1

Cognitive Radio Research in the U.S.: Overview, Challengesand Directions

ARIB Frequency Resource Development Symposium, June 8, 2007

Narayan B. MandayamProfessor of ECE & Associate Director, WINLAB

Rutgers University

narayan@winlab.rutgers.eduwww.winlab.rutgers.edu

2

Cognitive Radio Research A Multidimensional Activity

Spectrum PolicyEconomicsRegulationLegalBusiness

Theory and AlgorithmsFundamental LimitsInformation & Coding TheoryCooperative CommunicationsGame Theory & Microeconomics

Hardware/Software Platforms & Prototyping

Programmable agile radiosGNU platforms

Several Universities E.g. WINLAB-Rutgers, UC Berkeley, Virginia Tech, etc.

Govt Agencies E.g. FCC, DARPA, NSF, etc.

Industry – small and bigE.g. Blossom, Shared Spectrum, Vanu, Intel, Alcatel-Lucent, Philips, Qualcomm and a very many more

3

How it all started -The Spectrum Debate & Cognitive Radio

4

The Spectrum Debate Triumph of Technology vs. Triumph of Economics

Open Access (Commons)[Noam, Benkler, Shepard, Reed …]

Triumph of TechnologyAgile wideband radios will dynamically share a commonsSuccess of 802.11 vs 3G

Spectrum Property Rights[Coase, Hazlett, Faulhaber+Farber]

Triumph of EconomicsOwners can buy/sell/trade spectrumFlexible use, flexible technology, flexible divisibility, transferabilityA spectrum market will (by the force of economics) yield an efficient solution

What everyone agreed on (a few years ago):Spectrum use is inefficientFCC licensing has yielded false scarcity

5

Source: FCC website

Spectrum Management: In fairness to the FCC, Frequency allocation is complex…

6

Maximum Amplitudes

Frequency (MHz)

Am

plid

ue (d

Bm

)

Heavy Use

Sparse Use

Heavy Use

Medium Use

Less than 6% OccupancyLess than 6% Occupancy

FCC measurement shows that occupancy of approximately 700 MHz ofspectrum below 1 GHz is less than 6~10%~ 13% spectrum opportunities utilized in New York City during 2004 Political Convention to nominate U.S. Presidential Candidate

Spectrum Management: Then again, poor utilization in most bands

Atlanta

NewOrleans

SanDiego

Frequency

Time

7

Arguments against Triumph of TechnologyPartially developed theory

Information theoretic relay channelInformation theoretic interference channelAd hoc network capacity, with/without mobility

Technology PanaceaSpread spectrum, UWB, MIMO, OFDMShort range communicationsAd hoc multi-hop mesh networks

Infant technologyUWB, MIMO antenna arraysTransmitter agility

Technology not separable from user assumptionsCapabilities of technology vary with cooperation

8

Recent Research DevelopmentsResearch themes that have emerged from mobile ad hoc and/or

sensor networks research:Hierarchical Network Architecture wins

Capacity scaling, energy efficiency, increases lifetimes, facilitates discovery

Cooperation winsAchievable rates via information theoretic relay and broadcast channels

“Global” awareness and coordination winsSpace, time and frequency awareness and coordination beyond local measurements

Efficient operation requires radios that can:CooperateCollaborateDiscoverSelf-Organize into hierarchical networks

9

The Spectrum Debate and Cognitive Radio

What everyone agrees on now:Spectrum use is inefficientFCC licensing has yielded false scarcity

Possible middle ground?Dynamic spectrum accessShort-term property rightsSpectrum use driven by both technology and market forces

Cognitive Radios with ability to incorporate market forces?Microeconomics based approaches to spectrum sharingPricing and negotiation based strategies

(e.g. Ileri & Mandayam, “Dynamic Spectrum Access Models- Towards an Engineering Perspective in the Spectrum Debate” in IEEE Communications Magazine January 2008)

10

Dynamic Spectrum Management & Cognitive Radio

11

Static allocation of spectrum is inefficientSlow, expensive process that cannot keep up with technology

Spectrum allocation rules that encourage innovation & efficiency

Free markets for spectrum, more unlicensed bands, new services, etc.Anecdotal evidence of WLAN spectrum congestion

