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

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

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

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

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

NCEL: Network Communications & Economics Lab 4 postdocs 7 graduate students Jianwei Huang (NCEL) CVNO Games August 2010 5 / 44

Research Themes Wireless Communications Network Economics Distributed Optimization Game Theory Jianwei Huang (NCEL) CVNO Games August 2010 6 / 44

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

Cognitive Virtual Network Operator Games Jianwei Huang (NCEL) CVNO Games August 2010 8 / 44

� � � � � Mobile Virtual Network Operators Spectrum Acquisi.on � Service Provision � Spectrum
Owner � V NO Users � 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

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

Spectrum is a Scarce Resource Jianwei Huang (NCEL) CVNO Games August 2010 11 / 44

Spectrum is a Under-Utilized c � Share Spectrum Co. Jianwei Huang (NCEL) CVNO Games August 2010 12 / 44

Cognitive Virtual Network Operators Cognitive Virtual Network Operators ( C VNOs) ◮ 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

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

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

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

Part I: Monopoly CVNO with Supply Uncertainty Spectrum in Use Powe er Freq Sensing Time Spectrum Owner Spectrum holes Dynamic Spectrum Spectrum Sensing Leasing Useful S pectrum Se elling bandwidth Users (transmitter ‐ receiv ver node pairs ) Jianwei Huang (NCEL) CVNO Games August 2010 16 / 44

Two Investment Choices Reliable spectrum leasing with high negotiated cost Unreliable spectrum sensing with low energy/time cost Spectrum Owner s ' Spectrum Owner s Transference ' Band Service Band f HZ ( ) t = 1 t = 2 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 1 2 3 4 5 6 7 8 − Sub Channels PUs ' Activity Band Operator s Sensed Band Operator s ' Leased Band ' Jianwei Huang (NCEL) CVNO Games August 2010 17 / 44

Four-Stage Stackelberg Game Stage I Sensing Bandwidth B s (with unit cost C s ) Sensing Realization α Stage II Leasing Bandwidth B l Operator (with unit cost C l ) (leader) Stage III Pricing π Users Stage IV (followers) User Demand {w i } Jianwei Huang (NCEL) CVNO Games August 2010 18 / 44

Backward Induction & Subgame Perfect Equilibrium Stage I Sensing Bandwidth B s (with unit cost C s ) Subgame 4 Stage II Leasing Bandwidth B l Backward Induction (with unit cost C l ) Subgame 3 Stage III Pricing π Subgame 2 Subgame 1 Stage IV User Demand {w i } Jianwei Huang (NCEL) CVNO Games August 2010 19 / 44

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: g k = P max h k k n 0 ◮ P max : maximum transmission power k ◮ h k : channel condition ◮ n 0 : background noise density User k ’s data rate 1 + P max � h k � k r k ( w k ) = w k ln(1 + SNR k ) = w k ln n 0 w k Jianwei Huang (NCEL) CVNO Games August 2010 20 / 44

Users’ Payoff Functions Assumption : all users operate in the high SNR regime � P max h k � k r k ( w k ) ≈ w k ln n 0 w k ◮ Will be relaxed later. User k ’s payoff � P max h k � k u k ( π, w k ) = w k ln − π w k n 0 w k Jianwei Huang (NCEL) CVNO Games August 2010 21 / 44

Users’ Optimization Problems User i ’s Bandwidth Optimization Problem w ∗ w k ≥ 0 u k ( π, w k ) = g k e − (1+ π ) k ( π ) = arg max Jianwei Huang (NCEL) CVNO Games August 2010 22 / 44

Users’ Optimization Problems User i ’s Bandwidth Optimization Problem w ∗ w k ≥ 0 u k ( π, w k ) = g k e − (1+ π ) k ( π ) = arg max SNR ∗ k = g k / w ∗ k = e 1+ π : same (fair) for all users k ) = g k e − (1+ π ) : linear in g k Payoff u k ( π, w ∗ Jianwei Huang (NCEL) CVNO Games August 2010 22 / 44

Stages III, II and I Stage III: operator optimizes over price π : � � � w ∗ R III ( B l , B s , α ) = max π ≥ 0 min π k ( π ) , π ( B l + B s α ) − ( B s C s + B l C l ) . k Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

Stages III, II and I Stage III: operator optimizes over price π : � � � w ∗ R III ( B l , B s , α ) = max π ≥ 0 min π k ( π ) , π ( B l + B s α ) − ( B s C s + B l C l ) . k Stage II: operator optimizes over leasing bandwidth B l : R II ( B s , α ) = max B l ≥ 0 R III ( B l , B s , α ) . Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

Stages III, II and I Stage III: operator optimizes over price π : � � � w ∗ R III ( B l , B s , α ) = max π ≥ 0 min π k ( π ) , π ( B l + B s α ) − ( B s C s + B l C l ) . k Stage II: operator optimizes over leasing bandwidth B l : R II ( B s , α ) = max B l ≥ 0 R III ( B l , B s , α ) . Stage I: operator optimizes over sensing bandwidth B s : max B s ≥ 0 E α ∈ [0 , 1] [ R II ( B s , α )] . ◮ Assumption : sensing uncertainty α follows uniform distribution. ◮ Will be relaxed later. Jianwei Huang (NCEL) CVNO Games August 2010 23 / 44

Equilibrium Summary Unique equilibrium. 1 − e − 2 Cl C s ≥ C l ≤ C s ≤ C l Sensing Cost 2 4 2 B L ∗ Ge − (2+ C l ) , Ge − 2 � Sensing B ∗ � 0 ∈ s s 0 ≤ α ≤ Ge − (2+ C l ) / B L ∗ α > Ge − (2+ C l ) / B L ∗ Sensing Factor α 0 ≤ α ≤ 1 s s Ge − (2+ C l ) − B L ∗ Leasing B ∗ Ge − (2+ C l ) s α 0 l � � G Price π ∗ 1 + C l 1 + C l ln − 1 B L ∗ α s e (2+ C l ) e (2+ C l ) G User k ’s SNR B L ∗ α s g k e − (2+ C l ) g k e − (2+ C l ) g k ( B L ∗ User k ’s Payoff s α/ G ) Jianwei Huang (NCEL) CVNO Games August 2010 24 / 44