Cognitive Virtual Network Operator Games Dr. Jianwei Huang Network - - PowerPoint PPT Presentation

▶
cognitive virtual network operator games
SMART_READER_LITE
LIVE PREVIEW

Cognitive Virtual Network Operator Games Dr. Jianwei Huang Network - - PowerPoint PPT Presentation

Aug 14, 2022 271 likes •772 views

Cognitive Virtual Network Operator Games Dr. Jianwei Huang Network Communications and Economics Lab (NCEL) Information Engineering Department The Chinese University of Hong Kong Joint work with Lingjie Duan and Biying Shou Jianwei Huang (NCEL)

slide-1
SLIDE 1

Cognitive Virtual Network Operator Games

  • Dr. Jianwei Huang

Network Communications and Economics Lab (NCEL) Information Engineering Department The Chinese University of Hong Kong Joint work with Lingjie Duan and Biying Shou

Jianwei Huang (NCEL) CVNO Games August 2010 1 / 44

slide-2
SLIDE 2

Eight Universities in Hong Kong

The Chinese University of Hong Kong (CUHK) The University of Hong Kong (HKU) The Hong Kong University of Science and Technology (HKUST) The Hong Kong Polytechnic University (PolyU) City University of Hong Kong (CityU) Hong Kong Baptist University (HKBU) The Hong Kong Institute of Education (IED) Lingnan University

Jianwei Huang (NCEL) CVNO Games August 2010 2 / 44

slide-3
SLIDE 3

The Chinese University of Hong Kong (CUHK)

One of the two comprehensive universities in Hong Kong Strong engineering program

Jianwei Huang (NCEL) CVNO Games August 2010 3 / 44

slide-4
SLIDE 4

CUHK Information Engineering

23 Full-time Faculty Members 8 IEEE Fellows Research Areas

◮ Communication Theory (Birth Place of Network Coding) ◮ Wireless Communications (Cognitive Radio/MAC/MIMO) ◮ Internet and Networking (P2P/Network Economics) ◮ Image and Video Processing ◮ Optical Communications Jianwei Huang (NCEL) CVNO Games August 2010 4 / 44

slide-5
SLIDE 5

NCEL: Network Communications & Economics Lab

4 postdocs 7 graduate students

Jianwei Huang (NCEL) CVNO Games August 2010 5 / 44

slide-6
SLIDE 6

Research Themes

Wireless Communications Network Economics Distributed Optimization Game Theory

Jianwei Huang (NCEL) CVNO Games August 2010 6 / 44

slide-7
SLIDE 7

Research Areas

Cognitive Networking Network Economics Network Optimization Video Communications Network Security Interference Management Cooperative Communications Wireless MAC DSL Optimization Jianwei Huang (NCEL) CVNO Games August 2010 7 / 44

slide-8
SLIDE 8

Cognitive Virtual Network Operator Games

Jianwei Huang (NCEL) CVNO Games August 2010 8 / 44

slide-9
SLIDE 9

Mobile Virtual Network Operators

  • Spectrum
Owner

VNO

  • Users

Spectrum Acquisi.on Service Provision

Mobile Virtual Network Operators (MVNOs):

◮ Virtual: does not own radio spectrum or physical infrastructure ◮ Spectrum acquisition: spectrum leasing from spectrum owner ◮ Service provision: pricing & spectrum allocation among local users

Contributions to wireless market:

◮ Offering more flexible and innovative services ◮ Raising the competition level Jianwei Huang (NCEL) CVNO Games August 2010 9 / 44

slide-10
SLIDE 10

Current Status

Successful MVNO deployment worldwide:

◮ First commercial success:

from 1999

◮ Over 400 MVNOs owned by 360 companies by Feb. 2009

MVNO (USA) Spectrum Owner Technology IDT7 Verizon CDMA 800/1900 Call Plus AT&T GSM 850/1900 AirLink Mobile Sprint PCS CDMA 1900 . . . . . . . . .

