1

Adaptive Wireless Networks Using Cognitive Radios as a Building BlockMobiCom 2004 Keynote SpeechSept 29, Philadelphia

D. RaychaudhuriProfessor ECE Dept & Director, WINLAB

Rutgers Universityray@winlab.rutgers.eduwww.winlab.rutgers.edu

2

Talk Outline

Introduction: the future wireless network and related R&D challengesDynamic spectrum management & cognitive radio conceptsCognitive radio technologies: selected results

Coexistence of 802.11 and 802.16 in unlicensed bandsCSCC spectrum etiquetteAdaptive networks and ad-hoc self-organizationCognitive radio hardware

Concluding remarks

3

Introduction

4

Introduction: Future Wireless Network Scenario

Internet (IP-based)

Infostationcache

WLANAccess Point

WLANHot-Spot

VOIP(multi-mode)

Low-tier clusters(e.g. low power 802.11 sensor)

Ad-hocnetwork

extension

Public Switched Network(PSTN)

BTS

High-speed data & VOIP

Broadband Media cluster(e.g. UWB or MIMO)

BTS

BSC

MSC

CustomMobileInfrastructure(e.g. GSM, 3G)

CDMA, GSMor 3G radio access network

Generic mobile infrastructure

Today Future

GGSN,etc.

Voice(legacy)

High-speed data & VOIP

Relay node

Growing role for fast, low-costshort-range radios

Heterogeneous systems with multipleradio standards (3G, 4G, WLAN, UWB..)

Self-organizing ad-hoc access networksIncreasing use of unlicensed spectrum

and dynamic sharing methodsUniform IP core networkWide range of applications (

“ubiquitous wireless services”)

5

Introduction: Wireless Technology TrendsPervasive systems

~1 Mbps 3G/WCDMA~10-54 Mbps WLAN

~100 Mbps+4G/OFDM, 802.16 &

WLAN;~500 Mbps UWB, etc

IP-based networksfor both

Cellular & WLAN

IP+ Layer 7 overlayinfrastructure net;

Ad-hoc low-tiernetworks

Telephony;Multimedia;

Mobile Internet

Telephony; Multimedia;

Mobile InternetSensor Nets

Telephony;PC/LAN

PrimaryApplications

RadioTechnology

NetworkArchitecture

2G/CDMA & TDMA~1 Mbps WLAN

Cellular networksEthernet + WLAN

~1995-2000 ~2000-2005 ~2010+

Beyond IP networks(e.g. content aware routing)

Cross-layer techniques

Self-organizing multi-hop

Sensor net applications,Embedded wireless devices

VOIP, H264, HTTP, etc.Location-aware services

Mobile IPv6, etc.

Higher speed,OFDM Very wideband signals

New spectrum policiesCognitive radio

Adaptive RadioNetworks

6

Introduction: Key Technologies for Future Wireless Systems

New radios for heterogeneous accessLow-power sensor radiosHigh-speed WLAN and 4G/802.16Faster 4G cellular, 802.16, etc.

Spectrum-sharing for dense networksDynamic spectrum/cognitive radio for frequency coordinationSpectrum etiquette protocols

Ad-hoc wireless networksSelf-organizing networks capable of scaling organicallyDiscovery, MAC and routing protocols for reliable ad-hoc services

Pervasive computing softwareDynamic binding of application agents and sensorsReal-time orchestration of sensors and actuators

Focus of this talk

7

Dynamic Spectrum Management & Cognitive Radio

8

Motivation for Dynamic Spectrum and Cognitive Radio Techniques:Static allocation of spectrum is inefficient

Slow, expensive process that cannot keep up with technology

Spectrum allocation rules that encourage innovation & efficiencyFree markets for spectrum, more unlicensed bands, new services, etc.

