1 Power Control and Bargaining for Cellular - PowerPoint Presentation

Slide1

1

Power Control and Bargaining for CellularOperator Revenue Increase under LicensedSpectrum Sharing

NetGCOOP

, Avignon, 23.11.2016

Vaggelis

G. DourosStavros Toumpis George C. Polyzos

vaggelis.douros@inets.rwth-aachen.deSlide2

Towards the 5G Era (1)

2

CAGR: Compound Annual Growth Rate

Source:Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2015–2020 White Paper, Feb. 2016

Slide3

Towards the 5G Era (2)

Spectrum?Traditional spectrum availability is scarce

Bridging the spectrum gap with 5G3

2015



2020

# Devices: 1.5x# Data traffic: 10xSlide4

Co-Primary Shared Access

A complementary new alternative for spectrum sharing, where multiple operators jointly use a part (or the whole) of their licensed spectrum Mutual Renting: Operators have individual licenses to access exclusive frequency bands, and are mutually allowed to rent parts of their licensed resources to their peers upon requestLimited Spectrum Pool: a group license is given to an operator for using a common pool of spectral resources, which is shared with a limited set of operators that have equal access rights

4Slide5

The ChallengeThe challenge:

Ensure that wireless operators can seamless coexist in such licensed spectrum sharing scenarios

5Slide6

Our ContributionsPower control with bargaining for

improvement of operators’ revenuesOur joint power control and bargaining scheme outperforms both the

Nash Equilibrium without bargaining and classical pricing schemes in terms of revenue per operator and sum of revenuesA simple set of bargaining strategies maximizes the social welfare for the case of 2 operators with lower

communication overhead than pricing

6Slide7

System Model

N operators, 1 BS per operator, 1

MN per BSEach operator:controls the power of its BScharges its MN per round based on the QoS

aims at maximizing its revenue per round

7

Each device:will not change operator

downloads various filespays more for better QoS without min./max. QoS requirementsSlide8

Game Formulation

A non-cooperative game formulationThe game admits a unique Nash Equilibrium: All BSs transmit at

PmaxOur work: Can we find a more efficient operating point?

8

Players

BSs/OperatorsStrategy

Power Pi in [Pmin,Pmax]Utilityci Blog(1+SIR

i)Slide9

Analysis for

N Operators (1)

Red makes a “take it or leave it” offer to Black“I give you

o1,2

€ to reduce your power M times”

9

NE revenueEstimated revenueSlide10

Analysis for N Operators (2)

Black accepts the offer iff:

Win-win scenarioKey question: Are there cases that the

maximum offer that red can make is larger than the

minimum offer that black should receive?

10Slide11

Analysis for 2 Operators (1)

11

Good news: We can always find a better operating point than the NE without bargaining

Theorem: Let

and

the ratios of the path gain coefficient of the associated BS to the path gain coefficient of the interfering BS.

If M, then

If

M

, then

 Slide12

Analysis for 2 Operators (2)

12

Better news: By asking for the maximum power reduction, the operators will reach to an agreement at either point A1 or point

A2 and they will maximize the

social welfare

Theorem: The maximum sum of revenues of the operators corresponds to one of thefollowing operating points:

A1=(P1, P2)=(Pmax, Pmin) or A2=(P

1, P2

)=(

P

min

,

P

max

)

.Slide13

How to Pick a Good Offer?

Full knowledge “On the fly”Partial

knowledge Distributed iterative schemeStart with max. offerIf the offer is accepted, reduce it

13

Players

BSs/Operators

StrategyPower Pi in [Pmin,Pmax]Utilityci

Blog(1+

SIR

i

)Slide14

14

Numerical Examples (1)

limit

minimum offer

Maximum offer

Revenue

at the NEMN1 BS2 BS1

MN2

All these points are more efficient than the NE

BargainingA1(2): Revenue of

OP

1

(

OP

2

)

when

OP

1

makes offers

OP

1

offers

M

=32

Step

=1.15

q

=

r=

 Slide15

Numerical Examples (2)

BS2 MN

2

MN1

BS1

[Alpcan,02]

ParametersSpread factor L=4Charging factor c=1Pricing factor z=1.5

G11=0.5,

G

21

=0.2

G

12

=0.05,

G

22

=0.2

15

BargainingA1: Revenue of

OP

1

when

OP

1

makes offers

BargainingA1>NE1 in all scenarios

BargainingA1>Pricing1 in all scenariosSlide16

Numerical Examples (3)

BargainingA2>NE2 in all scenariosBargainingA2>Pricing2 in the first 3 scenarios

[Alpcan,02]

16

BS

2

MN2MN1BS1BargainingA2: Revenue of OP2 when OP1 makes offersSlide17

OP

2 makes offers

Numerical Examples (4)

Bargaining outperforms Pricing

17

Bargaining outperforms NE in all scenariosSlide18

Numerical Examples (5)-Sum of Revenues

BargainingA/B strictly outperforms NE and Pricing in terms of sum of revenuesBargainingB maximizes the social welfare

[Huang,06]

18Slide19

Sum of RevenuesAccumulated Results

~120,000 simulationsMaxBargaining strictly outperforms Pricing in the vast majority of scenarios

Even MinBargaining strictly outperforms Pricing in most cases

Parameters

Charging

factor c=1

Best pricing factor z*=1.5Gij={0.01,0.06,0.11,…,0.96}19

MaxBargaining=max{BargainingA,BargainingB}

100%Slide20

Qualitative Comparison (1)

20

Bargaining

Pricing [Alpcan,02]Social

welfare for 2 players

YesNo guaranteeEach payoff >= NE payoff

Yes NoAdditional path gain knowledge1 path gainNoGeneral convergenceYesOnly if N-1<LSlide21

Qualitative Comparison (2)

21

BargainingMaxSum

[Huang,06]Social welfare for

2 playersYes

YesSocial welfare for

N playersOpen issueNoAdditional path gain knowledge for 2 players1 path gain1 path gain + 1 pricing profileAdditional path gain knowledge for N players1 path gainN

-1 path gains + N

-1

pricing profilesSlide22

Agenda for Future Directions

N OperatorsMinimum/maximum data ratesCoalitional game theoryHow to share their revenues?Shapley value, coreNash Bargaining Solution

Communication overhead22Slide23

Take-home Messages

Our work: Licensed spectrum sharing through power control and bargainingAppealing property: By combining power control with bargaining, no player receives lower payoff than its payoff at the NE

Highlight: Bargaining outperforms standard pricing techniques 23Slide24

 Merci! 

Vaggelis G. DourosPost-Doctoral ResearcherInstitute for Networked Systems

RWTH Aachen University 52072 Aachen, Germanyvaggelis.douros@inets.rwth-aachen.de

http://www.aueb.gr/users/douros/

24