Often obtain spectrum via long-term contract. This talk: we consider more dynamic & efficient spectrum acquisition.

Jianwei Huang (NCEL) CVNO Games August 2010 10 / 44

slide-11
SLIDE 11

Spectrum is a Scarce Resource

Jianwei Huang (NCEL) CVNO Games August 2010 11 / 44

slide-12
SLIDE 12

Spectrum is a Under-Utilized

c Share Spectrum Co.

Jianwei Huang (NCEL) CVNO Games August 2010 12 / 44

slide-13
SLIDE 13

Cognitive Virtual Network Operators

Cognitive Virtual Network Operators (CVNOs)

◮ More flexible spectrum investment choices

Investment Choices Dynamic Leasing Spectrum Sensing Time Scale Small Small Cost High Low Reliability High Low

◮ More efficient pricing to maximize profit ( = revenue - cost) Jianwei Huang (NCEL) CVNO Games August 2010 13 / 44

slide-14
SLIDE 14

CVNO Games

Optimal investment and pricing decisions of CVNOs Two parts:

System Model Monopoly CVNO Duopoly CVNOs Operator One Two (Asymmetric) Investment Methods Sensing & Leasing Leasing Users Heterogeneous Heterogeneous Game Model Stackelberg Game Multi-leader-follower Game

Jianwei Huang (NCEL) CVNO Games August 2010 14 / 44

slide-15
SLIDE 15

Related Work and Contributions

Some recent related Work (incomplete list):

◮ Spectrum Acquisition (investment) only: ⋆ Auction: JiaZhangZhangLiu2009, SenguptaChatterjee2009, ... ⋆ Dynamic leasing: Simeone et al. 2008, JayaweeraLi2009, ... ◮ Service provision (pricing) only: ⋆ NiyatoHossainHan2009, XingChandramouliCordeiro2007, ... ◮ Joint investment and pricing: JiaZhang2008, NiyatoHossain2007 Jianwei Huang (NCEL) CVNO Games August 2010 15 / 44

slide-16
SLIDE 16

Related Work and Contributions

Some recent related Work (incomplete list):

◮ Spectrum Acquisition (investment) only: ⋆ Auction: JiaZhangZhangLiu2009, SenguptaChatterjee2009, ... ⋆ Dynamic leasing: Simeone et al. 2008, JayaweeraLi2009, ... ◮ Service provision (pricing) only: ⋆ NiyatoHossainHan2009, XingChandramouliCordeiro2007, ... ◮ Joint investment and pricing: JiaZhang2008, NiyatoHossain2007

Our contributions

◮ Analytical study of the joint investment and pricing decisions ◮ Operators and users are heterogeneous ◮ Derive results based on a practical physical layer model Jianwei Huang (NCEL) CVNO Games August 2010 15 / 44

slide-17
SLIDE 17

Part I: Monopoly CVNO with Supply Uncertainty

Spectrum Owner Dynamic Spectrum Leasing Powe Se Users (transmitter‐receiv Useful S Spectrum Sensing Spectrum holes er Spectrum in Use Freq elling bandwidth ver node pairs ) pectrum Time Sensing

Jianwei Huang (NCEL) CVNO Games August 2010 16 / 44

slide-18
SLIDE 18

Two Investment Choices

Reliable spectrum leasing with high negotiated cost Unreliable spectrum sensing with low energy/time cost

Sub Channels −

' Operator s Sensed Band

Leased Band ' Operator s ' PUs Band Activity 1 t = ( ) f HZ

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

2 t = Band Spectrum Band ' Owner s Spectrum ' Owner s Transference Service

Jianwei Huang (NCEL) CVNO Games August 2010 17 / 44

slide-19
SLIDE 19

Four-Stage Stackelberg Game

Operator (leader) Users (followers)

