8/22/2019 Cognitive Rrm Dasilva
1/27
06/ 05/ 2012
1
Cognitive Radio Resource
Management
Prof. Luiz DaSilva
Theory and Practice of Cognitive Radio
Aalborg University
May 9-11, 2012
About me...
8/22/2019 Cognitive Rrm Dasilva
2/27
06/ 05/ 2012
2
Objectivesof
Cognitive
Radio
ResourceManagement
ApproachesforAllocationof
WirelessResources
FunctionalArchitecturesfor
CognitiveRRM
RadioResourceManagement(RRM)
Efficientallocationofresourcesinawirelessnetwork
Orthogonalchannels,throughfrequency,time,code
oratleastallocationsthatresultintolerableamountof
interference
Achievedthrough
Transmitpowercontrol
Spectrum/channelallocation
Accesstothemedium/scheduling
Multiantennause(beamforming,MIMO)
Multitransceiveruse
8/22/2019 Cognitive Rrm Dasilva
3/27
06/ 05/ 2012
3
Goals
Goalsoftheindividualradio
MaximizeSINR,datarate,throughput
Mitigatetheeffectsofnoise,interference
Maximizeaccesstothemedium
Maximizebatterylife
Goalsofthenetwork
Ensureefficientuseofspectrumresources
Maximizeaggregatethroughput
Avoidunduecontentionforresources
Ensureappropriatecoverage
WhatisdifferentinCognitive RRM?
Greaterabilityoftheterminaltoreconfigure(softwaredefinedradio)
RadiosmaybecapableofworkingwithmultipleRadio
AccessTechnologies
(RATs)
Frequencyagility
Opportunisticuseofspectrum,ordynamicspectrummanagement
User
Service
Terminal
Network
RRM
8/22/2019 Cognitive Rrm Dasilva
4/27
06/ 05/ 2012
4
CognitiveRadioResourceManagement
Dynamicallocationofresources(e.g.,spectrum)whenradioshaveincreasedabilitytoadaptandreconfigure(frequencyagility,multipleRATs,etc.)
Higherdegreeofautonomyfromtheradios(radiosmayparticipateintheresourcemanagementdecisions)
Coexistence:competitionandcooperationintheuseofresources
Allocationof
traffic
to
different
RATs
that
are
supported
in
thenetwork
FlexibleSpectrumManagement
Coordinated
Verticalhandovers
Spectrumsharing/pooling
Spectrumbroker
Virtualoperators
Realtimespectrumauctions
Autonomous
Sharedbands Opportunisticmultiband
randomaccess
Cooperativeaccess,
coalition
formation
Licensedbands Secondaryaccess,interruptible
byprimaryuser
Femto/macrocellcoexistence
Dynamicmultihoming
8/22/2019 Cognitive Rrm Dasilva
5/27
06/ 05/ 2012
5
CoordinatedExample
Spectrumbrokercontrols
accesstopartofthespectrum
Grantsatimeboundlease
tooperators,specifying
spatialregion,maximum
power,exclusiveornon
exclusiveusagerights
Regionunderbroker
controlmayhavebase
stationsofseveral
providers
d
a
f
c
Autonomous,NonCooperative
Example1
Topologycontrolinadhoc
networks
Eachnodesetsits
transmissionpowerand
channelofoperationto
achievesome
connectivity
objective
Accesstothemediumis
random(unscheduled)
8/22/2019 Cognitive Rrm Dasilva
6/27
06/ 05/ 2012
6
Autonomous,NonCooperative
Example2
Opportunisticspectrumaccess
Secondaryuserdetects
whitespaceandshapesits
signaltofitintoit
Noexplicitcooperation
withincumbentisrequired,
aslongandinterference
doesnotexceedpre
determinedlimit 2.4096 2.4097 2.4098 2.4099 2.41 2.4101 2.4102 2.4103 2.4104x 109
60
50
40
30
20
10
0
Frequency (Hz)
Magnitude(dB)