Unlicensed systems need to scale and manage user “QoS”Density of wireless devices will continue to increase

~10x with home gadgets, ~100x with sensors/pervasive computingInteroperability between proliferating radio standards

Programmable radios that can form cooperating networks across multiple PHY’s

Motivation for Dynamic Spectrum and Cognitive Radio Techniques:

12

Spectrum Management: Problem ScopeSpectrumAllocation

Rules(static)

INTERNET

BTS

AuctionServer

(dynamic)

SpectrumCoordination

Server(dynamic)

AP

Ad-hocsensor cluster(low-power, high density)

Short-rangeinfrastructure

mode network (e.g. WLAN)

Short-range ad-hoc net

Wide-area infrastructuremode network (e.g. 802.16)

Dense deployment of wireless devices, both wide-area and short-rangeProliferation of multiple radio technologies, e.g. 802.11a,b,g, UWB, 802.16, 4G, etc.How should spectrum allocation rules evolve to achieve high efficiency?Available options include:

Agile radios (interference avoidance)Dynamic centralized allocation methodsDistributed spectrum coordination (etiquette)Collaborative ad-hoc networks

Etiquettepolicy

SpectrumCoordination

protocols

Spectrum Coordinationprotocols

Dynamic frequencyprovisioning

13

Broad range of technology & related policy options for spectrum

Need to determine performance (e.g. bps/Hz or bps/sq-m/Hz) of different technologies taking into account economic factors such as static efficiency, dynamic efficiency & innovation premium

Hardware Complexity

Protocol Complexity(degree of

coordination)

ReactiveRate/Power

Control

ReactiveRate/Power

Control

AgileWideband

Radios

AgileWideband

Radios

Unlicensed Band

with DCA (e.g. 802.11x)

Unlicensed Band

with DCA (e.g. 802.11x)

InternetServer-based

SpectrumEtiquette

InternetServer-based

SpectrumEtiquette

Ad-hoc,Multi-hop

Collaboration

Ad-hoc,Multi-hop

Collaboration

Radio-levelSpectrumEtiquetteProtocol

Radio-levelSpectrumEtiquetteProtocol

StaticAssignmentStatic

Assignment

InternetSpectrumLeasing

InternetSpectrumLeasing

“cognitive radio”schemes

UWB,Spread

Spectrum

UWB,Spread

Spectrum

“Open Access”+ smart radios

Unlicensed band +simple coord protocols

Cognitive Radio: Design Space

14

Selected Cognitive Radio Research

15

Cognitive Radio ResearchFundamental research and algorithms – based on foundations of:

Information and Coding TheoryRelay cooperation, User Cooperation, Coding techniques for cooperation, Collaborative MIMO techniques

Signal ProcessingCollaborative signal processing, Signal design for spectrum sharing, Interference avoidance, Distributed sensing algorithms

Game TheoryMicroeconomics and pricing based schemes for spectrum sharing, negotiation and coexistence, Incentive mechanisms for cooperation

MAC and Networking AlgorithmsDiscovery protocols, Etiquette protocols, Self-organization protocols, Multihop routing

16

Information Theoretic ApproachesVarious types of relay cooperation and user cooperation models

Cooperation – nodes share power and bandwidth to mutually enhance their transmissionsCan achieve spatial diversity – similar to multiple antennas

Fundamental limits are known in limited casesPrimary focus on achievable rates, outage and various cooperative coding schemes, e.g.

Decode-and-ForwardCompress-and-ForwardAmplify-and-Forward

Relay Cooperation

17

SN

Wired Backbone

SN

SN

SN

SN SN

SN

SN

SN

SN

SN

SN

SN SN

SN

SN

SN

SN

SNSN

SN

SN

SN

SN

SN

AP

AP

AP

SN

SN

SN

SN

SN

Information Theoretic ApproachesVarious types of relay cooperation and user cooperation models

Cooperation – nodes share power and bandwidth to mutually enhance their transmissionsCan achieve spatial diversity – similar to multiple antennas

Fundamental limits are known in limited casesPrimary focus on achievable rates, outage and various cooperative coding schemes, e.g.

