Wireless RevolutionWireless Revolution
RFID ApplicationsRFID Applications
V iV i
InternetInternet
Wireless EverywhereWireless EverywhereVoiceVoice EmailEmailVariousVarious
Info/MediaInfo/Media New MobileNew Mobile
Wireless EverywhereWireless EverywhereMoving Towards Moving Towards TrillionsTrillions
SMSSMSDevicesDevicesof Wireless Devicesof Wireless Devices
19971997 20072007 FutureFuture
VideoVideoMobile TVMobile TV Sensor Sensor
ApplicationsApplications997997 77 utu eutu e
0.20 Billion Mobile devices
1.8 Billion Mobile devices
Trillion’sWireless devices
Evolvement of DevicesEvolvement of DevicesSingle Radio Multi‐Radio (multi‐card) Cognitive Radio Single Radio Multi Radio (multi card) Cognitive Radio (single chip)
• Wireless Combo 11a/b/g card widely being used • Multi-radio smartphones available at the end of year 2004Multi radio smartphones available at the end of year 2004
• NTT DoCoMO, Motorola, Nokia
• Intel’s Cognitive Radio Chip• All-in-one chipset 802 11b/g and 802 11a• All-in-one chipset 802.11b/g and 802.11a• Single radio chip for: Wifi, Wimax, bluetooth, UWB …
• “XG1TM” by adapt4, inc
One Radio / One Network Multiple Radios / Networks One Single Cognitive Radio
ChallengesChallenges
h ll b ll l dBy 2020, there will be trillions wireless devicesChallenge for spectrum efficiencyChallenge for network architecture
Future wireless networks need ~100‐1000x increases in density and bit‐rate of radios
Motivates better spectrum coordination methodsMotivates better spectrum coordination methods
Spot shortages of spectrum will occur if present static allocation is continuedallocation is continued
Spectrum Spectrum EverywhereSpectrum, Spectrum, Everywhere, but . . .but . . .
Fixed allocationLittle sharingRigid requirement on how to use
Predicted Spectrum Requirements byPredicted Spectrum Requirements by 2020 on Regional Basis
Region 1 Region 2 Region 3
User Demand
Predicted Total
Already Identified
Net Addition
Already Identified
Net Addition
Already Identified
Net Addition
Setting (MHz) (MHz) (MHz) (MHz) (MHz) (MHz) (MHz)
LOW 1280 693 587 723 557 749 531
HIGH 1720 693 1027 723 997 749 971HIGH 1720 693 1027 723 997 749 971
-- Mobile Industry backing (mib) Terrestrial Spectrum for IMT
Spectrum Under UtilizationSpectrum Under‐Utilization
Most spectrum is unused actuallyMost spectrum is unused actually
11
i SDynamic Spectrum AccessS Spectrum users
Primary (PRI) user (with license) uses spectrum when desiredSecondary (SEC) user opportunistically uses remaining spectrumy pp y g p
Goals:maximize spectrum utilization while minimize interference
R d l ( b d l h l )Road analogue (spectrum bands lanes, spectrum users vehicles)
Real RoadSpectrum Road
Open SpectrumPRI SpectrumPRI
SEC
C i i diCognitive RadioC iti di i th k bli t h l f D i Cognitive radio is the key enabling technology for Dynamic Spectrum Access!!C bilit t h th t i t i ti Capability to use or share the spectrum in an opportunistic manner
Band idth har estingBandwidth harvestingDynamic spectrum access techniques allow the CR to operate in the best available channeloperate in the best available channelCR enables the usage of temporally unused spectrum
h l hSpectrum hole or white space
C iti R diCognitive RadioCognitive Radio (CR) Definition
A radio can change transmit parameter based on h hinteraction with the environment
Cognitive CycleSensing, analysis, decision
Particular requirementsSpectrum sensing
Fast and quite sensitive
Frequency agilityCentral frequency, bandwidth
Spectrum Management FrameworkSpectrum Management FrameworkDetermine which portions of the spectrum is available and detect the Select the best available spectrum is available and detect the presence of licensed users when a user operates in a licensed band
channelCoordinate access to this channel with other users
Spectrum DecisionSpectrum
Sh i
Channel CapacityVacate the channel
when a licensed user is detected
SharingSpectrum
holeDecisionRequest
Spectrum Sensing
Spectrum Mobility Primary user
detectionSpectrum
detection (Channel)Characterization
TransmittedSignal
Radio Environment
What is the spectrum usage in China?