Autonomous,Cooperative
Example
Cooperativespectrumsharing
Multiplesecondaryusersin
amultichannel
environmentnegotiate
spectrumusage
Informationdissemination
regardingprobabilityof
incumbentactivityand
economicincentivesmay
bepartofthenegotiation
8/22/2019 Cognitive Rrm Dasilva
7/27
06/ 05/ 2012
7
Approaches
Classicaloptimization
Gametheory
Heuristics
Metaheuristics
Multiobjectiveapproaches
andcombinationsofthese
Parameters
Physical
MAC
Network
Application
Transmitpower,modulation,channel
codingrate,channelselection,resource
blocks,
Transmissiontimeandchannel,service
rate,priority,
Routeselection,costmetric,cooperation,
Sourcecodingrate,priority,packetarrival
rate,
8/22/2019 Cognitive Rrm Dasilva
8/27
06/ 05/ 2012
8
OptimizationGoals
Physical
MAC
Network
Application
Minaggregate
power,
max
throughput,
maxrate/Joule,maxspectralefficiency,
minBER,
Maxaggregatethroughput,minbuffer
overflowprobability,mindelay,
Minroutelength,minenergyexpended,
Mindistortion,mindelay,
Constraints
Physical
MAC
Network
Application
Maxtransmitpowerpernodeand/or
channel,availablemodulation
constellation, availablechannelcoding
rate,
Limitedtime/frequencyslots,limited
informationaboutothermobiles,
Maxnumberofhops,
Limitedsourcerate,strictdelay
requirements,
8/22/2019 Cognitive Rrm Dasilva
9/27
06/ 05/ 2012
9
ClassicalOptimization
Findanoptimalallocationofresourcestomeetacertaingoal
Linear/nonlinear/integer/MILP
Mayassumecentralizeddecisionmakingand/orfull
information
Complexityissues
S
,,2,10)(
,,2,10)(
)(min
=
==
x
x
x
x
rjg
mihtosubject
f
j
i
K
K
ClassicalOptimizationExample
Given
Asetofsourcedestinationpairsinthenetworkand
associateraterequirements
Findtheoptimal
Assignmentoffrequencybandstoeachpair
Schedulingofsubbandsfortransmissionandreception
Multihoprouting
Thatminimizes
Thetotalbandwidthusedinthenetwork
8/22/2019 Cognitive Rrm Dasilva
10/27
06/ 05/ 2012
10
ClassicalOptimizationExample:Formulation
Y.T.Hou,Y.Shi,andH.Sherali,Spectrum
SharingforMultihop Networkingwith
CognitiveRadios,IEEEJSAC, Jan.2008.
DistributedOptimization
Optimizationproblemswhereeachvariableandconstraintis
ownedbyanagent
Consideranetworkofnagents:
Agentswanttocooperativelyfind:
Objectivefunction isknowntoagenti
Agentsperiodicallyexchangeinformationandestimatetheir
impactontheobjectivefunction
Moretailoredtodistributedscenarioswithincomplete
information
},,1{ nN K=
)(min1
xx =n
i
iR fm
RRf m
i :)(x
8/22/2019 Cognitive Rrm Dasilva
11/27
06/ 05/ 2012
11
GameTheory
Mathematicalmodelsofconflictandcooperationamong
intelligent,rationaldecisionmakers
Canbeusedtostudythedistributedresourceallocation
decisionsmadebydecisionmakers(networknodes,radio
providers)toachievesomegoal(maximizeperformance,
minimizeresourceutilization)
Agameconsistsof
Asetoftwoormoreplayers
Setsof
possible
actions
for
each
player
Asetofpreferencerelationshipsforeachplayerovereach
actiontuple
CooperativeandNonCooperativeGameTheory
Noncooperative
Playersmakedecisionsindependently
Cooperative
Playersareallowedtoenterintoagreements,form
coalitions,etc.