Decode-and-ForwardCompress-and-ForwardAmplify-and-Forward

User Cooperation

18

E.g. Relay Cooperation vs User Cooperation (Sankar-Kramer-Mandayam)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

x-coordinate

y-co

ordi

nate

User 1User 2DestinationRelay

Sector of circle of radius 1 and destination at circle center. Uniform node distribution.Average rate and outage over 100 locations.Path-loss exp. γ = 4. Transmit SNR for all users = P1.Processing cost factor η

Compare Outage Probability?

How much better than TDMA?

What is the effect of processing costs?

19

Amplify-and-Forward Cooperation η = 0.01

Relay and User Cooperation schemes gain over TDMARelay achieves gains relative to user-cooperation

-10 -5 0 5 10

10-4

10-3

10-2

10-1

100

10-5

Transmit SNR P1 (dB)

Out

age

Pro

babi

lity

Pou

t

Sub-plot 1

-10 -5 0 5 10 15

10-4

10-3

10-2

10-1

100

10-5

Total (transmit+proc.) SNR Ptot (dB)

Out

age

Pro

babi

lity

Pou

t

Sub-plot 2

Coop. 2-hopRelay Pr=0.5P1Relay Pr=P1

TD-MAC

Coop. 2-hopRelay Pr=0.5P1Relay Pr=P1

TD-MAC

K = 2Rate R = .25

K = 2Rate R = .25η = .01

20

Amplify-and-Forward Cooperation η = 0.5, 1

Even with η ↑ relay achieves gains relative user cooperation

-5 0 5 10 15

10-4

10-3

10-2

10-1

100

10-5

Total (transmit+proc.) SNR Ptot (dB)

Out

age

Pro

babi

lity

Pou

t

Sub-plot 1

-5 0 5 10 15

10-4

10-3

10-2

10-1

100

10-5

Total (transmit+proc.) SNR Ptot (dB)

Out

age

Pro

babi

lity

Pou

t

Sub-plot 2

Coop. 2-hopRelay Pr=0.5P1Relay Pr=P1

TD-MAC

Coop. 2-hopRelay Pr=0.5P1Relay Pr=P1

TD-MAC

K = 2Rate R = .25η = 1

K = 2Rate R = .25η = .5

21

Game Theory ApproachesNegotiation strategies for mediation

Pricing and microeconomic strategies to promote cooperation in spectrum sharingReimbursing costs in cooperation – energy costs, delay costs

Domination strategies for situations of conflictSpectrum warfare with agile waveforms and competition for spectrum

Coalition formation strategies for cooperationCoalitional games for receiver and transmitter coalitions in spectrum sharing

Approaches result in algorithms that specify:• Power control • Rate control• Channel selection• Cooperation techniques• Route selection

22

Each node•Enjoys its own data reaching the access point.•Pays for its outgoing throughput to adjacent devices.•Gets reimbursed for data it relays only if the data reaches theaccess point.

E.g. Inducing Forwarding through Reimbursement (Ileri-Mandayam)

23

Horizontal Trajectory

Radio tower

5 m 10 m

-5 m

-

USER 2(POTENTIAL

FORWARDER)

USER 1(NON-FORWARDER)

ACCESS POINT

5 m POSITIONS OFUSER 1 IN

HORIZONTALGEOMETRY

• Incentives for forwarding works well when nodes tend to “cluster”

• The aggregate bits/Joule in network is higher

• Network revenue is also higher

0 1 2 3 4 5 6 710

14

1015

1016

1017

1018

1019

Aggregate bits/JouleAggregate bits/Joule, no reimbursement.

E.g. Inducing Forwarding through Reimbursement

user 2 distance

24

Reactive (autonomous) methods used to avoid interference via:Frequency agility: dynamic channel allocation by scanningPower control: power control by interference detection and scanningTime scheduling: MAC packet re-scheduling based on observed activityWaveform agility: dynamism in signal space

A

D

C&D’s spectrum bandRange with Power Control

A

BB

C

DA&B’s spectrum band

Range without Power Control

Range without Power Control

Range with Power Control

PHY & MAC: Reactive Algorithms

Frequencyagility

Scheduling

D DC D

25

Cognitive Radio: Limitations of Reactive Schemes

Reactive schemes (without explicit coordination protocols) suffer from certain limitations:

Near-far problems possible at the receiverInability to predict future behavior of other nodesOnly detects transmitters, not receivers, but interference is a receiver property

