Dynamic Spectrum Access and the Role of

Game Theory

Luiz A. DaSilva!Professor of Telecommunications CONNECT, Trinity College Dublin

Riunione Annuale 2015 dell'Associazione Gruppo Nazionale Telecomunicazioni e Tecnologie dell’Informazione (GTTI)

L’Aquila, Italy, 18 June 2015

To  evolve  future  wireless  networks:

More  spectrum  (e.g.,  mm-­‐wave,  licensed  +  unlicensed)  More  antennas  (massive  MIMO)  More  technologies

New  spectrum  licensing  regimes  Cell  densifica@on  Sharing  of  infrastructure,  backhaul,  processing,  storage  Virtualised  wireless  networks

Wireless  networks  of  the  future  will  be  characterised  by  heterogeneity  of  spectrum  usage  regimes  of  ownership  models  of  radio  access  technologies  

where  resources  are  shared  and  orchestrated  to  create  bespoke,  virtual  networks  designed  for  specific  services

[Doyle, Forde, Kibilda, DaSilva, Proc. of IEEE, 2014]

Resource allocation and management

Who gets what resources when

Dynamic spectrum access

Incentives Cooperative communications Coexistence (between equals or hierarchical)

Pricing How much to charge for resources

A  set  of  analy@cal  tools  from  economics  and  mathema@cs  to  predict  the  outcome  of  complex  interac@ons  among  ra@onal  en@@es

Game theory

Non-cooperative, cooperative game theory

vs.

Single-stage, multi-stage

Equilibrium  concepts  Pareto  op@mality  Stackleberg  games  Bargaining  solu@ons  Best/beOer  response    Disagreement  point  Mechanism  design

Basic concepts

players

A game: +

action sets

+

utility functions

More  sophis4cated  game  theore4c  models…

Hierarchy  of  decision  makers  Stackelberg  games

Uncertainty  as  to  player  types  Bayesian  games

Sub-­‐set  of  players  coopera@ng  Coali@on  games

SeSng  the  rules  of  the  game  Mechanism  design

Consider  a  network  where  nodes  [players]  set  transmit  power  levels  autonomously  [ac@ons]    What  is  the  appropriate  u@lity  func@on?  

A  func@on  of  transmit  power  and  SIR/SINR

Power control games

A candidate utility function [Shah, Mandayam, Goodman, ’98]

If  receivers  are  modelled  as  co-­‐located,  it  is  possible  to  derive  the  Nash  equilibrium  

However,  it  is  not  Pareto  efficient  A  u@lity  that  also  contains  a  term  for  per-­‐unit  price  of  transmit  power  does  somewhat  beOer  Note:  posi@ve  versus  norma@ve  models  of  u@lity

Consider  a  CDMA  system,  with  users  [players]  differen@ated  by  their  choice  of  spreading  codes  [ac@ons]  The  u@lity  will  be  related  to  the  orthogonality  of  the  spreading  code  selec@ons  

SINR  for  a  correla@on  receiver  It  turns  out  that  the  game  can  be  formulated  as  an  ordinal  poten@al  game  

Desirable  proper@es  of  NE  Convergence  to  the  NE  through  beOer  response  /  best  response  dynamics

Interference avoidance games

[Menon et al., ’09]

❶  Primary  users  (PUs)  can  charge  secondary  users  (SUs)  for  access  to  spectrum  !❷  SUs  distributedly  select  on  which  sub-­‐bands  to  operate  

 Mul@ple  SUs  can  occupy  the  same  sub-­‐band  and  cooperate  in  communica@ng  

!❸  SUs  control  their  transmit  power  !  Model  as  inter-­‐related  Stackelberg  game  and  coali@on  forma@on  game  

!  Derive  an  algorithm  to  arrive  at  the  NE  for  the  individual  games  and  the  SE  for  the  hierarchical  game

Hierarchical spectrum sharing

[Xiao, Bi, Niyato, DaSilva, JSAC’12]

N  transmiOer/receiver  pairs  [players]  Bandwidth  B  divided  into  K  channels  TransmiOers  allocate  power  over  the  K  channels  [ac@ons],  subject  to  a  maximum  power  constraint  U@lity  func@on  is  the  aggregate  capacity  

Depends  on  interference  levels  in  each  channel  

U@lity  space  convexifies  as  K  increases  Distributed  algorithm  to  arrive  at  the  Nash  bargaining  solu@on,  by  exchanging  informa@on  in  the  neighbourhood  

Cooperative spectrum sharing

[Suris, DaSilva, Han, MacKenzie, Somali, TWC’09]

N  transmiOer/receiver  pairs  [players]    Channel  selec@on  and  transmit  power  [ac@ons]  U@lity  can  include  network-­‐wide  spectrum  efficiency,  fairness,  network  connec@vity  Study  the  coali@on  forma@on  process

Coalitions for resource sharing

[Khan, Glisic, DaSilva, Lehtomakki, TCIAIG’10]

D2D  links  [players]  compete  for  sub-­‐bands  occupied  by  a  cellular  subscriber  (if  interference  is  tolerable)  or  for  a  sub-­‐band  for  exclusive  use  (otherwise)  Mul@ple  D2D  links  can  share  a  sub-­‐band  D2D  links  do  not  know  about  others’  preferences,  loca@on,  link  condi@ons  Bayesian  non-­‐transferable  u@lity  overlapping  coali@on  forma@on  game  Propose  a  hierarchical  matching  algorithm  to  achieve  a  stable,  unique  matching  structure