Externalentityassumedtoexisttoenforcecontracts
Itispossibletomodel/predicttheoutcomeofabargaining
processwithoutmodelingtheprocessitself
8/22/2019 Cognitive Rrm Dasilva
12/27
06/ 05/ 2012
12
NonCooperativeGameTheoryExample
SINRmaximizingpowercontrolgame
Setofplayers:cognitiveradiosinacertainregion
Setofactions:eachradio setsitstransmitpower to
maximizeitsutilityfunction
whereMisthestatisticalcrosscorrelationofthesignals
}{,, iuAN=
Ni ip
+=
}\{
)1()(
iNk
kk
iii
phM
phu
p
J.O.Neel,J.H.Reed,andR.P.Gilles,GameModelsfor
CognitiveRadioAlgorithmAnalysis,SDRForumTechnicalConf., Nov.2004.
CooperativeGameTheoryExample
Cooperativespectrumsharinggame
Setofplayers:cognitiveradiosinagivenregion
Setofactions:eachplayerselectsasubsetofatotalofK
channelstooccupy(eachwithbandwidthB/K),andthepower
allocationoneachofthosechannelstomaximizeutility
function
Radioscanbargainwithoneanothertoachieveanefficient
solutionJ.E.Suris,L.A.DaSilva,Z.Han,A.B.MacKenzie,andR.
Komali,AsymptoticOptimalityforDistributedSpectrum
SharingUsingBargainingSolutions,IEEETrans.WirelessCommunications,vol.8,no.10,Oct.2009.
2log (1 )i kk
Bu SINR
K= +
8/22/2019 Cognitive Rrm Dasilva
13/27
06/ 05/ 2012
13
TypicalResultsofGameTheoreticAnalysis
Equilibria
Stableoperatingpointsforthenetworkfromwhich,once
reached,noplayerhasanincentivetodeviate
Convergenceproperties(noncooperativegames)
Indicationofwhetherthroughanadaptiveprocessplayers
willarriveatanequilibrium
Expectedoutcomeofbargaining(cooperativegames)
Indicationofwhetheramoreefficientoutcomecanbe
arrivedat
through
cooperation
Designofincentives/disincentives
Examples:auctions,mechanismdesign
Heuristics
Approximationsthatseekagoodenough,ratherthanan
optimal,solutionfortheproblem
Appropriatewhen
Problemistoocomplexforanoptimizationsolutiontobe
foundinreasonableamountoftime
Environmentchanges
too
rapidly
Adaptationsandresourceallocationdecisionsmustbe
madeunderpartialinformation
Heuristicsolutioncanbeshowntodonotmuchworse
thantheoptimum
8/22/2019 Cognitive Rrm Dasilva
14/27
06/ 05/ 2012
14
HeuristicsExamples
Linearorparameterizedapproximationstoobjective
functions
Mayallownonconvexproblemstobetreatedasconvex
Maybereasonableapproximationsundersomeconditions
(highSNR,largenumberofchannels,etc.)
Simpleadaptationsthatcanbeshowntogivegoodresults
withhighprobability
Adaptationsbasedonlocalconditions,withouttakinginto
accountglobal
state
of
the
network
MetaHeuristics
Heuristicmethodsforsolvingaclassofcomputational
problemsforwhichthereisnopractical(e.g.,polynomial
time)solution
Usesomeblackboxproceduresthatarethemselves
heuristics
Examples
HillClimbingorgreedyAlgorithm
Tabu search
Simulatedannealing
Geneticalgorithms
8/22/2019 Cognitive Rrm Dasilva
15/27
06/ 05/ 2012
15
MetaHeuristicsExample
Islandgeneticalgorithmappliedtothechannelallocation
problemincognitivenetworks
Protocolmodelofinterference
Anislandgeneticalgorithmdividesthepopulationinto
subpopulations,orislands,thatinteractthroughthemigration
ofindividualstootherislands
Seekchannelassignmentthatmaximizessumcapacity
Startwithavalidchannelassignment,performmutationand
crossover
of
individuals
with
high
fitness
function
D.Friend,M.ElNainay,Y.Shi,andA.B.MacKenzie,
Architectureand PerformanceofanIslandGenetic
AlgorithmBasedCognitiveNetwork,Proc.IEEEConsumerComm.AndNetworkingConf.,pp.993997,2008.
MultiObjectiveOptimization
Alternatives
Combineallobjectivefunctionsintoanaggregateobjective
function,suchasaweightedsum(buthowtoassign
weights?)