Is there any correlation among the spectrum usage?Is there any correlation among the spectrum usage?
Whether we can leverage the spectrum correlation for spectrum usage prediction?
M C d dMeasurement Conducted4 measurement sites
2 in downtown area and 2 in suburban area2 in suburban area
Measurement period From 15:00 Feb 16thFrom 15:00 Feb 16th 2009 to 15:00 Feb 23th 2009
F b dFrequency band Between 20MHz and 3GHz
D. Chen, S. Yin, Q. Zhang, M. Liu, and S. Li, “Mining Spectrum Usage Data: a Large scale“Mining Spectrum Usage Data: a Large-scale Spectrum Measurement Study”, in ACM Mobicom, 2009.
M E iMeasurement EquipmentSpectrum analyzer type: R&S EM550
Scanning frequency range: 20MHz~3.6GHz
Measurement resolution: one per 0.2MHz
14,900 frequency readings per time slot (roughly 75 seconds)14,900 frequency readings per time slot (roughly 75 seconds)
8,058 (7 days/ 75s) time slots
14 900×8 058 data points in the data set (roughly 2GB in size) per 14,900×8,058 data points in the data set (roughly 2GB in size) per location
D t St ti ti A l iData Statistic AnalysisData OverviewData Overview
Energy level overview for location 3 over 7 days
Data Statistic AnalysisConvert power level to 0/1 state
Data Statistic Analysis
Channel state information (CSI)CSI(t, c) = 0 denotes that Channel c is idle at time slot t
( )CSI(t, c) = 1 otherwise
Convert energy level to CSI dataThreshold is set to be 3dB higherThreshold is set to be 3dB higher
than the minimum value seen in
this channel over the entire duration of the trace collected
D t St ti ti A l iData Statistic AnalysisChannel Vacancy Duration (CVD)Channel Vacancy Duration (CVD)
Period in which a channel remains idle
Data Statistic AnalysisData Statistic AnalysisCVD distribution
CVD follows an exponential‐like distribution in all channels we studied, but are not independently distributed
cxb −cxbeay +=
Data Statistic AnalysisData Statistic AnalysisTwo messages delivered by CVD distribution:Two messages delivered by CVD distribution:1. CVD distribution can be nearly exactly estimated
2 Ch l t t t b d l d 1 t d M k Ch i (1 t2. Channel state cannot be modeled as 1st‐order Markov Chain (1st‐MC) model
If the channel state can be modeled as 1st‐MC, it should be:1. b=c. However, b=1.67, c=1.07 )( cxbeay −+=
2.
However
,...],,|Pr[]|Pr[ 2111 −−++ = tttttt xxxxxx
9189530]0|0P [However 918953.0]0|0Pr[ 1 ===+ tt xx55454.0]1,0|0Pr[ 11 ==== −+ ttt xxx
24
Data Statistic AnalysisData Statistic AnalysisTemporal / Spectral / Spatial CorrelationTemporal / Spectral / Spatial CorrelationIntuitively speaking, correlation coefficient (from ‐1 to 1) indicates the likeness of two series (curves)e ess o o se es (cu es)
25
Data Statistic AnalysisData Statistic Analysis• Channel state CS(t, c) is a microscopic level measure of
the channel utilization• Service congestion rate (SCR)
• Examine it across channels within the same service for a given • Examine it across channels within the same service for a given time slot
Data Statistic AnalysisIntra service spectral correlation
Data Statistic AnalysisIntra‐service spectral correlation
CSI: Channel State Infomation. The value is 0 or 1.
High spectral correlation is found between the CSI series of the channels within the same service.
