Opportunistic Communication

7/30/2019 Opportunistic Communication

1/48

Opportunistic Communication:

Smart Scheduling and Dumb Antennas

David Tse

Department of EECS, U.C. Berkeley

January 28, 2002

Intel


2/48

Communication over Wireless Channels

Fundamental characteristic of wireless channels: fading.

A modern view of communication over fading channels is

emerging.

This view has ramifications to the design of not only the physical

layer but to the design of the entire wireless network.


3/48

Opportunistic Communication

Smart Scheduling

Downlink scheduling for Qualcomms HDR (High Data Rate) system.

(Tse 99)

Dumb Antennas

Opportunistic beamforming using multiple transmit antennas

(Viswanath, Tse and Laroia 2001)


4/48

Wireless Fading Channels

Channel Quality

Time

fading due to constructive and destructive interference between

multiple signal paths;

Rayleigh fading: superposition of many small paths

Rician fading: many small paths plus one dominant path


5/48

Qualcomm HDRs DownLink

HDR (1xEV-DO): a wireless data system operating on IS-95 band (1.25

MHz)

Fixed Transmit

Power

User 2

User 1

Base Station

Data

Measure Channel

Request Rate

HDR downlink operates on a time-division basis.

Scheduler decides which user to serve in each time-slot.


6/48

Downlink Multiuser Fading Channel

Fading Channel

MobileUser 1

User 2

User KBase Station

What is the sum capacity with channel state feedback?


7/48

Information Theoretic Capacity of Downlink

(Tse 97)

2 4 6 8 10 12 14 160

0.5

1

1.5

2

2.5

Rayleigh Fading

Number of Users

Totalspectalefficienyi

n

bps/Hz

Each user undergoes independent Rayleigh fading with average received

signal-to-noise ratio SNR = 0dB.


8/48

To Fade or Not to Fade?

2 4 6 8 10 12 14 160

0.5

1

1.5

2

2.5

AWGN Channel

Rayleigh Fading

Number of Users

TotalSpectralEfficien

yin

bps/Hz

Sum Capacity of fading channel much larger than non-faded channel!


9/48

Multiuser Diversity

0 200 400 600 800 1000 1200 1400 1600 1800 20000

500

1000

1500

time slots

requested

rate

(kbps)

symmetric channels

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.

effective SNR at time t = max1kK

|hk(t)|2.


10/48

Multiuser Diversity

0 200 400 600 800 1000 1200 1400 1600 1800 20000

500

1000

1500

time slots

requestedr

ate(

kbps)

symmetric channels

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.

effective SNR at time t = max1kK

|hk(t)|2.


11/48

Multiuser Diversity

Diversity in wireless systems arises from independent signal paths.

Traditional forms of diversity includes time, frequency andantennas.

Multiuser diversity arises from independent fading channels across

different users.

Fundamental difference: Traditional diversity modes pertain to

point-to-point links, while multiuser diversity provides

network-wide benefit.


12/48

Multiuser Diversity

Diversity in wireless systems arises from independent signal paths.

Traditional forms of diversity includes time, frequency andantennas.

Multiuser diversity arises from independent fading channels across

different users.

Fundamental difference: Traditional diversity modes pertain to

point-to-point links, while multiuser diversity provides

network-wide benefit.


13/48

Fairness and Delay

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requested

rate

(k

bps)

asymmetric channels

tc

Challenge is to exploit multiuser diversity while sharing the benefits

fairly and timely to users with asymmetric channel statistics.


14/48

Hitting the Peaks

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requestedr

ate

(kbps)

asymmetric channels

tc

Want to serve each user when it is near its peak within a latency

time-scale tc.

In a large system, at any time there is likely to be a user whose

channel is near its peak.


15/48

Hitting the Peaks

1000 1500 2000 25000

500

1000

1500

2000

2500

time slots

requestedr

ate

(kbps)

asymmetric channels

tc

Want to serve each user when it is near its peak within a latency

time-scale tc.

In a large system, at any time there is likely to be a user whose

channel is near its peak.


16/48

Proportional Fair Scheduler

At time slot t, given

1) users average throughputs T1(t), T2(t), . . . , T K(t) in a past window.

2) current requested rates R1(t), R2(t), . . . , RK(t)transmit to the user k with the largest

Rk(t)

Tk(t).

Average throughputs Tk(t) can be updated by an exponential filter with

time constant tc.


17/48

Comments

If users have symmetric channel statistics, this reduces to the

greedy policy of transmitting to the mobile with the highestrequested rate.

If channels have different statistics, competition for resource is

made fair by normalization

feedback is built into the metric Rk(t)/Tk(t) to provide a fair

bandwidth allocation over the time-scale tc.


18/48

Comparison with Round-Robin Policy

Round-Robin Policy

Give same number of time slots to all the users in a round-robinfashion, regardless of their channel conditions.

Proportional fair policy:

Give roughly the same number of time slots to all users, but try to

transmit to a user when its channel condition is near its peak.

Resource fair, but not necessarily performance fair.


19/48

Throughput of HDR Scheduler: Symmetric Users

2 4 6 8 10 12 14 160

100

200

300

400

500

600

700

800

900

1000

1100

number of users

totalthrough

put(kbps)

mobile environment

fixed environment

round robin

average SNR = 0dB

timescale tc = 1.6sec

latency

Mobile environment: 3 km/hr, Rayleigh fading

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


20/48

Channel Dynamics

0 1000 2000 30000

200

400

600

800

1000

1200

1400

1.6 sec

requested

rate

ofa

user(kbps)

time slots

mobile environment

0 1000 2000 30000

200

400

600

800

1000

1200

1400

1.6 sec

time slots

requested

rate

ofa

user(kbps)

fixed environment

Channel varies faster and has more dynamic range in mobile

environments.