Sensing Realization α Stage I Sensing Bandwidth Bs (with unit cost Cs) Stage II Leasing Bandwidth Bl (with unit cost Cl) Stage III Pricing π Stage IV User Demand {wi}

Jianwei Huang (NCEL) CVNO Games August 2010 18 / 44

slide-20
SLIDE 20

Backward Induction & Subgame Perfect Equilibrium

Backward Induction Subgame 1 Subgame 2 Subgame 3 Subgame 4 Stage I Sensing Bandwidth Bs (with unit cost Cs) Stage II Leasing Bandwidth Bl (with unit cost Cl) Stage III Pricing π Stage IV User Demand {wi}

Jianwei Huang (NCEL) CVNO Games August 2010 19 / 44

slide-21
SLIDE 21

Stage IV: Users’ Bandwidth Demands

Physical layer model: users share the spectrum using OFDM

◮ No interferences ◮ Users request bandwidth from the operator

User k’s wireless characteristics: gk = Pmax

k

hk n0

◮ Pmax

k

: maximum transmission power

◮ hk: channel condition ◮ n0: background noise density

User k’s data rate rk(wk) = wk ln(1 + SNRk) = wk ln

  • 1 + Pmax

k

hk n0wk

  • Jianwei Huang (NCEL)

CVNO Games August 2010 20 / 44

slide-22
SLIDE 22

Users’ Payoff Functions

Assumption: all users operate in the high SNR regime rk(wk) ≈ wk ln Pmax

k

hk n0wk

  • ◮ Will be relaxed later.

User k’s payoff uk(π, wk) = wk ln Pmax

k

hk n0wk

  • − πwk

Jianwei Huang (NCEL) CVNO Games August 2010 21 / 44

slide-23
SLIDE 23

Users’ Optimization Problems

User i’s Bandwidth Optimization Problem w∗

k (π) = arg max wk≥0 uk(π, wk) = gke−(1+π)

Jianwei Huang (NCEL) CVNO Games August 2010 22 / 44

slide-24
SLIDE 24

Users’ Optimization Problems

User i’s Bandwidth Optimization Problem w∗

k (π) = arg max wk≥0 uk(π, wk) = gke−(1+π)

SNR∗

k = gk/w∗ k = e1+π: same (fair) for all users

Payoff uk(π, w∗

k ) = gke−(1+π): linear in gk

Jianwei Huang (NCEL) CVNO Games August 2010 22 / 44

slide-25
SLIDE 25

Stages III, II and I

Stage III: operator optimizes over price π: RIII(Bl, Bs, α) = max

π≥0 min

  • π
  • k

w∗

k (π), π (Bl + Bsα)

  • −(BsCs + BlCl) .

Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

slide-26
SLIDE 26

Stages III, II and I

Stage III: operator optimizes over price π: RIII(Bl, Bs, α) = max

π≥0 min

  • π
  • k

w∗

k (π), π (Bl + Bsα)

  • −(BsCs + BlCl) .

Stage II: operator optimizes over leasing bandwidth Bl: RII(Bs, α) = max

Bl≥0 RIII(Bl, Bs, α).

Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

slide-27
SLIDE 27

Stages III, II and I

Stage III: operator optimizes over price π: RIII(Bl, Bs, α) = max

π≥0 min

  • π
  • k

w∗

k (π), π (Bl + Bsα)

  • −(BsCs + BlCl) .

Stage II: operator optimizes over leasing bandwidth Bl: RII(Bs, α) = max

Bl≥0 RIII(Bl, Bs, α).

Stage I: operator optimizes over sensing bandwidth Bs: max

Bs≥0 Eα∈[0,1] [RII(Bs, α)] .

◮ Assumption: sensing uncertainty α follows uniform distribution. ◮ Will be relaxed later. Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

slide-28
SLIDE 28

Equilibrium Summary

Unique equilibrium.