27
Diverse Ideas, Confusing TermsDynamic spectrum accessDynamic spectrum allocationSpectrum property rightsSpectrum commonspOpportunistic spectrum accessSpectrum poolingSpectrum poolingSpectrum underlayS t lSpectrum overlayCognitive radio……
Exclusive Use ModelExclusive Use Model
Maintains the basic structure: license for exclusive use
Introduces flexibility in allocation and spectrum usage
Exclusive Use ModelExclusive Use Model
Maintains the basic structure: license for exclusive use
Introduces flexibility in allocation and spectrum usage
Spectr m Propert RightsSpectrum Property Rights
Allows selling and trading spectrum and freely choosing technology
Let economy & market determine the most profitable use of spectrum
Dynamic Spectrum AllocationDynamic Spectrum Allocation
Dynamic spectrum assignment to different services
Exploiting spatial and temporal traffic statistics
Open Sharing ModelOpen Sharing Model
Open sharing among peer users (spectrum commons)
Draws support from the success of unlicensed ISM bands
Hierarchical Access ModelHierarchical Access Model
Hierarchical access with primary and secondary usersHierarchical access with primary and secondary users
sharing with limited interference to primary users (licensees)
Hierarchical Access ModelHierarchical Access Model
Spectrum underlay: constraint on transmission power
Spectrum overlay: constraint on when and where to transmit
Dynamic Spectrum Access (DSA)Dynamic Spectrum Access (DSA)Concept: secondary (SEC) users with cognitive radio (CR) opportunistically use spectrum of primary (PRI) usersoppo tu st ca y use spect u o p a y ( R ) use s
Public Network TV Broadcast (PRI)
InfrastructureInfrastructure(SEC)
Ad hocVideo
Ad hoc (SEC)
File sharingg
39Public Safety (PRI)
Research Challenges (1)Research Challenges (1)Spectrum rights: PRI’s incentive to share spectrum for SECPublic Network TV Broadcast (PRI)
InfrastructureInfrastructure(SEC)
Ad hocAd hoc (SEC)
40Public Safety (PRI)
Research Challenges (2)Research Challenges (2)Spectrum sensing: discovery of spectrum opportunity
Public Network TV Broadcast (PRI)
InfrastructureInfrastructure(SEC)
Ad hocAd hoc (SEC)
41Public Safety (PRI)
Research Challenges (3)Research Challenges (3)Spectrum sharing: opt. SEC’s performance w/o harmful interference for PRIR
Public Network TV Broadcast (PRI)
InfrastructureInfrastructure(SEC)
Ad hocAd hoc (SEC)
42Public Safety (PRI)
i ddQuestions to Address
Economics (Policy and Marketing) Aspect
T h l P t t dTechnology Aspect
Prototype and Experiments
E isting Primar Spectr m MarketExisting Primary Spectrum MarketFirst come first served Lottery Beauty contestFirst come first served, Lottery, Beauty contestAuction
Cases:$100 billion from European auctions in 2000$19.592 billion from US 700 MHz auction in 2008
Pros:Pros:Maximize social welfareTransparent
Cons:long‐term: years or decades
i l/ i lnational/regionalhuge up‐front investments
Secondary Spectrum MarketSecondary Spectrum Market Incentive of sellers Incentive of sellers (primary users)
Non governmentNon governmentRevenue generation
Fine spectrum usageFine spectrum usageSmall regionShort term
To propose an auction to h lmeet the goals?
J. Jia, Q. Zhang, Q. Zhang, and M. Liu, "Revenue Generation for Truthful Spectrum Auction in Dynamic Spectrum Access", in ACM MobiHoc, 2009.
Secondary Spectrum AuctionSecondary Spectrum Auction Spectrum resourcep
Space: interferenceFrequency: channels CCA
C
1 2 3
Time: slotted
Bidding language
AACC
BBA
C
Bg g g
Total Channels in All CellsSpecific cells and
di h lcorresponding channels, etc.
Desired propertiesRevenue: primary’s incentiveTruthfulness: prevent market manipulation & ease biddersC i l ffi i l i iComputational efficiency: real‐time operation
Wireless Service Providers (WSPs) in ( )Dynamic Spectrum Market
Dynamic spectrum marketShort‐term, real‐time
WSPs play an important role in wireless industryAcquire radio spectrumD l i f d id iDeploy infrastructure and provide service
Tight relation between spectrum acquisition and service pricingpricing
Short term spectrum contracts due to time‐varying spectrum availabilityConsider both spectrum acquisition and service provision
Severe competition among heterogeneous WSPsEnd users with intelligent devices have more selectionsShort‐term or traffic based contracts
Market StructureMarket StructureThree‐layer modelThree layer model
Spectrum Holder (SH)Primary users
Wi l S i P id (WSP)Wireless Service Providers (WSP)Competing Duopoly
End Users (EU) Freely switch
Relations between layersSpectrum trading (#of spectrum to purchase)to purchase)Service subscription (service price)
48J. Jia and Q. Zhang, “Competitions and Dynamics of Duopoly Wireless Service Providers in Dynamic Spectrum Market", in ACM MobiHoc, 2008.
