Spectrum Sharing for Unlicensed Bands
Spectrum Sharing for Unlicensed Bands
Raul Etkin, Abhay Parekh, and David TseDept of EECSU.C. Berkeley
Project supported by NSF ITR ANI-0326503 grant
DySPAN 2005, Nov. 10, 2005
2Spectrum Sharing for Unlicensed Bands
Problem: Spectrum SharingCan multiple heterogeneous wireless systems coexist and
share spectrum in a fair and efficient manner?
•Unlicensed setting
•Equal rights
•Different goals
Introduction
3Spectrum Sharing for Unlicensed Bands
Main Goals
• Find spectrum sharing rules that are:– Efficient– Fair– Robust against selfish behavior
• Study how to obtain good performance without
cooperation.
Introduction
4Spectrum Sharing for Unlicensed Bands
The Model• Flat Fading• Systems use Gaussian signals with
PSD {pi(f)}.
• Power constraint for each system.
• Total bandwidth W.
• Interference treated as noise.
• Design choice: power allocations over frequency.
Introduction
C1,1
C2,2
C1,2
C2,1
N0
N0
noise interference
5Spectrum Sharing for Unlicensed Bands
Static Gaussian Interference Game
• M Players: the M systems
• Strategy of system: power allocation satisfying power
constraint
• Utility of system i non-decreasing, concave on Ri.
• All parameters ({ci,j},{Pi},N0) are common knowledge.
• Players select their actions simultaneously.
Non-cooperative Scenarios
6Spectrum Sharing for Unlicensed Bands
Static Game Analysis
Non-cooperative Scenarios
full spread Nash equilibrium
Achievable rates
proportional fair
orthogonal
Unique if
XXinterference
limited
noise limited
price of anarchy
7Spectrum Sharing for Unlicensed Bands
Dynamic Game• What rate vectors are achievable as a N.E. in the dynamic game ?
Non-cooperative Scenarios
achievable with self enforcing strategies
Punishment strategies: encourage cooperation by threatening to spread
good behavior
punishment
8Spectrum Sharing for Unlicensed Bands
Example A
Non-cooperative Scenarios
asymmetry in power and gains
802.11 bluetooth
full spread N.E.
proportional fair
9Spectrum Sharing for Unlicensed Bands
Example B
Non-cooperative Scenarios
asymmetry in power
802.11
bluetooth
full spread N.E.
proportional fairQ: Can be achieved with other self enforcing strategies ?
No !
best PF self enforcing point
10Spectrum Sharing for Unlicensed Bands
Asymmetry and Fairness
Non-cooperative Scenarios
No Loss
No Loss
11Spectrum Sharing for Unlicensed Bands
Conclusions• With complete information and moderate asymmetry it is
possible to find policies that are fair, efficient and robust against selfish behavior.
• Results can be extended to:– Non-Gaussian signals– Any achievable rate region (with interference cancellation, etc.)
• Future research: – Find distributed algorithms that do not require complete
information and approximate the performance predicted here.– Investigate how to deal with cases of extreme asymmetry.
Conclusions
12Spectrum Sharing for Unlicensed Bands
Related Work
• Distributed optimization of power spectral allocations for DSL using
iterative waterfilling [Cioffi, et al. 2001]
• Use of Game Theory to analyze outcomes of iterative waterfilling
algorithm [Cioffi, et al., 2002]
• Iterative waterfilling may lead to poor performance. Signal space
partitioning often leads to better results. [Popescu, Rose &
Popescu, 2004]
• Use of genetic algorithms to find good strategies in repeated games
with small strategy space. [Clemens & Rose, DySPAN 05]
Introduction
Top Related