Sensing Cost Cs ≥ Cl

2 1−e−2Cl 4

≤ Cs ≤ Cl

2

Sensing B∗

s

BL∗

s

∈

  • Ge−(2+Cl), Ge−2

Sensing Factor α 0 ≤ α ≤ 1 0 ≤ α ≤ Ge−(2+Cl)/BL∗

s

α > Ge−(2+Cl)/BL∗

s

Leasing B∗

l

Ge−(2+Cl) Ge−(2+Cl) − BL∗

s α

Price π∗ 1 + Cl 1 + Cl ln

  • G

BL∗

s

α

  • − 1

User k’s SNR e(2+Cl) e(2+Cl)

G BL∗

s

α

User k’s Payoff gke−(2+Cl) gke−(2+Cl) gk(BL∗

s α/G)

Jianwei Huang (NCEL) CVNO Games August 2010 24 / 44

slide-29
SLIDE 29

Stage I: Equilibrium Sensing Decision B∗

s

Theorem (Equilibrium sensing bandwidth B∗

s )

Linear in wireless characteristics G =

k gk;

Threshold structure: senses only when sensing cost Cs is small;

◮ Threshold increases with leasing cost Cl. 0.2 0.3 0.4 0.5 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Sensing Cost Cs Bs

* /G

Cl=0.5 Cl=0.7 Cl=1 Jianwei Huang (NCEL) CVNO Games August 2010 25 / 44

slide-30
SLIDE 30

Stage II: Equilibrium Leasing Decision B∗

l

Theorem (Equilibrium leasing bandwidth B∗

l )

Linear in wireless characteristics G =

k gk;

Threshold structure: leases only when α is small;

◮ Threshold decreases with leasing cost Cl. ◮ Threshold increases with sensing cost Cs. 0.2 0.4 0.6 0.8 1 0.02 0.04 0.06 0.08 0.1 Sensing Reliazation Factor α Bl

*/G

L:(Cl, Cs)=(0.5, 0.2) L:(Cl, Cs)=(0.5, 0.1) L:(Cl, Cs)=(1, 0.1) Jianwei Huang (NCEL) CVNO Games August 2010 26 / 44

slide-31
SLIDE 31

Stage III: Equilibrium Price π∗

Theorem (Equilibrium price π∗) Independent of wireless characteristics G =

k gk;

Threshold structure: decreases when α is large;

◮ Threshold decreases with leasing cost Cl. ◮ Threshold increases with sensing cost Cs. 0.2 0.4 0.6 0.8 1 0.8 1 1.2 1.4 1.6 1.8 2 Sensing Reliazation Factor α Optimal Price π* L:(Cl, Cs)=(0.5, 0.2) L:(Cl, Cs)=(0.5, 0.1) L:(Cl, Cs)=(1, 0.1) Jianwei Huang (NCEL) CVNO Games August 2010 27 / 44

slide-32
SLIDE 32

Robustness of Results

To obtain closed form solutions, we have assumed

◮ All users achieve high SNR ◮ Sensing realization factor α follows uniform distribution

Previous observations still hold in the general case

◮ Users operate in general SNR regime: rk(wk) = wk ln

  • 1 + Pmax

k

hk n0wk

  • ◮ α follows any continuous distribution

Starting point: implicit function theorem

Jianwei Huang (NCEL) CVNO Games August 2010 28 / 44

slide-33
SLIDE 33

Impact of Sensing Uncertainty on Operator

Expected profit increases with sensing

0.2 0.4 0.6 0.8 1 0.01 0.02 0.03 0.04 0.05 0.06 0.07 Sensing Cost Cs Normalized Exp. Profit RI /G Cl=2, with sensing Cl=2, w./o. sensing Jianwei Huang (NCEL) CVNO Games August 2010 29 / 44

slide-34
SLIDE 34

Impact of Sensing Uncertainty on Operator

Realized profit increases with α

◮ Can be smaller than no sensing

Smaller Cs leads to more aggressive sensing and less reliable supply

0.2 0.4 0.6 0.8 1 −0.04 −0.02 0.02 0.04 0.06 0.08 0.1 Sensing Realization Factor α Normalized Realized Profit RII