Cooperation among WSPsCooperation among WSPs
An interesting observationSpectrum occupancy measurement (for CDMA, GSM900 and p p y 9GSM1800, which are occupied by China Mobile and China Unicom)
shown that the collaboration coefficient is high
Unbalanced spectrum usage of these two WSPs
If the two WSPs can cooperate by spectrum exchangeIf the two WSPs can cooperate by spectrum exchangeThey may not need to purchase more spectrum while
idi hi h li f iproviding higher quality of services
Both of them will be benefited by the cooperation
Sharing among WSPsSharing among WSPsCooperation among WSP is carried out by their base Cooperation among WSP is carried out by their base stationsThe particular WSP’s BSs are still fully controlled by p y ythemselves and serving their own end users
The spectrum resource can be shared by other WSP’s BSs, in terms f h i iof the time portion
The BSs serving large user demands and short of spectrum resource can reuse the underresource can reuse the under‐utilized spectrum of other WSPs
P. Lin, J. Jia, Q. Zhang, and M. Hamdi, "Cooperation l damong Wireless Service Providers: Opportunity,
Challenge and Solution", IEEE Wireless Communications Magazine, 2010.
i ddQuestions to Address
Economics (Policy and Marketing) Aspect
T h l P t t dTechnology Aspect
Prototype and Experiments
S t M t F kSpectrum Management Framework
Spectrum S
Channel Capacity
DecisionSpectrum Sharing
Spectrum Decision
Spectrum S
HoleDecisionRequest
pSensingSpectrum
Mobility Primary User Detection
Spectrum (Channel)
Characterization
Radio Environment
TransmittedSignal
Radio Environment
Cogniti e MACCognitive MACC i i l l f f f i i l Critical layer for performance of cognitive protocol stack
f dNew situations for cognitive MAC designProtection of primary user
Sense spectrum to detect
Advanced hardware ffeature
Spectrum aggregation
H d iHardware constraints…
Hardware Constraints of CognitiveHardware Constraints of Cognitive Radio
Few off‐the‐shelf product“XG1TM” by adapt4 incXG1TM by adapt4, inc
Sensing constraintFine sensing within a small portion of spectrumFine sensing within a small portion of spectrum
Transmission constraintMax bandwidth aggregation limitMax bandwidth aggregation limitMax fragmentation number limit
Assumptions in this workAssumptions in this workSingle cognitive radioCan not sense and transmit simultaneouslyy
S t S i d A i T d ffSpectrum Sensing and Accessing Tradeoffsecondary sense t: sensing timeprimary occupied
secondary transmit
gT: maximum transmission timeB: unit bandwidth
BT/(T+2t) Chn3 Avail. Chn3 N.A.
2BT/(T+3t) BT/(T+3t)
55
2BT/(T+3t) BT/(T+3t)
Spectrum Sensing and Accessing Tradeoff
More channels explored
Instant data rate increased (in no primary users)Instant data rate increased (in no primary users)Shannon theory
Ti h d i dTime overhead increasedSensing time to check primary
Negotiation between sender and receiver
How to sense and access in MACBalance spectrum opportunity and sensing overhead
Different from the physical layer sensingDifferent from the physical layer sensing
T f hi W kTargets of this Work
SEC’s dilemma: whether to sense the next channel, or just stop and using current available channels?
