Opportunistic Communication:

From Theory to Practice

David Tse

Department of EECS, U.C. Berkeley

March 9, 2005

Viterbi Conference

Fundamental Feature of Wireless Channels:

Time Variation

Time

time-varying channel

StrengthChannel

• multipath fading

• large-scale channel variations

• time-varying interference

Traditional Approach to Wireless System Design

constant channel

Time

time-varying channel

StrengthChannel

StrengthChannel

Time

Compensates for channel fluctuations.

Case Study: CDMA Systems

Two main compensating mechanisms:

1. Channel diversity:

• time diversity via coding and interleaving

• frequency diversity via Rake combining,

• macro-diversity via soft handoff

• transmit/receive antenna diversity

2. Interference management:

• power control

• interference averaging

Case Study: CDMA Systems

Two main compensating mechanisms:

1. Channel diversity:

• time diversity via coding and interleaving

• frequency diversity via Rake combining,

• macro-diversity via soft handoff

• transmit/receive antenna diversity

2. Interference management:

• power control

• interference averaging

What Drives this Approach?

constant channel

Time

time-varying channel

StrengthChannel

StrengthChannel

Time

Main application is voice, with very tight latency requirements.

Needs a consistent channel.

Opportunistic Communication: A Different View

Transmit more when and where the channel is good.

Exploits fading to achieve higher long-term throughput, but no

guarantee that the ”channel is always there”.

Appropriate for data with laxer latency requirements.

Opportunistic Communication: A Different View

Transmit more when and where the channel is good.

Exploits fading to achieve higher long-term throughput, but no

guarantee that the ”channel is always there”.

Appropriate for data with laxer latency requirements.

Opportunistic Communication: A Different View

Transmit more when and where the channel is good.

Exploits fading to achieve higher long-term throughput, but no

guarantee that the ”channel is always there”.

Appropriate for data with laxer latency requirements.

Point-to-Point Fading Channels

��

���� ��� ���

��� �������

��� �����

��� ���

Capacity-achieving strategy is waterfilling over time. (Goldsmith and

Varaiya 97)

Performance over Rayleigh Channel

� � � � ��� ���

����� ���

���

�������

����� ��� �"!� �"!$#

� � %�&�')(+*-,)�

�� ���� ���� ���

.

/

�

0

1

�

�

Performance: Low SNR

-10 -5 0 5

Full CSI

10

SNR [dB]

CSIR

C

Cawgn

0.5-15-20

3

2.5

2

1.5

1

At low SNR, capacity can be greater when there is fading.

Hitting the Peaks

(a)

(b)

(a)

(b)

Optimal AllocationNear Optimal Allocation

N0|h[m]|2

Time m

Time m

P [m]

1λ

N0|h[m]|2

N0|h[m]|2

Time m

Time m

1λ

P [m]

1λ

1λ

N0|h[m]|2

At low SNR, one can transmits only when the channel is at its peak.

Primarily a power gain.

In practice, hard to realize such gains due to difficulty in tracking the

channel when transmitting so infrequently.

Hitting the Peaks

(a)

(b)

(a)

(b)

Optimal AllocationNear Optimal Allocation

N0|h[m]|2

Time m

Time m

P [m]

1λ

N0|h[m]|2

N0|h[m]|2

Time m

Time m

1λ

P [m]

1λ

1λ

N0|h[m]|2

At low SNR, one can transmits only when the channel is at its peak.

Primarily a power gain.

In practice, hard to realize such gains due to difficulty in tracking the

channel when transmitting so infrequently.

Multiuser Opportunistic Communication

Fading Channel

MobileUser 1

User 2

User KBase Station

(Knopp and Humblet 95, T 97)

Performance

2

4

6

8

AWGNCSIRFull CSI

C [bits/s/Hz]

SNR [dB]

sum

K=16

K=4

K=2

K=1

Multiuser Diversity

Total average SNR = 0 dB.

2 4 6 8 10 12 14 160

0.5

1

1.5

2

2.5

AWGN Channel

Rayleigh Fading

Number of Users

Tot

al S

pect

ral E

ffici

eny

in b

ps/H

z

• In a large system with users fading independently, there is likely to

be a user with a very good channel at any time.

• Long term total throughput can be maximized by always serving

the user with the strongest channel.

Multiuser Diversity: A More Insightful Look

0 50 100 150 200 250 3000.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time Slots

Req

uest

ed r

ates

in b

ps/H

z

• Independent fading makes it likely that users peak at different

times.

• In a wideband system with many users, each user operates at low

average SNR, effectively accessing the channel only when it is near

its peak.

• In the downlink, channel tracking can be done via a strong pilot

amortized between all users.

Multiuser Diversity: A More Insightful Look

0 50 100 150 200 250 3000.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time Slots

Req

uest

ed r

ates

in b

ps/H

z

• Independent fading makes it likely that users peak at different

times.

• In a wideband system with many users, each user operates at low

average SNR, effectively accessing the channel only when it is near

its peak.

• In the downlink, channel tracking can be done via a strong pilot

amortized between all users.

Multiuser Diversity: A More Insightful Look

0 50 100 150 200 250 3000.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time Slots

Req

uest

ed r

ates

in b

ps/H

z

• Independent fading makes it likely that users peak at different

times.