CS1/2 (α)/G

Cl=2,Cs=0.8 Cl=2,w./o. sensing Cl=2, Cs=0.5 Jianwei Huang (NCEL) CVNO Games August 2010 30 / 44

slide-35
SLIDE 35

Impact of Sensing Uncertainty on Users

Users’ payoffs never decrease under sensing

0.2 0.4 0.6 0.8 1 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Sensing Realization Factor α Normalized Realized Payoff of User ui

*(α) /gi

Cl=2, Cs=0.8 Cl=2, Cs=0.5 Cl=2, w./o. sensing Jianwei Huang (NCEL) CVNO Games August 2010 31 / 44

slide-36
SLIDE 36

Part II: Duopoly CVNO Competition

Secondary users (transmitter-receiver pairs)

Spectrum

  • wner

Operator i Operator j Investment (leasing bandwidth) Pricing (selling bandwidth) Spectrum

  • wner

Jianwei Huang (NCEL) CVNO Games August 2010 32 / 44

slide-37
SLIDE 37

Three-Stage Multi-leader-follower Game

Stage I: Leasing Game Leasing Bandwidth B1 and B2 (with unit costs C1 and C2) Stage II: Pricing Game Pricing π1 and π2 Stage III User k Chooses One Operator i and Demand wki

Operators (leaders) Users (followers)

Backward Induction

Jianwei Huang (NCEL) CVNO Games August 2010 33 / 44

slide-38
SLIDE 38

Stage IIII: Users’ Bandwidth Demands

Similar as before: assume high SNR first User k’s payoff of choosing operator i = 1, 2 uk(πi, wki) = wki ln Pmax

i

hi n0wki

  • − πiwki

◮ Optimal demand: w ∗

ki(πi) = arg maxwki≥0 uk(πi, wki) = gke−(1+πi)

◮ Optimal payoff: uk(πi, w ∗

ki(πi))

User k prefers the “better” operator: i∗ = arg maxi=1,2 uk(πi, w∗

ki(πi))

Users demands may not be satisfied due to limited resource

◮ Difference between preferred demand and realized demand Jianwei Huang (NCEL) CVNO Games August 2010 34 / 44

slide-39
SLIDE 39

Stages II: Pricing Game

Players: two operators Strategies: πi ≥ 0, i = 1, 2 Payoffs: profit Ri for operator i = 1, 2: Ri(Bi, Bj, πi, πj) = πiQi(Bi, Bj, πi, πj) − BiCi

Jianwei Huang (NCEL) CVNO Games August 2010 35 / 44

slide-40
SLIDE 40

Stage II: Pricing Equilibrium

Symmetric equilibrium: π∗

1 = π∗ 2.

Threshold structure:

◮ Unique positive equilibrium exists B1 + B2 ≤ Ge−2.

( 1) M ( 3) M ( ) L ( 2) M ( ) H

2

Ge−

1

Ge−

i

B

j

B

2

Ge−

1

Ge−

(L) : Unique nonzero equilibrium (M1)‐(M3) : No equilibrium (H) : Unique zero equilibrium

Jianwei Huang (NCEL) CVNO Games August 2010 36 / 44

slide-41
SLIDE 41

Stage I: Leasing Game

Players: two operators Strategies: Bi ∈ [0, ∞), i = 1, 2, and B1 + B2 ≤ Ge−2. Payoffs: profit Ri for operator i = 1, 2: Ri(Bi, Bj) = Bi

  • ln
  • G

Bi + Bj

  • − 1 − Ci
  • Jianwei Huang (NCEL)

CVNO Games August 2010 37 / 44

slide-42
SLIDE 42

Stage I: Leasing Equilibrium

Linear in wireless characteristics G =

i gi;