More channels sensed, possible more available channels, but more overheadOptimal stopping: Classical Secretary ProblemOptimal stopping: Classical Secretary Problem
Formulate the tradeoff of sensing and accessing and solve the optimal sensing problemthe optimal sensing problemDesign a Cognitive MAC to facilitate optimal sensing solutionsolutionUse simulation to verify its performance
Preliminar of Optimal StoppingPreliminary of Optimal StoppingThe theory of optimal stopping:The theory of optimal stopping:
Choose a time to take an action based on sequentially observed random variables to maximize an expected payoff or to minimize an expected cost
Two objects:A sequence of random variables, X1; X2; : : :, whose joint distribution is assumed to be known, and
A sequence of real‐valued reward functions, y0; y1(x1); y2(x1; x2); : : :
For n = 1; 2; : : :, after observing X1 = x1; : : : ; Xn = xn, the decision may be to stop and receive reward yn(x1; : : : ; xn), or continue and observe Xn+1
Goal: choose a time to stop such that the Goal: choose a time to stop such that the
expected reward is maximized
58
O ti l St i f S t S iOptimal Stopping of Spectrum SensingTh t i d i i bl b f l t d The spectrum sensing decision problem can be formulated as an optimal stopping problem
We consider a finite horizon problemMax number of channels a user can probe is limited
Solvable by backward induction principle (has exponential complexity)
Reduce the computational complexity k stage look ahead rules to approximate the optimal stopping rulek‐stage look‐ahead rules to approximate the optimal stopping rule
1‐stage or 2‐stage look‐ahead is almost optimal
HC MACHC‐MACHC MAC H d t i d M lti Ch l C iti HC‐MAC: Hardware‐constrained Multi‐Channel Cognitive MAC
B i tiBasic assumptionsA common control channel is usedHardware constraints of secondary usersHardware constraints of secondary usersPrimary protection time (T)
3 types of control packets on common channel3 types of control packets on common channelC‐RTS/C‐CTSS‐RTS/S‐CTSS RTS/S CTST‐RTS/T‐CTS
HC MACHC‐MAC
ContentionUse C‐RTS/C‐CTS for contention and spectrum contention and spectrum reservation DCF model scheme
SensingUse physical layer sensingExchange spectrum information with S‐RTS/S‐
TransmissionSpectrum aggregationN tif i hb it information with S RTS/S
CTSUse optimal stopping rule
Notify neighbors its completion with T‐RTS/T‐CTS
61
Sim lation Res ltsSimulation ResultsDiff t l ti Different evaluation cases
Fully‐connected, spectrum‐homogeneous topologyhomogeneous topologyFully‐connected, spectrum‐heterogonous topologyheterogonous topologyRandom topology
62
Simulation Results (Cont )Simulation Results (Cont.)Case 1: The throughput with Case 1: The throughput with different total numbers of channel
Case 2: Individual throughput for 2 secondary pairs
• Case 3: A random topology with 15 secondary flowsy
63
Brief Summary for HC MACBrief Summary for HC‐MACEffi i t iti MAC t l th t tili lti l t Efficient cognitive MAC protocol that utilizes multiple spectrum opportunities (channels) to improve spectrum utilization
Consider hardware constraints of cognitive radio: sensing constraints and transmission constraints
Identify the problem for secondary user on how to maximize throughput by optimizing the sensing decision in a sequence of
i sensing processesCan be mapped to as a well‐defined optimal stopping problemB th ti l l ti d i ti l i bt i dBoth optimal solution and approximation rule is obtained
C ti R l i CRNCooperative Relay in CRNResource unbalance is severe in cognitive radio network Resource unbalance is severe in cognitive radio network (CRN)
Spectrum heterogeneityp g y
How to fulfill the heterogeneous traffic demand from secondary users? y
By fully leveraging the unbalanced spectrum usage and availability within the secondary CRN
J. Jia, J. Zhang, and Q. Zhang, "Cooperative g gRelay for Cognitive Radio Networks", in IEEE Infocom 2009.
I S Di iImprove Spectrum DiversityAssumptions Assumptions
Single radio (both AP and nodes)OFDMA100 Kbps per channel
[1] [3]
AP
[1] [3]
AP
rate demand
Traditional:
S1: 100 KbpsS2: 50 Kbps
S2S1
[ ] [ ]
S2S1
[ ] [ ]rate demand S2: 50 Kbps
AP [1,3,4]AP [1,3,4]
APAP APAPRelay:
S1: 150 Kbps
150Kbps 50Kbps150Kbps 50Kbps150Kbps150Kbps
S2S1
[1] [3,4]
S2S1
[1] [3,4]
[2] S2S1
[1]
[2] S2S1
[1]
il bl h l
5 pS2: 50 KbpsS2S1 [1,2] [2,3,4]S2S1 [1,2] [2,3,4]
[ ][ ]available channelsSlot 1 Slot 2
Key Issues to be DiscussedHow to enable three‐node multiple‐channel (probably discontinuous) transmission?