LookforsolutionsintheParetofrontier
[ ]
S
,,2,10)(
,,2,10)(
)()()(min 21
=
==
x
x
x
xxx
rjg
mihtosubject
j
i
T
p
K
K
K
8/22/2019 Cognitive Rrm Dasilva
16/27
06/ 05/ 2012
16
TheParetoFrontier
Givenanallocationofresourcesinvolvingsometradeoff,say
performanceandcost,aParetoimprovementisamovementfromoneallocationtoanotherthatcanmakeatleastone
individualbetteroffwithoutmakinganyindividualworseoff
AsolutionisParetooptimalorParetoefficientifnoParetoimprovementispossible
Thesetofallsolutionstoamultiobjectiveoptimizationthat
areParetooptimalistheParetofrontier No
objective
can
be
further
improved
without
making
at
leastoneotherobjectiveworseoff
ParetoFrontier:Visualization
8/22/2019 Cognitive Rrm Dasilva
17/27
06/ 05/ 2012
17
ConsiderationsintheSelectionofanApproach
Timescaleinwhichresourceallocationmustbemadeand
timescaleinwhichenvironmentchanges
Fairlystatic(optimization)versusdynamic(heuristics)
Centralized(optimization)versusdecentralized(gametheory,
distributedoptimization,heuristics)
Processingcomplexity
Realtimeadaptations(heuristics)versusestablishmentof
baselineidealresults(optimization)
IEEE1900.4
ArchitectureandEnablersforOptimizedRadioandSpectrumResourceUsage
DefinesamanagementsystemthatdecidesactionsrequiredtooptimizeradioresourceusageandQoS inheterogeneouswirelessenvironments
Spectrumassignment(carrierfrequency,signalbandwidth,
radiointerface
used
in
the
assigned
spectrum)
to
radio
accessnetworks(RANs)canbedynamicallychanged OrspectrumassignmenttoRANsisfixed,butaRANcan
operateconcurrentlyinmultiplebands Terminalsarereconfigurableandmayormaynotbecapable
ofmultihoming Multihoming =capabilitytosimultaneouslymaintain
morethanoneactiveconnectionwithRANs
8/22/2019 Cognitive Rrm Dasilva
18/27
06/ 05/ 2012
18
IEEE1900.4Resource
Management
Reconfigurationdecisionsare
madebytheterminal(TRM)
andbythenetwork(NRM)
Logicalcommunicationchannel
betweenTRMandNRMcanbe
realizedinband(usingan
existingRANalsousedfordata)
oroutofband(specifyinga
dedicatedphysicalchannelfor
thiscommunication)
S.Buljoreetal.,ArchitectureandEnablersforOptimizedRadioResourceUsageinHeterogeneousWirelessAccess
Networks,IEEECommunicationsMagazine, Jan.2009.
IEEE1900.4ReferenceUsage
Cases
Dynamicspectrumassignmentexample:aspectrumbandisshifted
from802.16eto802.11n
duetoshiftsindemand
Dynamicspectrumsharingexample:802.22systems
usingVHF/UHFbands
opportunistically
Distributedradioresourceusageoptimizationexample:jointoptimization
ofresourceassignmentby
terminalsandnetwork
S.Buljoreetal.,ArchitectureandEnablersforOptimizedRadioResourceUsageinHeterogeneousWirelessAccess
Networks,IEEECommunicationsMagazine, Jan.2009.