• In a wideband system with many users, each user operates at low

average SNR, effectively accessing the channel only when it is near

its peak.

• In the downlink, channel tracking can be done via a strong pilot

amortized between all users.

1x EV-DO’s DownLink

Fixed Transmit Power

User 2

User 1

Base Station

Data

Measure Channel Request Rate

Information theory suggests that resource should be scheduled in a

channel-dependent way.

Challenge is to exploit multiuser diversity while sharing the benefits

fairly and timely to users.

1x EV-DO’s DownLink

Fixed Transmit Power

User 2

User 1

Base Station

Data

Measure Channel Request Rate

Information theory suggests that resource should be scheduled in a

channel-dependent way.

Challenge is to exploit multiuser diversity while sharing the benefits

fairly and timely to users.

Symmetric Users

0 200 400 600 800 1000 1200 1400 1600 1800 20000

500

1000

1500

time slots

requ

este

d ra

te (

kbps

)

symmetric channels

Serving the best user at each time is also fair in terms of long-term

throughputs.

Asymmetric Users

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requ

este

d ra

te (

kbps

)

asymmetric channels

tc

• Want to serve each user when it is near its peak.

• A peak should be defined with respect to a latency time-scale tc.

Asymmetric Users

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requ

este

d ra

te (

kbps

)

asymmetric channels

tc

• Want to serve each user when it is near its peak.

• A peak should be defined with respect to a latency time-scale tc.

Asymmetric Users

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requ

este

d ra

te (

kbps

)

asymmetric channels

tc

• Want to serve each user when it is near its peak.

• A peak should be defined with respect to a latency time-scale tc.

Proportional Fair Scheduling

(T 99)

Schedule the user with the highest ratio Rk/Tk, where

Rk = current requested rate of user k

Tk = average throughput in past tc time slots

Performance

2 4 6 8 10 12 14 160

100

200

300

400

500

600

700

800

900

1000

1100

Low mobility environment

Fixed environment

Number of users

Tot

al th

roug

hput

(kb

ps)

High mobility environment

latency time scale tc = 1.6s

Average SNR = 0dB

Fixed environment: 2Hz Rician fading with Efixed/Escattered = 5.

Low Mobility environment: 3 km/hr, Rayleigh fading

High mobility environment: 30 km/hr, Rayleigh fading

Channel Dynamics

0 1000 2000 30000

200

400

600

800

1000

1200

1400

1.6 sec

requ

este

d ra

te o

f a u

ser

(kbp

s)

time slots

mobile environment

0 1000 2000 30000

200

400

600

800

1000

1200

1400

1.6 sec

time slots

requ

este

d ra

te o

f a u

ser

(kbp

s)

fixed environment

Channel varies faster and has more dynamic range in mobile

environments.

Inducing Randomness

• Scheduling algorithm exploits the nature-given channel fluctuations

by hitting the peaks.

• If there are not enough fluctuations, why not purposely induce

them?

Dumb Antennas

(Viswanath, T and Laroia 02)

���������

� �������

����������

� ����� ������������ !#"

$ �����

� � �����

The information-bearing signal at each of the transmit antennas are

multiplied by a random complex gain.

Inducing Randomness

transmission times

opportunisticbeamforming

t

Strength

Channel

t

Strength

Channel

user 1

user 2

afteropportunisticbeamforming

Strength

Channel

t

Strength

Channel

t

before

Slow Fading: Opportunistic Beamforming

user 2

user 1

• Dumb antennas create a beam in random time-varying direction.

• In a large system, there is likely to be a user near the beam at any

one time.

• By transmitting to that user, close to true beamforming

performance is achieved.

Fast Fading

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Rayleigh

2 antenna, Ricean

1 antenna, Ricean

Channel Amplitude

Den

sity

Improves performance in fast fading Rician environments by spreading

the fading distribution.

Overall Performance Improvement

2 4 6 8 10 12 14 160

100

200

300

400

500

600

700

800

900

1000

1100

number of users

tota

l thr

ough

put (

kbps

)

mobile

fixed

fixed but with opp. beamforming

latency time−scale tc = 1.6s

average SNR = 0 dB

Mobile environment: 3 km/hr, Rayleigh fading

Fixed environment: 2Hz Rician fading with Efixed/Escattered = 5.

Smart vs Dumb Antennas

• Space-time codes increase reliability of point-to-point links but

decreases multiuser diversity gains.

• Dumb antennas add fluctuations to point-to-point links but

increases multiuser diversity gains.

Smart vs Dumb Antennas

• Space-time codes increase reliability of point-to-point links but

decreases multiuser diversity gains.

• Dumb antennas add fluctuations to point-to-point links but

increases multiuser diversity gains.

Conclusions

• Implementation of a new point of view has to obey system

constraints.

• The new point of view impacts rest of the system design and

suggests new research problems.

• Interplay between theory and system is what makes

communications research so fun!

Conclusions

• Implementation of a new point of view has to obey system

constraints.

• The new point of view impacts rest of the system design and

suggests new research problems.

• Interplay between theory and system is what makes

communications research so fun!

Conclusions

• Implementation of a new point of view has to obey system

constraints.

• The new point of view impacts rest of the system design and

suggests new research problems.

• Interplay between theory and system is what makes

communications research so fun!