The receiver receives one flow of data from the source directlyRelay node decodes another flow of data on another channel and shift it to a third channel to forward it to the receivershift it to a third channel to forward it to the receiver
Cooperative relay brings new issues of resource allocationAmong secondary users, how to select proper node as relay node Among secondary users, how to select proper node as relay node and also how to allocate proper spectrum for secondary users
New routing metric that can exploit the performance gain bring by cooperative relay
Relay‐Assisted D‐OFDMSenderSender
Simultaneous TX from a single radio
R l d iRelay and receiverDecode packet correctly using part of the whole using part of the whole subcarriers
Alleviate the interference S:
CH1 CH2 CH3
fP2 (T1)
S:
CH1 CH2 CH3
fP2 (T1)Alleviate the interference
from other simultaneous transmitting channels to h h h h
R: f
f
P1(T1) P3(T2)P2 (T1)
P2(T2)
R: f
f
P1(T1) P3(T2)P2 (T1)
P2(T2)
achieve a higher SNR on the specific channel
D: f
fu1 fv1 fu2 fv2 fu3 fv3
D: f
fu1 fv1 fu2 fv2 fu3 fv3
A hit t f I f t t b dArchitecture for Infrastructure‐based CRNCRN
Upper Layer Upper Layer
AP End User
Upper Layer Upper Layer
AP End User
MACCoordination
Spectrum Allocation
Reconfigurable Radio
End User MAC
Reconfigurable Radio
MACCoordination
Spectrum Allocation
Reconfigurable Radio
End User MAC
Reconfigurable Radio
Programmable Hardware (USRP)
Relay-assisted DOFDMReconfigurable Radio Reconfigurable Radio
Programmable Hardware(USRP)
Relay-assisted DOFDM
Programmable Hardware (USRP)
Relay-assisted DOFDMReconfigurable Radio Reconfigurable Radio
Programmable Hardware(USRP)
Relay-assisted DOFDM
(USRP)
Spectrum Channel
(USRP)(USRP)
Spectrum Channel
(USRP)
Joint Relay Selection and Channel AllocationJointly design Jointly design
relay selection spectrum allocation
Two types of transmissionsypTraditional transmission between AP and node viAdvanced transmission among AP, node vi and relay vjj
Constraints for relay relationshipone node can have only one relay; one relay can serve only one node; when one node is served by a relay, this node can not be the relay node of another noderelay node of another node.
⑴
rij = 1 means that vj performs as relay node for vi
J i t R l S l ti d Ch l All tiJoint Relay Selection and Channel Allocation (Cont.)( )
Constraint related to interference avoidanceA channel can be used by only one active transmission A channel can be used by only one active transmission link in one frameActive link can be either from AP or from relay
⑵
xik = 1 if channel k is assigned to node vi to transmit or
⑵
ik g ireceive data
Channel availability constraint yChannel allocation should also satisfy the channel availability constraint in each user
⑶
Joint Relay Selection and Channel AllocationJoint Relay Selection and Channel Allocation (Cont.)