Anecdotal evidence of WLAN spectrum congestionUnlicensed systems need to scale and manage user “QoS”

Density of wireless devices will continue to increase~10x with home gadgets, ~100x with sensors/pervasive computing

Interoperability between proliferating radio standardsProgrammable radios that can form cooperating networks across multiple PHY’s

9

Spectrum Management: Frequency allocation today …

Source: FCC website

10

Spectrum Management: Policy ConceptsUnlicensed bands with spectrum etiquette

More ISM/U-NII bands with simple coordination rules

Property RightsFee simple ownership with non-interference easements

Spectrum clearinghousePackets are sent with access tokens with pricing determined by congestion

Open accessNo coordination rules, technology expected to evolve towards co-existence

Cognitive radio bandsAgile/smart radios capable of adaptive strategies for interference avoidance

11

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

12

Cognitive Radio: DefinitionsThe term “cognitive radio” used to denote new generation of adaptive wireless devices capable of dynamic spectrum coordination

Baseline capability includes spectrum scanning and frequency agilityFast adaptation of transmitted signal to fit into changing radio environmentCapable of higher-layer spectrum etiquette or negotiation protocolsMay also participate in ad-hoc networks formed with other cognitive radiosInteroperability with multiple radio technologies based on SDR capabilities

13

Cognitive Radio: R&D Status

Policy and technology R&D on cognitive radio still at an early stage. Recent activities include:

FCC notice of rulemaking for specific “underlay” data services in UHF TV bandsMore general notice of proposed rulemaking on new unlicensed cognitive bandsSoftware defined cognitive radios developed at Vanu Inc., GNU/UtahXG policy framework being developed by DARPASystem studies and prototyping at Mitre, Rutgers/WINLAB, Stevens, others….New National Science Foundation research initiative (“NeTS ProWIN”), 2004

Cognitive radio has the potential for significant improvements in spectrum efficiency, performance and interoperability between unlicensed band services

14

Cognitive Radio: Design SpaceBroad 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

15

Reactive (autonomous) methods may be 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 activity

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

Cognitive Radio: Reactive Algorithms

Frequencyagility DD DC

Scheduling

16

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

17

Cognitive Radio: Spectrum Policy ServerInternetInternet--based Spectrum Policy Serverbased Spectrum Policy Server can help to coordinate wireless networks

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

18

Cognitive Radio: Common Spectrum Coordination Channel (CSCC)

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

19

Adaptive Wireless Networks: Ad-Hoc CollaborationCognitive radios can organize themselves into a multia multi--hop adaptive hop adaptive network network in order to achieve better system performance

Multi-hop collaboration can increase spectrum efficiency, reduce powerconsumption and potentially also improve throughput Cognitive radio scans for active nodes and executes discovery algorithmBootstrapped PHY to selected nodes adapts to high bit-rate, low power/rangeControl protocol between nodes used to negotiate ad-hoc network parameters and to exchange routing tables

PHY A

PHY BPHY C

Control(e.g. CSCC)

Multi-mode radio PHYAd-Hoc Discovery

& Routing Capability

AA

BB

D

C

D

E

F

Bootstrapped PHY &control link

End-to-end routed pathFrom A to F

Adaptive WirelessNetwork Node

(…functionality can be quitechallenging!)

20

Adaptive Wireless Network: Simple cellular/WLAN example

Multi-mode (cellular, WLAN) or cognitive radio capable of ad-hoc association can be used to improve cellular srevicesSystem may also include provisioned “radio forwarding nodes”Radio paths (single or multi-hop) selected adaptively based on current cellular radio link quality and proximity to other nodes

BTS

ForwardingNode

Dual-ModeMobile

(relay node)

Dual-ModeMobile Ad-Hoc

Cluster 1

To wirednetwork

To wirednetwork Ad-Hoc

Cluster 2

Ad-HocCluster n

21

Cognitive Radio Techniques: Selected Research Results

22

Reactive Schemes: Case study of 802.11 & 16 in shared unlicensed band

802.11b(~500m) and 802.16a (~10Km) coexist by applying reactive schemes to avoid interference

Dynamic Frequency Selection (DFS):Radio Scans each channel and calculates interference power levelTypical scanning interval is averaged 100msChoose the channel with least interference for communication

Power Control (PC):The receiver senses interference power level and calculates the minimum required transmit power and feedback to the transmitterThe transmitter uses minimum transmit power for communication

23

802.11 & 16 Co-Existence: Simulation Parameters

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

24

802.11 & 16 Co-Existence: Power Control ResultsObservations:

802.16 throughput can improve up to 3 times at the expense of 802.11 throughput degradation < 10% (e.g. at D=2.5Km)If two systems are too near to each other, power control may not work

4 links for 802.11 hotspot, each has Poisson arrival with mean 3ms802.11 hotspot is 3Km away from 802.16 BS

2000 2200 2400 2600 2800 30000

50

100

150

200

250

300

350

400

450

802.