Coverage area of D

Y

A

B

C

D

A cannot hear DA’s agile radio waveform

D’s agile radio waveformwithout coordination protocol

with coordination

Hidden Terminal Problem

Coverage area of A

Coverage area of D

Y

A

B

C

D

A cannot hear DA’s agile radio waveform

D’s agile radio waveformwithout coordination protocol

with coordination

Hidden Terminal Problem

26

Cognitive Approaches: OutlookFundamental research and algorithms – based on foundations

ofInformation and coding theory, Game theory, MAC, Networking, Signal processing

all point to the following:Cognitive radio networks require a large of amount of network (and channel) state information to enable efficient

DiscoverySelf-organization Cooperation Techniques

In addition to advances in cognitive radio technology, Network Architectures and Information Aids that

support these are required

27

Cognitive Radios need help too!

Infrastructure that can facilitate cognitive radio networks

Coordination mechanisms for coexistence and cooperation

Information aids Spectrum Coordination Mechanisms

Network architectures “Spectrum Servers” to advise/mediate sharing

28

Cognitive Radio: Common Spectrum Coordination Channel (CSCC) (Jing-Raychaudhuri)

Common spectrum coordination channel (CSCC) Common spectrum coordination channel (CSCC) can be used to coordinate radios with different PHY

Requires a standardized out-of-band etiquette channel & protocolPeriodic tx of radio parameters on CSCC, higher power to reach hidden

nodesLocal contentions resolved via etiquette policies (independent of

protocol)Also supports ad-hoc multi-hop routing associations

Frequency

CH#N

CH#N-1

CH#N-2

CH#2

CH#1

CSCC

::

Ad-hocnet B Ad-hoc

net A

Ad-hocPiconet

MasterNode

CSCCRX range

for X

CSCCRX range

for Y

Y

X

29

CSCC Spectrum Etiquette ProtocolCSCC( Common Spectrum Coordination Channel) can enable mutual observation between neighboring radio devices by periodically broadcasting spectrum usage information

Edge-of-bandcoordination channel

Service channels

30

CSCC: Proof-of-Concept Experiments

WLAN-BT Scenario

Different devices with dual mode radios running CSCCd=4 meters are kept constantPriority-based etiquette policy

31

CSCC Results: Throughput Traces

0 20 40 60 80 100 120 140 160 180 200 220 240

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

WLA

N T

hrou

ghpu

t (M

bps)

Time (Seconds)

CSCC on CSCC off

WLAN session with BT2 in initial position

WLAN = high priority

0 50 100 150 200 250 30030

35

40

45

50

55

60

65

Blue

toot

h Th

roug

hput

(Kbp

s)

Time (Seconds)

CSCC on CSCC off

BT session with BT2 in initial position

Bluetooth = high priority

Observations:WLAN session throughput can improve ~35% by CSCC coordinationBT session throughput can improve ~25% by CSCC coordination

32

802.11 & 16 Co-Existence Scenario

802.16a 802.11bTraffic Type UBR (Poisson arrival), UDP packet, 512 Bytes datagram

MAC protocol TDMA IEEE 802.11 BSS modeChannel Model AWGN, two ray ground propagation model, no fading

Default channel Channel 1 : centered at 2412GHz Channel 1 : centered at 2412GHzAvailable channels 4 (non-overlap) 12 (overlapping)

Bandwidth/channels 20 MHz / 4 non-overlapping chs 22MHz / 11 overlapping chs

Receiver Sensitivity -80dBm (@BER 10^-6) -82dBm (@BER 10^-5)Antenna Height BS 15m, SS 1.5m 1.5m

Tx Power/Max range 33dBm / 3.2Km 20dBm / 500m

Bit Rate 13Mbps 2MbpsRadio parameters OFDM (256-FFT, QPSK) DSSS (QPSK)Background Noise -174 dBm/Hz

Rx Noise Figure 9 dB 9 dB

33

CSCC frequency adaptation when DSS-AP =200m and traffic load 2Mbps

AP1km

100m

802.11b Hotspot

802.16a Cell

BSSS

DSS-AP

802.16 DL 802.11 link0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Ave

rage

Lin

k Th

roug

hput

(Mbp

s)

