP
LGCP & GPP
aggregation
Spectrum sharing sceheme
selection

47. Coverage probability Networks with coordinated spectrum sharing PPP
LGCP
GPP 𝑢→0
aggregation
Spectrum sharing scheme
selection

48. Average user rate Networks with coordinated spectrum sharing PPP LGCP
aggregation
Spectrum sharing scheme
selection

49. Summary

50. And the conclusions so far…
Infrastructure and spectrum cannot be simply substituted for each other, as they bring a tradeoff in coverage and throughput
The efficiency gains from the combination of infrastructure and spectrum sharing do not add linearly
The spatial distribution of the networks has a significant impact on the gains brought about by sharing
Channel aggregation brings about gains to data rate, while best spectrum selection yields significant improvement to coverage
The loss of coverage due to channel aggregation may be combated, without hurting the user rate, with spatial interference coordination

51. Final thoughts

52. Looking forward
More sophisticated models of spectrum (and infrastructure) sharing recognizing the incentives for both incumbent and prospective new users
Virtualized wireless networks combining assets from multiple operators in different technologies and spectrum regimes
Slicing of resources: protection of QoS from other virtual networks versus gains from statistical multiplexing
How to manage the use of resources in virtual networks
Resource management combining wireless and optical network resources (the FUTEBOL project)

53. Thank You luizdasilva.wordpress.com