16 D

L Th

roug

hput

(Kbp

s)

802.16 BS-SS Distance (meters)

Both PC OFF Both PC ON

2000 2200 2400 2600 2800 3000

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

0.22

802.

11 h

otsp

ot A

vera

ge L

ink

thro

ughp

ut (M

bps)

802.16 BS-SS Distance (meters)

Both PC OFF Both PC ON

802.16 Down Link Throughput 802.11 Average Link Throughput

25

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

Service channels

Edge-of-bandcoordination channel

26

CSCC: Protocol Stack

CSCC-PHY: 1Mbps 802.11b with 10 mW power (~100 m range)

CSCC-MAC: Simple periodic broadcast with randomization (100ms~seconds) to eliminate repeated collisions

27

CSCC: Packet FormatDestaddr

Srcaddr

Type IE length IE(1) IE(n)

6B 6B 2B 2B 2B 2B

CSCC-PKT: A standard Ethernet packet format with control payload (consisting of variable length information elements)

. . . Duration (32b)

Service Time . . .Price_bid(8b)Priority (8b)

Channel(8b)Type (8b). . . and Description

. . . Device Name and Description (64bits) . . .

Device Name and . . .. . . MAC Address

CSCC radio (802.11) MAC Address (48bits). . .

0 8 16 24 31

Tx Pwr (8b) Rx Pwr (8b)

CSCC packet used in WLAN-Bluetooth prototype at WINLAB

28

CSCC: Proof-of-Concept Experiments

WLAN-BT Scenario

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

29

CSCC: Experimental Parameters

WLAN nodes Bluetooth nodes

Mobility Static without mobility BT1 static, BT2 position varies

Traffic Model 100M bytes data by TCP 1.5M bytes data using Stop-and-wait scheme

MAC protocol IEEE 802.11b at 11Mbps Bluetooth ACL data link

Data card Cisco Aironet 350 series DS (at channel #1)

Ericsson BT w/ USB(hopping over whole band)

CSCC MAC IEEE 802.11 & periodic announcements at 1Mbps

CSCC card Cisco Aironet 350 series DS (at channel #11)

30

CSCC Results: Throughput TracesObservations:

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

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

31

Adaptive Wireless Networks:Ad-hoc Discovery and Self-Organization

Spectrum etiquette channel used to initiate discovery and network bootstrapExpanded beacon signals transmitted by each radio in CSCCNote that each link may use a different PHY/MAC -> cognitive radio switches between links dynamically, while using mutually agreed routing protocol

D

C

A

B’s transmissionrange

End-to-end routed pathFrom A to D

B

C’s transmissionrange

B’s beacon(may bein CSCC)

C’s beacon(may bein CSCC)

D

Node ID PHY spec MAC spec NeighbortableTx power Routing

bootstrapExample of Beacon Payload

PHY A(~50 Mbps)

PHY B(~100 Mbps) Control connection

Data path

….

32

TransmitPower

HopsToAP

NodeType

SequenceNumber

Cluster ID

PacketType

NodeID

BroadcastMAC

SourceMAC

Beacon Frame Format

Low-tier access links(AP/FN Beacons, MN Associations, Data)

Ad-hoc infrastructure links between FNs and APs(AP/FN Beacons, FN Associations, Routing Exchanges, Data)

Forwarding Node (FN)

Access Point (AP)

FN

AP

FNcoverage

area

APcoverage

area

Low-tier(e.g. sensor)Mobile Node (MN)

FN

Self-organized ad-hoc network

MN

MN

MN

MN

MN

MNMN MN

Internet

FN

AP

Channel 4

Channel 2

Beacon

Transmit Power Required: 1mW

Beacon

Assoc

Transmit Power Required: 4mW

FN

AP

SN•Scan all channels•Associate with FN/AP•Send data

FN•Scan all channels•Find minimum delay links to AP•Set up routes to AP•Send beacons•Forward SN data