802.16 DL and 802.11 link

No Coordination CSCC frequency adaptation

0 200 400 600 800 10000.0

0.2

0.4

0.6

0.8

1.0

1.2

Ave

rage

Lin

k Th

roug

hput

(Mbp

s)

Distance between 802.16 SS and 802.11 hotspot (meters)

802.16a DL 802.16a DL with CSCC 802.11 link 802.11 link with CSCC Average No Coordination Average with CSCC

Throughputs vs. DSS-AP by using CSCC power adaptation and traffic load 2Mbps

802.11 & 16 Co-Existence: Reactive vs. CSCC-based Power Control (Jing-Raychaudhuri)

Single 802.11 Hot Spot Case

34

802.11 & 16 Co-Existence: Reactive vs. CSCC Power Control

0.2 0.4 0.6 0.8 1.070000

80000

90000

100000

110000

120000

130000

Aver

age

Net

wor

k Th

roug

hput

(bps

)

Clustering Index

No Coordination with CSCC with RTPC with TA

0.2 0.4 0.6 0.8 1.070000

80000

90000

100000

110000

120000

Ave

rage

Net

wor

k Th

roug

hput

(bps

)

Clustering Index

No Coordination with CSCC with RTPC with TA

802.16 SS nodes are (i) uniform or (ii) clustered in the cell.

clustering index CI = R11/R16

1km

Varying Max Cluster Radius

R16

Max Hotspot Radius R11

Region (i)

Region (ii)

802.16a BS802.16a SS802.11b AP802.11b Client

0.2 0.4 0.6 0.8 1.070000

80000

90000

100000

110000

120000

130000

Aver

age

Net

wor

k Th

roug

hput

(bps

)

Clustering Index

No Coordination with CSCC with RTPC with TA

CSCC power adaptation with clustered 802.16a SS in region (ii), with numbers of 802.16a SS :

802.11b nodes = 1:1

600 Kbps load

1 Mbps load

Multiple 802.11 Hot Spot Case

35

Cognitive Radio: Spectrum Policy ServerInternetInternet--based Spectrum Policy Serverbased Spectrum Policy Server can help to coordinate wireless networks (a “Google for spectrum”)

Needs connection to Internet even under congested conditions (...low bit-rate OK)Some level of position determination needed (..coarse location OK?)Spectrum coordination achieved via etiquette protocol centralized at server

InternetInternet

Access Point(AP2)

AP1

WLANoperator A WLAN

operator B

Ad-hocBluetoothPiconet

Wide-areaCellular data

service

SpectrumPolicy Server

www.spectrum.netAP1: type, loc, freq, pwrAP2: type, loc, freq, pwrBT MN: type, loc, freq, pwr

MasterNode

EtiquetteProtocol

36

What can a Spectrum Policy Server do?

Spectrum Server

rate4rate1

rate2 rate3

Spectrum Policy Server facilitates co-existence of heterogeneous set of radios by advising them on several possible issues

Spectrum etiquetteInterference informationLocation specific servicesMany more things ….

37

Users share a common frequency bandOrthogonal signal dimensions = time slotsTime domain scheduling is used for channelization

Wireless network of L directed linksLinks employ ON-OFF transmission schedule in each time slot

Use constant transmission power in the ON stateLinks employ interference-adaptive modulation/coding

Link rate in each time slot depends on interference, receiver mitigation.Interference depends on the transmission mode

mode = subset of links that are ON simultaneously

Scheduling Variable Rate Links with a Spectrum Server (Raman-Yates-Mandayam)

1 4

3

2

1

3

tli = 1, if link l is in mode i= 0, otherwise.