8/22/2019 Cognitive Rrm Dasilva
19/27
06/ 05/ 2012
19
IEEE1900.4SystemArchitecture
Terminal context
information
(terminal
capabilities,
user preferences,
required QoS,
measurements,
geo-location
info, )
RAN context
information(optimization
objectives,
radio
capabilities,
measurements,)
Radio resource
selection
policies
IEEE1900.4FunctionalArchitecture
8/22/2019 Cognitive Rrm Dasilva
20/27
06/ 05/ 2012
20
BringingDynamicSpectrum
Accessto
4G
(LTE+)
and
Beyond
Licensedspectrumis
augmentedwithdynamic
accesstoadditionalbands
CoordinatedbySpectrum
AccountabilityServer(SAS)
Signallingtocontrolaccessto
additionalbandsusesthe
licensedcarriers
Datatransportedoverlicensed
anddynamicallyallocated
carriers
J.Deaton,R.IrwinandL.A.DaSilva,TheEffectsofa
DynamicSpectrumAccessOverlayinLTEAdvanced
Networks,Proc.ofIEEEDySPAN,2011
SAS
cBS
Spectr umAuct i on
Li censed Car r i er
BCCHCCCHDCCHDTCHPCCH
DCCHDTCHPCCH
CONNCONNCONNCONN IDLEIDLEIDLEIDLE
CONNCONNCONNCONN
BCCH: Broadcast Control Channel
CCCH: Common Control Channel
DCCH: Dedicated Control Channel
DTCH: Dedicated Traffic Channel
PCCH: Paging Control Channel
ArchitectureforDSAinLTE+
HSS
PDG
( SGW+PGW)
I nter net
MME
UE
E- UTRAN
eNBeNB
eNB
eNB
eNB
eNB
eNB
GDB
cUE
SAS
cRAN
S1- U
S1- MME
S1- U
S1- MME
S11
Di amet er ( RFC 3588)
eNB: Evolved Node B
E-UTRAN: Evolved Universal Terrestrial Radio Access
HSS: Home Subscriber Server
MME: Mobility Management Entity
PDG: Packet Data Gateway
PGW: Packet Gateway
SGW: Signaling Gateway
LTE Net wor k El ement s
cBS: cognitive Base Station
cRAN: cognitive Radio Access Network
cUE: cognitive User Equipment
GDB: Geolocation Data Base
SAS: Spectrum Accountability Server
SA Net wor k El ement s
cBS
cBScBS
cBS
cBS
cBS
cBS
8/22/2019 Cognitive Rrm Dasilva
21/27
06/ 05/ 2012
21
ArchitecturalComponents
cBS
I P Reachabl eSpect r umTrendi ngTr af f i c Trendi ng
Spect r umAgi l e Radi oRRC Spect r umSensi ng
SAP: Spect r umAccount i ng Prot ocol
X2e- Cooperat i ve SenseX2e- Spectr umTradi ng
PDG
I P Gat ewayI P Tunnel i ng t o cUE
SAS
Spect r umPol i cy &Leasi ng
SAP: Spect r umAccount i ng Protocol
Spectr umAccount abi l i t y
I nter f erence Al arms
cUE
Spect r um Agi l e Radi oRRC - Spect r umSensi ng
I RI P Connect TVSer vi ce Loss
Detect i onSer vi ce Loss
Repor t i ng
SignallingEndpoints
H- PDGH- cBS
I R
I nt ernet
GDB
SAS
N- PDGN- cBS
I R
I R
OPERATOR A
OPERATOR B
Networ k Operat orsRegul ator Agenci es
DSA Stat i st i cs
H- cUE
H- cUE
N- cUE
N- cUE
X2e: Cooper at i ve Sensi ngX2e: Spectr um Tradi ng
H- : HomeN-: Nei ghborI R: I nt egrat ed Recei verSAP: Spect r umAccountabi l i t y Pr ot ocol
8/22/2019 Cognitive Rrm Dasilva
22/27
06/ 05/ 2012
22
cBS RegistrationandNeighbourDiscovery
Update Local DB
SAP: cBS Registration Request
IP Addr., Geolocation
SAP: cBS Registration Response
Registration Successful
SAP: cBS Neighbor Request
cBS Neighbors Not Known
SAP: cBS Neighbor Response
Request Successful
SASH-cBS
Create Spectrum Account for H-cBS
Find cBS Neighbors
N-cBS
Update Neighbor Tables
X2e: Link Setup Request
Sent to each neighbor
X2e: Link Setup Response
Request Successful
SpectrumLeaseRequestProcedure
SAScBScUE
SAP: Spectrum Lease Request
Calculate Lease Request
Evaluate Spectrum Availability
SAP: Spectrum Release Order
SAP: Spectrum Release ACK
Update Spectrum Account with Lease
information
Service Requests
SAP: Spectrum Lease Response
Spectrum available
Procedure Note:
This process could be executed during the maintenance window or on demand on a call by call basis.