21/48

Throughput of Scheduler: Asymmetric Users

(Jalali, Padovani and Pankaj 2000)


22/48

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?


23/48

Dumb Antennas

Received signal at user k:

(t)h1k(t) +

1 (t) exp(j(t))h2k(t)

x(t).


24/48

Slow Fading Environment: Before

0 500 1000 1500 2000 2500 300080

100

120

140

160

180

200

220

Time Slots

SupportableRate

User 1

User 2


25/48

After

0 500 1000 1500 2000 2500 300050

0

50

100

150

200

250

300

350

Time Slots

SupportableRate

User 1

User 2


26/48

Opportunistic Beamforming: Slow Fading

0 5 10 15 20 25 30 350.8

0.9

1

1.1

1.2

1.3

1.4

1.5

Number of Users

AverageThr

oughputinbps/Hz

Opp. BF

Coherent BF

Consider first a slow fading environment when channels of the

users are fixed (but random).

Dumb antennas can approach the performance of true

beamforming when there are many users in the systems.


27/48

Opportunistic versus True Beamforming

If the gains h1k and h2k are known at the transmitter, then true

beamforming can be performed:

=

| h1k |2

| h1k |2 + | h2k |2

= h1k h2k

Dumb antennas randomly sweep out a beam and opportunistically

sends data to the user closest to the beam.

Opportunistic beamforming can approach the performance of true

beamforming when there are many users in the systems, but with

much less feedback and channel measurements.


28/48

Opportunistic versus True Beamforming

If the gains h1k and h2k are known at the transmitter, then true

beamforming can be performed:

=

| h1k |2

| h1k |2 + | h2k |2

= h1k h2k

Dumb antennas randomly sweep out a beam and opportunistically

sends data to the user closest to the beam.

Opportunistic beamforming can approach the performance of true

beamforming when there are many users in the systems, but with

much less feedback and channel measurements.


29/48

Asymptotic Result

Assume that the slow fading states of each user are i.i.d. randomly

generated (but fixed for all time).

In a large system of K users, with high probability, the users achievethroughputs

Tk 1

KRbfk

, k = 1, . . . K

.

where Rbfk

is the rate user k gets when it is perfectly beamformed to.


30/48

Opportunistic Beamforming: Fast Fading

0 1 2 3 4 5 6 7

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

with opp. bf.

without opp. bf.

Channel Amplitude

Density

Improves performance in fast fading Rician environments by spreading

the fading distribution.


31/48

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

totalthroughp

ut(kbps)

mobile

fixed

fixed but with opp. beamforming

latency timescale tc

= 1.6s

average SNR = 0 dB

Mobile environment: 3 km/hr, Rayleigh fading

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


32/48

Space Time Codes

Space time codes: intelligent use of transmit diversity to improve

reliability of point-to-point links.

For 2 transmit antennas, Alamouti scheme is the best space-time

code.

Let us compare smart and dumb antennas in terms of both

performance and complexity.


33/48

Comparison: Performance

Slow Fading:

Alamouti: diversity gain

dumb antennas: diversity gain plus 3 dB power gain

Fast Fading:

Alamouti: reduces channel fluctuations and thereby reduces the

multiuser diversity gain.

dumb antennas: keeps the fluctuations the same in Rayleigh

fading and increases the fluctuations in Rician fading.


34/48

Comparison: Performance

Slow Fading:

Alamouti: diversity gain

dumb antennas: diversity gain plus 3 dB power gain

Fast Fading:

Alamouti: reduces channel fluctuations and thereby reduces the

multiuser diversity gain.

dumb antennas: keeps the fluctuations the same in Rayleigh

fading and increases the fluctuations in Rician fading.


35/48

Comparison: Complexity

Alamouti:

requires two separate pilots to estimate the multi-antenna channel.

special encoder/decoder.

Dumb Antennas:

only requires a single pilot to estimate the overall channel SNR.

no special encoder/decoder.

In fact the mobiles are completely oblivious to the existence ofmultiple transmit antennas.


36/48

Comparison: Complexity

Alamouti:

requires two separate pilots to estimate the multi-antenna channel.

special encoder/decoder.

Dumb Antennas:

only requires a single pilot to estimate the overall channel SNR.

no special encoder/decoder.

In fact the mobiles are completely oblivious to the existence ofmultiple transmit antennas.


37/48

Cellular System: Opportunistic Nulling

In a cellular systems, users are scheduled when their channel is

strong and the interference from adjacent base-stations is weak. Multiuser diversity allows interference avoidance.

Dumb antennas provides opportunistic nulling for users in other

cells.

Particularly important in interference-limited systems with no soft

handoff.


38/48





cells.


handoff.


39/48





cells.


handoff.


40/48

Traditional CDMA Downlink Design

orthogonalize users (via spreading codes)

Makes individual point-to-point links reliable by averaging:

interleaving

multipath combining,

soft handoff

Role of transmit antennas is to provide further link diversity.

Important for voice with very tight latency requirements.


41/48




interleaving


soft handoff




42/48




interleaving


soft handoff




43/48




interleaving


soft handoff




44/48

Downlink Design: Modern View

Shifts from the point-to-point view to a multiuser network view.

Wants large and fast fluctuations of both channel and interference

so that we can ride the peaks.

Role of transmit antennas is to amplify the fluctuations.

Exploits more relaxed latency requirements of data as well as MAC

layer packet scheduling mechanisms.


45/48









46/48









47/48









48/48

A Broader Perspective

Efforts on increasing wireless capacity has been on boosting

spectral efficiency of point-to-point links.

Rely on sophisticated physical layer signal processing techniques:

smart antennas, interference suppression, etc.....

Future progress will come from taking a broader network

perspective.

Opportunistic Communication

Documents

Transcript of Opportunistic Communication