[0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1] [0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1][0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1][0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1]

Transmission mode [1 0 1 0]

(one of 24 possible modes)

38

network with 4 links

1 4

3

2

Transmission mode [1 0 1 0]

1

3

Rate matrix C =[6.6 0 0.01 0 0.56 0 0.01 0 2.05[ 0 6.6 0.06 0 0 1.86 0.06 0 0[ 0 0 0 6.6 1.0 1.86 0.83 0 0[ 0 0 0 0 0 0 0 6.65 0.32

0 0.01 0 0.49 0 0.01]0.97 0.06 0 0 0.77 0.06 ]0 0 0.04 0.04 0.04 0.04 ]0.05 0.04 0.40 0.19 0.05 0.04]

Glk = link gain fromTx k to Rx l

Spectrum Server Mode Matrix Rate Matrix

39

Spectrum server specifies xi = fraction of time mode i is ON

Average rate in link l is rl = Σi cli xi

In vector form, r = Cx

Spectrum server specifies x to optimize an objective such as for example:

Maximum sum rate of the scheduleMaximize the common rate on the linksFair schedulingEfficient scheduling

Spectrum Server: Optimizing the Mode Schedule

40

As common rmin increases, the sum rate decreasesRates of dominant mode links decreaseRates of disadvantaged links increase

Spectrum Server: Scheduling Results

41

Spectrum Server: Fairness in Scheduling

42

Hardware & Software Platforms

43

Vanu Software Defined Radio

Vanu Inc. SDR programmable radio based on commodity processors. Supports multiple standards on handheld device.

Vanu Inc. Software Defined Radio Source: http://www.vanu.com/products.html

44

GNU Hardware/Software Platform (Blossom)

GNU/USRP boards with GNU Software with API’s for flexible PHY and MAC are currently available for experimentationVarious RF front-ends (0-100MHz, 400MHz, 900MHz, 2.4GHZ) with data rates upto 64 MSamples/secAll DSP functions in software on general-purpose CPU

45

WARP Platform (Rice University)

•Various RF front-ends (2.4GHZ, 5GHz) with baseband of up to 20 MHz•High data rates possible, Programmable PHY & Protocols, Extensible to MIMO•Backed by Xilinx, Nokia

46

Cognitive Radio Networks & ProtocolsDiscovery strategies

Algorithms and protocols for frequency selection, coordination and cooperation

Multihop strategiesAlgorithms for self-organization and routing

AA

BB

D

C

D

E

F

Bootstrapped PHY &control link

End-to-end routed pathFrom A to F

PHY A

PHY BPHY C

Control(e.g. CSCC)

Multi-mode radio PHYAd-Hoc Discovery

& Routing Capability

E.g.

• Cognitive Radio scans for active nodes and executes discovery algorithm

• Control protocol for exchange of network information and routing tables

Functionality can be quitechallenging!

47

WINLAB Cognitive RadioWINLAB’s “network centric” concept for

cognitive radio prototype (..under development in collaboration with GA Tech

& Lucent Bell Labs)

Requirements include:~Ghz spectrum scanning,Etiquette policy processingPHY layer adaptation (per pkt)Ad-hoc network discoveryMulti-hop routing ~100 Mbps+

48

Cognitive Radio Networks: “CogNet”Architecture

New NSF FIND project called “CogNet” aimed at development of prototype cognitive radio stack within GNU frameworkJoint effort between Rutgers, U Kansas, CMU and Blossom Inc

49

Cognitive Radio Networks: “CogNet”Protocol Stack

Global Control Plane (GCP)Common framework for spectrum allocation, PHY/MAC bootstrap, topology discovery and cross-layer routing

Data planeDynamically linked PHY, MAC, Network modules and parameters as specified by control plane protocol

Control PHYControl MAC

Boot-strap

Discovery

PHY

MACNetworkTransport

Application

Control Plane Data Plane

Global Control Plane

Data Plane

Control Signalling

Data Path

Establishment

Naming&

Addressing

50

Cognitive Radio Network Experiments Hardware/Software Platforms@WINLAB

Urban

300 meters

500 meters

Suburban

20 meters

ORBIT Radio Grid

Office

30 meters

Radio Mapping Concept for ORBIT Emulator

400-node Radio Grid Facility at WINLAB Tech Center

ProgrammableORBIT radio node

URSP2CR board

Planned upgrade(2007-08)

Current ORBIT sandbox with GNU radio

ORBIT radio grid testbed currently supports ~10/USRP GNU radios, 100 low-cost spectrum sensors, WARP platforms, WINLAB Cognitive platforms and GNU/USRP2Each platform will include baseline CogNet stack

51

Concluding Remarks

52

Wireless/Mobile/Sensor Scenarios and the Future Internet – NSF’s GENI Project

Some architectural and protocol implications for the future Internet...