Demand Trigger:
8/22/2019 Cognitive Rrm Dasilva
23/27
06/ 05/ 2012
23
SpectrumSharingProcedure
H-cBScUE
SAP: Spectrum Lease Response
Spectrum unavailable
Evaluate Spectrum Availability
SAP: Spectrum Lease Request
X2e-ST: Spectrum Release Order
X2e-ST: Spectrum Release ACK
Add Spectrum to Available Pool
SAS N-cBS
X2e-ST: Spectrum Lease Request
Evaluate Spectrum Availability
X2e-ST: Spectrum Lease Response
Spectrum available
Service Request
Procedure Notes:
This process could be done during the maintenance window or on demand
Calculate Lease Request
Demand Trigger:
ServiceRequestandReportingProcedure
RRC Carrier Optimization
RRC: Connection Request
Licensed spectrum
SAScBSIDLE_cUE
Determine Carrier for Use
Service Response
Carrier Assignment Available
cUE Service
Change Carrier
Report Service KPI
Successful/Dropped/Lost/BlockedService
Record and Report Statistics
CONN_cUEs
RRC: Carrier Handoff Order
Channel reassignment
RRC: Carrier Handoff Complete
Determine Carrier Use and
Availability
Procedure Notes:
This process is done on
demand.
The RRC optimization
procedure is one of the first
open issues.
Reporting Service Usage could
also be combined in the
spectrum lease release
SAP: Report Service KPI
Piggybacked on Spectrum ReleaseACK or Periodic
8/22/2019 Cognitive Rrm Dasilva
24/27
06/ 05/ 2012
24
PerformanceAnalysis:Scenario
cBS
cBS
cBS
Demand
Demand
Demand
10MHzLicensedCarrier
10MHzLicensedCarrier
10MHzLicensedCarrier
10MHzSharedDSACarrier
Service
Service
Service
OverflowTraffic
PoissonArrivals
With
ParetoSessionDemand
DSAChannelAvailability
BetaDistribution
Ti me
Packet
Size
Par et o Demand
Poi sson Ar r i val s
PerformanceAnalysis:SpectrumAvailability
Ti me
Spectrum
200 kHz
1 sec
Spect r umAvai l abi l i ty Prof i l e
OFF ON
Modi f i ed Beta Di st r i but i on
8/22/2019 Cognitive Rrm Dasilva
25/27
06/ 05/ 2012
25
PerformanceAnalysis:Results
HeterogeneousNetworks
Operatordeployedand/oruserdeployedsmallcells
Significantimprovementinoverallcapacity
SmallcellsaddcomplexitytoRRMproblem
Verticalhandover
Femto
femto and
macro
femto channel
allocation
and
interference
TheRRMtaskalsobecomesincreasinglydecentralized
Lessplanningofinfrastructuredeployment
8/22/2019 Cognitive Rrm Dasilva
26/27
06/ 05/ 2012
26
J.Deaton,R.Irwin,andL.DaSilva,Supporting
DynamicSpectrumAccessinHeterogeneousLTE+
Networks,underreview,2012.
IncorporatingIndoorSmallCells
Radioresourcemanagementreferstothe
allocationofwirelessresourcestofulfill
networkanduserobjectives
Incognitiveradionetworks,musttakeinto
accountincreasedreconfiguration
capabilitiesoftheradios,frequencyagility,
coexistence
between
primary
and
secondary
users,multipleradioaccesstechnologies
Essentially,anoptimizationproblem
Functionalarchitecturesbeingstandardized
(IEEE1900)andproposedintheliteraturefor
4Gnetworksandbeyond
8/22/2019 Cognitive Rrm Dasilva
27/27
06/ 05/ 2012
About CTVR www.ctvr.ie
On email [email protected]
Papers luizdasilva.wordpress.com
Wireless @ Virginia Tech wireless.vt.edu
Ontheinterwebs