Calculate throughput of node vi according to whether it is a relay or is helped by a relay
Normal receiver without any relay’s help
Receives data from the AP with the help of a relay vj
Acts as a relay for node vj
Single unified format for throughput calculation
J i t R l S l ti d Ch l All tiJoint Relay Selection and Channel Allocation (Cont.)( )
System optimizationNonlinear integer programming problemg p g g p
Propose a heuristic centralized solutionFirst partition nodes into two parts according to their traffic and spectrum availability Greedily select the best pair of destination and relay from Greedily select the best pair of destination and relay from the two parts to increase the system throughput
T b d I l iTestbed ImplementationGNU Radio + USRPGNU Radio + USRP
GNU RadioSignal processing libraryCommunicating with the USRPCommunicating with the USRP
Universal Software Radio Peripheral (USRP)Front‐end functionality A/D and D/A conversion
T tb d F kTestbed FrameworkUSRP + GNU Radio
Underlying MAC L
Upper Layer
MAC LayerOwn designed protocol
Network Layer Network Layer ent
men
t
Network LayerClick
Cross Layer Management
Ne wo aye
MAC Layer
er M
anag
eme
rum
Man
agem
Cross Layer ManagementControl path
Radio Spectrum Management Cro
ss-L
aye
Rad
io S
pect
r
GNURadioRadio Spectrum Management
Spectrum policy and market rules USRP
E i C fi iExperiment ConfigurationW i l i USRP d GNU We implement a system using USRP and GNU Radio
A d d l kAP mode, downlinkDAC rate (128 samples/ms), interpolation rate (256, 2 samples/symbol 1bit/symbol)samples/symbol, 1bit/symbol)FFT length (512 subcarriers), channel bandwidth (100 KHz = 100 contiguous subcarriers supporting about 60 KHz = 100 contiguous subcarriers, supporting about 60 Kbps data rate)Packet size (200 bytes)( y )Duration of each run (1000 packets)
Experiment ResultsRelay selection for multiple transmissionsRelay selection for multiple transmissions
AP 1on
2 11 20.40.60.8
1
ativ
e Fr
actio
Total Gain over TA
D2R1D1 1 R2 1
21.3 1.32 1.34 1.36 1.38 1.40
0.2
Throughput GainC
umul
a
2
AP0 8
1
actio
n
Node D1 Gain over TA
1 11 1 1
0.20.40.60.8
mul
ativ
e Fr
a
Node D2 Gain over TA
D2R1D1 R21 1 1.3 1.35 1.4 1.45 1.50
Throughput Gain
Cum
P bl St t tProblem StatementPropose a cooperative cognitive radio framework
Primary users, aware of the existence of secondary users, may select f th t b th ti lsome of them to be the cooperative relay
In return lease portion of the channel access time to them for their own data transmissiono data t a s ss o
Both primary and secondary users target at maximizing their utilities in terms of their transmission rate and their utilities in terms of their transmission rate and revenue/payment
J. Zhang and Q. Zhang, "Stackelberg Game for Utility-Based Cooperative Cognitive Radio Networks", in ACM MobiHoc, 2009.
Win‐Win CaseBenefit: win‐win
PU: Higher transmission rate , lower delaySU: More time left for secondary accessSU: Increased access opportunity and satisfied trafficP d PU: Increased revenue
Primary User Secondary User
Transmission Rate Access Opportunity
Primary User Secondary User
Transmitted TrafficRevenue
Ch llChallengesI i h iIncentive mechanism
Two independent networks trying to maximize their own utilityFor SUFor SU
What’s the direct benefit of serving as a relay? Access time What’s the overall utility it wants to maximize? Traffic ‐ Revenue
F PUFor PUWhat’s the direct benefit of selecting a relay? RevenueWhat’s the overall utility it wants to maximize? Traffic + Revenue
Optimal decisionHow to select relay nodes?How much money should SU pay for PU?How much money should PU charge the successful relays?Stackelberg game Nash EquilibriumStackelberg game Nash Equilibrium
CCRN M d lCCRN ModelOne primary transmission pair PT PR with required rate ROne primary transmission pair PT PR, with required rate R0
Several secondary transmission pairs STi SRi, totalSi∈
In a unit slotIn a unit slot,PT STi In first fractionPT PR, STi PR in next fraction
αβ)1( βα −
STi SRi in the last fraction in TDMA mode in proportional to the payment
α−1
ST2ST4
SR2
Fixed transmission powerPrimary user selects i i i b fi
),,( Sβα
ST4
SRPR
to maximize its benefitSecondary users select the payment to maximize its utility
PTSR1
ST1 SR4
payment to maximize its utilityST3
SR3
S k lb G A l iStackelberg Game AnalysislGame player:
Leader: primary userll dFollower: secondary user
Utility function:Primary user: weighted sum of rate and revenue
