O pen Dy namic S pectum S haring with minimal E nerg Y

Open Dynamic Spectum Sharing with minimal EnergY

Our Odyssey* towards

Cognitive Radio

*1. An extended adventurous voyage or trip. 2. An intellectual or spiritual quest.

A range of technologiesfor a range of requirements

1995 2000 2005 2010

10 kbps 100 kbps 1 Mbps 10 Mbps 100 Mbps 1 Gbps

WIMAX

Low speed/

Stationary

2G(digital)

3GMultimedia

3G+

1G(analog)

Medium speed

802.16e

2.4 GHzWLAN

5 GHzWLAN

High rate WLAN

GSMCDMAone

Bluetooth

60 GHz WPAN

4Gresearch

target

UMTSCDMA2000

High speed GPRS

EDGE

3GPP -LTE+

UWB WPAN

1995 2000 2005 2010

10 kbps 100 kbps 1 Mbps 10 Mbps 100 Mbps 1 Gbps

WIMAX

Low speed/

Stationary

2G(digital)

3GMultimedia

3G+

1G(analog)

Medium speed

802.16e

2.4 GHzWLAN

2.4 GHzWLAN

5 GHzWLAN5 GHzWLAN

High rate WLAN

GSMCDMAone

Bluetooth

60 GHz WPAN

4Gresearch

target

UMTSCDMA2000

High speed GPRS

EDGE

3GPP -LTE+

UWB WPAN

Many technologies share the same spectrum

2.4 GHz example

5 MHz

Sensors0dBm

2 MHz

2.4 GHz

WLAN~15dBm

22 MHz

2.4 GHz

Bluetooth0dBm

1 MHz

2.4 GHz

Spectrum Sharing Further Classification in 4 Sub Projects

OdysseySpectrum Sharing

Heterogeneous networks Time/Space/Frequency

Adaptation Capabilites

Pol

icie

s/P

riorit

ies

Only one adapts (asymm)

Het

erog

eneo

us

Cap

abili

ties

Heterogeneous

Policies

PER: 10%

802.15.4 Low load

+802.11

SU should Not harm

PU

PU QoS

SU

QoS

Only SU acts

Both adapt (symm)

Het

erog

eneo

us

Pol

icie

s

Heterogeneous

Capabilities

802.11 Packets per Second

0

50

100

150

200

250

640 320 0

802.15.4 Rate

Packets per Second

802.11 at 1Mbps

802.11 at 5.5MbpsPunishment

802.15.4 High load

+ 802.11

802.11ResourcePlanning

With priorities PU QoS

SU

QoS

PU-SU game

Capturing context or spectrum information is costly

• Active observations: try out a configuration

trade-off between

exploitation

and

exploration

• Passive observations– Require extra hardware or introduce delay– Energy costly

Requir

ed B

W

Channel Sequence

Active observations:Gambling for the best channel

PER: 5% PER: 10% PER: 50%

• Just send packets and learn the best performing channel

• No (wrong) model, no (false) assumptions, perfect adaptation to any situation

• If you do not have to obey a policy rule

Multi-armed bandit theory gives us the optimal strategy

802.15.4 searches it’s best channel in ISM band (11-26)

Predicting PU idle time

Switching to a channel that will remain idle longerminimizes switching overhead

Improve channel selection with efficient passive scanning

Measuring interference power by Nyquist sampling is expensive (ADC cost, large amount of data to transfer)

Information is only in the edges: new sampling method that converts analog signals directly into information

ADC

802.15.4 packets

Collision with 1 802.11packet

AIC

What is the best multichannel MAC design?

Spatial planningKeep-Out-Region in theory

Fie

ld S

tren

gth

41dBu

F(50,90)

Desired/Undesired:28dB for DTV

Required Sensitivity

F(50,10)

Distance SU from contour

Contour size

Pathloss trend in reality

Resulting Spatial Reuse much improved

PU QoS

SU

Qo

SOnly SUacts

Proposed a run-time technique to predict and flood pathloss trends and coverage

SU sensing requirements as function of SU density

Spectrum Sharing Further Classification in 4 Sub Projects

OdysseySpectrum Sharing

Heterogeneous networks Time/Space/Frequency

Adaptation Capabilites

Pol

icie

s/P

riorit

ies

Only one adapts (asymm)

Het

erog

eneo

us

Cap

abili

ties

Heterogeneous

Policies

PER: 10%

802.15.4 Low load

+802.11

SU should Not harm

PU

PU QoS

SU

QoS

Only SU acts

Both adapt (symm)

Het

erog

eneo

us

Pol

icie

s

Heterogeneous

Capabilities

802.11 Packets per Second

0

50

100

150

200

250

640 320 0

802.15.4 Rate

Packets per Second

802.11 at 1Mbps

802.11 at 5.5MbpsPunishment

802.15.4 High load

+ 802.11

802.11ResourcePlanning

With priorities PU QoS

SU

QoS

PU-SU game

Are there scenarios where both 802.15.4 and 802.11 want to adapt

• 802.15.4 harms 802.11 in measurements

802.15.4 onthen 802.11throughput down

Coexistence approach does not work

802.15.4 is not expected to harm 802.11 because:• 802.15.4 output power is >15dB lower than 802.11• 802.15.4 does a listen-before-send

802.11 packet @ 1Mbps

802.15.4 packet train pausedfor 802.11 transmission:802.15.4 detects 802.11

collision because 802.15.4 sensing resolution is coarse

PU identification: which PU present on which channels

802.11 at channel 1

802.11 at channel 6 802.15.4

No presence802.11 at

channel 1 and 6

we can identifydifferent PU’s

SU QoS depends on PU requirements and sensing performance

Can we build a PU-SU game?

PU QoS

SU

Qo

SOnly SUacts

Open Dynamic Spectrum Sharing with minimal Energy

• Contibutors– UC Berkeley

• Maryam,Youwei, Ian, Carl, Mehdi, Nikhil, Wilson

– IMEC• Michael

– TU Delft• Przemek

– VTT• Marko

– Stanford University• Ali, Bart