Integrated support for dynamic end-user mobilityWireless/mobile devices as routers (mesh networks, etc.)Network topology changes more rapidly than in today’s wired InternetSignificant increase in network scale (10B sensors in 2020!)New ad hoc network service concepts: sensors, P2P, P2M, M2M,…Addressing architecture issues – name vs. routable addressIntegrating geographic location into routing/addressingIntegrating cross-layer and cognitive radio protocol stacksData/content driven networking for sensors and mobile data Pervasive network functionality vs. broadband streamingPower efficiency considerations and computing constraints for sensorsMany new security considerations for wireless/mobileEconomic incentives, e.g. for forwarding and network formation

53

NSF GENI Implementation: Wireless Subnets – Overall Wireless Deployment PlanFive types of experimental wireless networks planned –necessary to support full range of protocol research and to enable new applications

1. Wireless emulation and simulation (repeatable protocol validations)2. Urban 802.11-based mesh/ad-hoc network (real-world networking experience with emerging short-range radios)3. Wide-area suburban network with both 3G/WiMax (wide area) and 802.11 radios4. Sensor networks (…application specific, specific system TBD via proposal process; may include environmental, vehicular, smart spaces, etc.)5. Cognitive radio network – advanced technology demonstrator (…adaptive, spectrum efficient networks using emerging CR platforms)…also some common network facilities such as location & dynamic binding services

Each network at a different geographic location – new spectrum allocation may be needed at some sites

54

Advanced TechnologyDemonstrator (spectrum)

LocationService

NSF RadioTestbeds

Emulation &Simulation

Protocol &ScalingStudies

5 3

4

12

Other services

SensorNetworks

“Open” InternetConcepts forCellular devices

Embedded wireless,Real-world applicationsBroadband

Services,Mobile Computing

NSF GENI Implementation Wireless Sub-Networks Overview

Ad-HocMesh

Network

EmergingTechnologies

(cognitive radio)

GENIInfrastructure

Open APIWide-AreaNetworks

55

Cognitive Radio in NSF’s GENI ProjectPropose to build advanced technology demonstrator of cognitive radio networks for reliable wide-area services (over a ~50 Km**2 coverage area) with spectrum sharing, adaptive networking, etc.

Basic building block is a cognitive radio platform, to be selected from competing research projects now in progress and/or future proposalsRequires enhanced software interfaces for control of radio PHY, discovery and bootstrapping, adaptive network protocols …….. suitable for protocol virtualizationFCC experimental license for new cognitive radio band

Cognitive Radio Network Node

Cognitive Radio Client

Cognitive Radio Network Node

Cognitive Radio Client

Connections to GENIInfrastructure

Spectrum MonitorsSpectrum Server

Research Focus:1. New technology validation of cognitive

radio2. Protocols for adaptive PHY radio

networks3. Efficient spectrum sharing methods4. Interference avoidance and spectrum

etiquette5. Dynamic spectrum measurement6. Hardware platform performance studies

56

Concluding RemarksFuture wireless networks need ~100-1000x increases in density and bit-rate of radios motivates better spectrum coordination methodsSpot shortages of spectrum will occur if present static allocation is continued significant improvement achieved with dynamic allocationCognitive radio technologies can be characterized in terms of the combination of hardware complexity & level of protocol coordinationPromising cognitive radio schemes include

Agile radio with interference avoidance

Spectrum etiquette protocols: spectrum server, CSCC..

Adaptive networks via ad-hoc collaboration

Early technical results now available for some of these methods, but very different complexity factors and market implications…

57

Concluding RemarksFuture research areas in cognitive radio include:

New concepts and algorithms for agile radio and spectrum etiquette protocolsArchitecture and design of adaptive wireless networks based on cognitive radiosDetailed evaluation of large-scale cognitive radio systems using alternative methodsSpectrum measurement and field validation of proposed methodsCognitive radio hardware and software platforms

User-level field trials of emerging cognitive radios and related algorithms/protocols may also be useful to gain experience

Controlled testbed experiments comparing different co-existence methodsLarge-scale “spectrum server” trial for 802.11x coordinationExperimental deployments in proposed US FCC cognitive radio band

Success with cognitive radio technologies should lead to major improvements in spectrum efficiency, performance and interoperability