Threshold structure:

◮ Low costs: infinitely many equilibria ◮ High comparable costs: unique equilibrium ◮ High incomparable costs: unique monopoly equlibrium

1

j i

C C = + ( ) L 1 1

i

C

j

C ( ) HC ( ) HI ( ') HI 1

j i

C C = − (L) : Infinitely many equilibria (HC) : Unique equilibrium (HI)‐(HI’) : Unique equilibrium

Jianwei Huang (NCEL) CVNO Games August 2010 38 / 44

slide-43
SLIDE 43

Equilibrium Summary (Assuming Ci ≤ Cj)

Costs LOW HC HI Ci + Cj ≤ 1 Ci + Cj > 1, Cj > 1 + Ci Cj − Ci ≤ 1 equilibria Infinite Unique Unique (B∗

i , B∗ j )

(ρGe−2, (1 − ρ)Ge−2),

  • (1+Cj−Ci)G

2e

Ci +Cj +3 2

, (1+Ci−Cj)G

2e

Ci +Cj +3 2

  • (Ge−(2+Ci), 0)

ρ ∈ [Cj, (1 − Ci)] (π∗

i , π∗ j )

(1, 1)

  • Ci+Cj+1

2

, Ci+Cj+1

2

  • (1 + Ci, N/A)

User SNR e2 e

Ci +Cj +3 2

e2+Ci User Payoff gke−2 gke

− “ Ci +Cj +3

2

”

gke−(2+Ci)

Users achieve the same SNR User k’s payoff is linear in gk

Jianwei Huang (NCEL) CVNO Games August 2010 39 / 44

slide-44
SLIDE 44

Robustness of Results

To obtain closed form solutions, we have assumed

◮ All users achieve high SNR

Previous observations still hold in the general case

◮ Users operate in general SNR regime: rki(wki) = wki ln

  • 1 + Pmax

k

hk n0wki

  • Jianwei Huang (NCEL)

CVNO Games August 2010 40 / 44

slide-45
SLIDE 45

Impact of Duopoly Competition on Operators

Benchmark: Coordinated Case

◮ Operators cooperate in investment and pricing to maximize total profit

Define Efficiency Ratio = Total Profit in Competition Case Total Profit in Coordinated Case Price of Anarchy = minCi,Cj Efficiency Ratio= 0.75.

Jianwei Huang (NCEL) CVNO Games August 2010 41 / 44

slide-46
SLIDE 46

Conclusions and Future Work

What we have shown

◮ New network architecture for cognitive radio networks ◮ Simple threshold decisions rules that are easy to implement Jianwei Huang (NCEL) CVNO Games August 2010 42 / 44

slide-47
SLIDE 47

Conclusions and Future Work

What we have shown

◮ New network architecture for cognitive radio networks ◮ Simple threshold decisions rules that are easy to implement

Future directions

◮ Different network architecture ◮ Different information structure ◮ Different time/frequency dynamics Jianwei Huang (NCEL) CVNO Games August 2010 42 / 44

slide-48
SLIDE 48

Related Publications

Part I: Lingjie Duan, Jianwei Huang, and Biying Shou, “Cognitive Mobile Virtual Network Operator: Investment and Pricing with Supply Uncertainty,” IEEE INFOCOM, San Diego, CA, USA, 2010. Journal submitted to IEEE Transactions

  • n Mobile Computing.

Part II: Lingjie Duan, Jianwei Huang, and Biying Shou, “Competition with Dynamic Spectrum Leasing,” IEEE DySPAN, Singapore, 2010. Journal submitted to IEEE Transactions on Networking.

Jianwei Huang (NCEL) CVNO Games August 2010 43 / 44

slide-49
SLIDE 49

Contact

ncel.ie.cuhk.edu.hk jwhuang@ie.cuhk.edu.hk

Jianwei Huang (NCEL) CVNO Games August 2010 44 / 44