Adaptive Networks: Ad-hoc Discovery Protocol Implementation

WINLAB’s “SOHAN” 802.11-based ad-hoc prototype demonstrates aspects of self-organization that can be extended to cognitive radio

33

Adaptive Networks: Discovery Algorithm Performance Results

NS-2 extended to supportHierarchical net with APs, FNs, MNsMultiple interfacesMultiple 802.11 channels

Distributed and optimal centralized algorithms2 APs, 4 FNs, 10 SNsCBR traffic of 64 byte packets1000 m x 1000 m area1 Mbps 802.11 radios

..significant reductions in routing overhead & energy used

34

Adaptive Networks: Ad-Hoc Routing/Discovery Implementation

Protocol stack at each layer

35

Adaptive Networks : “SOHAN”Experimental Results

Experimental Setup

Packet delivery ratioGains in system capacity and per-user throughput achievable relative to WLAN BSS mode or comparable “flat”ad-hoc mesh, particularly when FN’suse multiple radio channels...

36

Cognitive Radio: Hardware Platforms

Next-generation software-defined radio supporting fast spectrum scanning, adaptive control of modulation waveforms and collaborative network processingFacilitates efficient unlicensed band coordination and multi-standard compatibility between radio devices

MPC8260

TMS320C6701XC2V6000FPGA

100BaseT EthernetMegarray

Connector-244 Configurable

I/O pins

Bell Laboratories Software Defined Radio (Baseband Processor)Courtesy of Dr. T. Sizer

37

Cognitive Radio: Hardware PlatformsVanu 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

38

Cognitive Radio: Hardware Platform

radio

BasebandFPGA

BasebandProcessor Core

(DSP)

SRAM

PacketFPGA

Clock Mgmt

A/D

D/A

A/D

D/A

A/D

D/A

Wakeup

Packet BufferDRAM)

Host(CR Strategies)

radio

radio

Local ethernet drop

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

Agile radioI/O

Software defined modem Network Processor

WINLAB’s “network centric” concept for cognitive radio prototype (..under development in collaboration with GA Tech & Lucent Bell Labs)

39

Concluding Remarks

40

Concluding Remarks: ORBIT TestbedOpen-access next-generation wireless network testbed being developed at Rutgers for NSF network research testbeds (NRT) programLarge scale “radio grid emulator” for evaluating new concepts for future wireless networks, e.g. ad-hoc networks, cognitive radio protocols, ...Also, outdoor “field trial network” with open-interface 3G & WLAN for real-world application work

Static radio node

Radio link emulation

ORBIT radio node

“Open”APIAccess Point(802.11b)

End-user devices

Ad-hoclink

3Gaccess

link

HighSpeed

Net

Firewall

MobilityServer

Wiredrouters

EmulatorMapping

“Open”API

3G BTS

Global InternetGlobal Internetns-2+ scripts

&code

downloads

ResearchUser of Testbed

1. Radio Grid for Lab Emulation 2. Field Trial Network

41

Concluding Remarks: ORBIT Radio Grid

42

ORBIT: Field Trial System

Lucent “Base Station Router”with IP interface

“Open API” 802.11a,b,gORBIT radio node

43

Concluding RemarksFuture wireless networks need ~100-1000x increases in density and bit-rate of radios motivates better spectrum coordination methods

Spot shortages of spectrum will occur if present static allocation is continued significant improvement achieved with dynamic allocation

Cognitive radio technologies can be characterized in terms of the combination of hardware complexity and level of protocol coordination

Possible cognitive radio schemes includeAgile 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…

44

Concluding RemarksFuture research areas in cognitive radio include:

New concepts and algorithms for agile radio and spectrum etiquette protocols

Architecture and design of adaptive wireless networks based on cognitive radios

Detailed evaluation of large-scale cognitive radio systems using alternative methods

Spectrum measurement and field validation of proposed methods

Cognitive 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 methods

Large-scale “spectrum server” trial for 802.11x coordination

Experimental deployments in proposed US FCC cognitive radio band

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