[Xiao, Chen, Yuen, Han, DaSilva, TWC’15]

Suppor4ng  D2D  communica4ons  in  cellular  bands

Subscribers  [players]  dynamically  request  channels  of  operators  Bayesian  game:  subscribers  are  unaware  of  each  other’s  preferences  

Belief  func@ons,  learning  Matching  market:  subscribers  are  matched  to  operators,  then  to  sub-­‐bands  controlled  by  the  operator  Mechanism  incen@vises  truth-­‐telling

Matching subscribers to operators

[Xiao, Han, Chen, DaSilva, JSAC’15]

Inter-­‐operator  sharing  and  virtualised  wireless  networks

Games  between  operators    How  much  infra-­‐structure  to  deploy  individually  and  how  much  to  deploy  collec@vely?  Spectrum  versus  infra-­‐structure  sharing  

Games  between  operators  and  over-­‐the-­‐top  service  providers  

Should  the  OTT  deploy  its  own  infra-­‐structure?

Single-stage, non cooperative

initial works on power and interference games potential games, reach equilibrium via best response

Multi-stage games

can incorporate rewards, punishment for prior behaviour routing, peer-to-peer services, sharing repeated games, dynamic games, Markov games

Stackelberg games hierarchy in decision-making primary/secondary use of spectrum

Cooperative games opportunity for bargaining spectrum sharing among equals

Coalition formationcooperation enabled with a subset of all players D2D communication, spectrum sharing decisions

Mechanism designdevelopment of incentive-compatible mechanisms spectrum auctions, matching between subscribers and providers

Also: games of imperfect information, games of imperfect monitoring, evolutionary game theory, etc.

• Game theory is being used to model increasingly complex interactions among autonomous decision makers

• Models are particularly tailored to autonomous decision making and reasoning by different network entities - in line with trends in wireless networks (HetNets, D2D, resource sharing, etc.)

• Machine learning meets game theory: some learning processes can be shown to converge to Nash equilibria (e.g., application of learning automata to dynamic channel selection)

• Models can be applied at different scales: individual transmissions by nodes, longer-term decisions by transmitters or by users, interactions among networks, operators, etc.

Y. Xiao, K.-C. Chen, C. Yuen, Z. Han, and L. A. DaSilva, “A Bayesian Overlapping Coalition Formation Game for Device-to-Device Spectrum Sharing in Cellular Networks,” IEEE Transactions on Wireless Communications, 2015 (to appear).

Z. Khan, J. J. Lehtomäki, L. A. DaSilva, E. Hossain, and M. Latva-aho, “Opportunistic Channel Selection by Cognitive Wireless Nodes under Imperfect Observations and Limited Memory: A Repeated Game Model,” IEEE Transactions on Mobile Computing, 2015 (to appear).

Y. Xiao, Z. Han, K.-C. Chen, and L. A. DaSilva, “Bayesian Hierarchical Mechanism Design for Cognitive Radio Networks,” IEEE Journal on Selected Areas in Communications (JSAC), vol. 33, no. 5, pp. 986-1001, May 2015.

H. Ahmadi, Y. H. Chew, N. Reyhani, C. C. Chai, and L. A. DaSilva, “Learning Solutions for Auction-based Dynamic Spectrum Access in Multicarrier Systems,” Computer Networks, vol. 67, pp. 60-73, July 2014.

Y. Xiao, G. Bi, D. Niyato, and L. A. DaSilva, “A Hierarchical Game Theoretic Framework for Cognitive Radio Networks,” IEEE Journal on Selected Areas in Communications (JSAC), vol. 30, no. 10, November 2012, pp. 2053-2069.

Z. Khan, S. Glisic, L. A. DaSilva, and J. Lehtomaki, “Modeling the Dynamics of Coalition Formation Games for Cooperative Spectrum Sharing in an Interference Channel,” IEEE Trans. on Computational Intelligence and AI in Games, vol. 3, no. 1, Mar. 2011, pp. 17-30.

J. E. Suris, L. A. DaSilva, Z. Han, A. B. MacKenzie, and R. S. Komali, “Asymptotic Optimality for Distributed Spectrum Sharing Using Bargaining Solutions,” IEEE Trans. on Wireless Communications, vol. 8, no. 10, Oct. 2009, pp. 5225-5237.

V. Srivastava, J. Neel, A. MacKenzie, R. Menon, L.A. DaSilva, J. Hicks, J.H. Reed and R. Gilles, “Using Game Theory to Analyze Wireless Ad Hoc Networks,” IEEE Communications Surveys and Tutorials, vol. 7, no. 4, pp. 46-56, 4th quarter 2005.

I. Macaluso, L. A. DaSilva, and L. E. Doyle, “Learning Nash Equilibria in Distributed Channel Selection for Frequency-Agile Radios,” ECAI 2012 Workshop on Artificial Intelligence for Telecommunications and Sensor Networks (WAITS), Montpellier, France, August 27-31, 2012, pp. 7-10

luizdasilva.wordpress.com