Secondary user: transmission rate penalized by the paymentSecondary user: transmission rate penalized by the payment
Stackelberg Game AnalysisUse backward induction to analyze the Stackelberg game
Secondary users form non‐cooperative payment selection game (NPG)
select payment under given parameters Primary user maximize its utility function by selecting Primary user maximize its utility function by selecting , assuming that secondary users act according to the NE point in NPG
Secondary User Payment Selection Game
The NPG is proved to have a unique NE point.pThe NE point is given by
Under the constraints:U de t e co st a ts:
M i i i P i U ’ UtilitAssume the secondary users select payment according to
Maximizing Primary User’s Utilityy p y g
the NE point, primary user’s utility is
Primary user maximize its utility at
WhereWhere
Under the condition:
Simulation ResultsSecondary nodes are placed at normalized distance d from y pPT and 1‐d from PRFading channelS consists k nodesValue of utility vs. d and k: y
ConclusionsA novel cooperative cognitive radio framework has been A novel cooperative cognitive radio framework has been proposed, which enable the primary user to actively involve secondary users as the cooperative relaysy p y
A payment mechanism, where secondary users pay charges to primary user in order to achieve the access opportunity, is to primary user in order to achieve the access opportunity, is designed
By formulating the model as a Stackelberg game, we prove By formulating the model as a Stackelberg game, we prove that there exists a unique Nash Equilibrium for the game and derive the analytical result
Numerical result shows that both primary and secondary systems achieve better performancey
i ddQuestions to Address
Economics (Policy and Marketing) Aspect
T h l P t t dTechnology Aspect
Prototype and Experiments
Importance of TestbedImportance of TestbedSimulation based evaluation (e.g., NS2, Opnet)Simulation based evaluation (e.g., NS2, Opnet)
With simplification of realityEasy for implementationy pNot convincing enough for real performance
Cognitive radio has sophisticated PHY and MACMore concern about simulation (no widely used simulator)
With real testbedM f l i lMore useful testing resultHarder work, and long term implementation
90
GNU R diGNU RadioGNU R diGNU Radio
Software “blocks” that run on the GPPUniversal Software Radio Peripheral (USRP)
Channelizer, DUCs/DDCs, Interpolator/DecimatorRF Daughter boards
RF Interface
Enter the USRPEnter the USRPUniversal Software Radio Peripheral (USRP)p ( )
ADC 64 Msps (4)DAC 128 Msps (4)DAC 128 Msps (4)USB 2.0 Interface*Small FPGA†Small FPGA
* Bottleneck: 8 MHz, 16-bit complex samples† Loads bitfile from the GPP
USRP2USRP2
Uses same RF Daughter boards as USRPGigabit Ethernetg100 Msps ADC (2)400 Msps DAC (2)400 Msps DAC (2)Larger FPGASD card readerSD card reader
RF Da ghter boardsRF Daughter boardsBasic RXReceiver for use with external RF hardware
Basic TXTransmitter for use with external RF hardware
RFX24002.3‐2.9 GHz Transceiver, 50+mW output
RFX900800‐1000MHz Transceiver, 200+mW output
RFX400 MH MH T i W t t 400 MHz ‐ 500 MHz Transceiver, 100+mW output
XCVR24502 4 2 5 GHz and 4 9 to 5 85 GHz Dual band Transceiver 2.4‐2.5 GHz and 4.9 to 5.85 GHz Dual‐band Transceiver, 100+mW output on 2.4 GHz, 50+mW output on 5GHz
Sora ApproachEnabling high system throughput
Design new PCIe‐based Interface cardDevelop new software architecture for efficient PHY/MAC processing
New optimizations to implement PHY algorithmsStreamline over multiple CPU cores
Support real‐timeCore dedication
We need to Solve Real ProblemWe need to Solve Real Problem
There is a big gap between
Having a flexible cognitive radio, effectively a building block, g g , y g ,and
The large‐scale deployment of cognitive radio networks g p y gthat dynamically optimize spectrum use
Building and deploying a network of cognitive radios is a complex while essential taskis a complex while essential task
ConclusionConclusionC iti di d d i t i th “N t Cognitive radio and dynamic spectrum access is the “Next Big Thing” in future wireless communications Both economics and technology aspects of cognitive radio Both economics and technology aspects of cognitive radio networking for dynamic spectrum management need to be well studiedReal system deployment is essential to have better understanding about the impact of CR g pto the future wireless networks