Cooperative wireless networks Cooperative wireless networks based - - PowerPoint PPT Presentation

SLIDE 1

The 2004 International Workshop The 2004 International Workshop

• n Wireless Ad
• n Wireless Ad-
• hoc Networks

hoc Networks

INFOCOM INFOCOM Dpt Dpt. . University of Rome University of Rome “La “La Sapienza Sapienza” ”

Cooperative wireless networks Cooperative wireless networks based on based on distributed space time coding distributed space time coding

S.

• S. Barbarossa

Barbarossa, L. , L. Pescosolido Pescosolido, D. , D. Ludovici Ludovici, L. , L. Barbetta Barbetta, G. , G. Scutari Scutari

Oulu Oulu, Finland June 3, 2004 , Finland June 3, 2004

SLIDE 2

Context Context

Romantik Romantik Project Project R Res esO Ource urce M Management and anagement and A Adva dvaN Nced ced T Transceiver ransceiver Algor AlgorI Ithms thms for for Multihop Multihop Networ NetworK Ks s

Partners Partners: :

• Universitat Politecnica de Catalunya

Universitat Politecnica de Catalunya

• University of Bristol

University of Bristol

• University of Rome “La

University of Rome “La Sapienza Sapienza” ”

• DUNE

DUNE

• INTRACOM

INTRACOM

• FUJITSU Lab. of Europe LTD

FUJITSU Lab. of Europe LTD

• Telenor

Telenor

SLIDE 3

Overview Overview

• Cooperative communications

Cooperative communications

• Cooperation coding gain

Cooperation coding gain

• Distributed space

Distributed space-

• time coding strategies

time coding strategies

• Orthogonal STC ( OSTC )

Orthogonal STC ( OSTC )

• Full Rate Full Diversity ( FRFD )

Full Rate Full Diversity ( FRFD )

• V

V-

• BLAST

BLAST

• Performance

Performance

• Conclusion

Conclusion

SLIDE 4

State of the art State of the art

Cooperative transmission and DSTC

[ [A.Sendonaris,E.Erkip,B.Aazhang A.Sendonaris,E.Erkip,B.Aazhang ’98] ’98] cooperation among users in a wireless network cooperation among users in a wireless network can increase the capacity in the uplink multi can increase the capacity in the uplink multi-

• user

user channel channel coding strategies essentially based on time coding strategies essentially based on time repetition code and orthogonal channels with a repetition code and orthogonal channels with a loss in terms of rate loss in terms of rate [ [J.N.Laneman,G.W.Wornell J.N.Laneman,G.W.Wornell ’02,’03] ’02,’03] transmission strategy optimization in case of transmission strategy optimization in case of source and relay know the channel source and relay know the channel [ [S.Barbarossa,G.Scutari S.Barbarossa,G.Scutari ‘03] ‘03] approach merging the idea of cooperation with approach merging the idea of cooperation with space time coding for flat fading and frequency space time coding for flat fading and frequency selective channels selective channels [ [P.A.Anghel,G.Leus,M.Kaveh P.A.Anghel,G.Leus,M.Kaveh ‘03] ‘03] comparison of different DSTC techniques based comparison of different DSTC techniques based

• n D&F and A&F strategies in case of detection
• n D&F and A&F strategies in case of detection

errors at the relay nodes errors at the relay nodes [ [S.Barbarossa,G.Scutari S.Barbarossa,G.Scutari ‘04] ‘04]

SLIDE 5

Cooperation coding gain Cooperation coding gain

Example:

• Circle of radius R centered around 0 where there are n radio nodes
• C(x0,r) = circle of radius r, centered around the point x0
• Radio nodes distributed uniformly and independently of each other within C(0,R)
• p =r0

2 / R2 = probability of finding a node within a circle of radius r0

P{k dots out of n lie in C(x,r0) included in C(0,R)} (1 )

k n k

n p p k

−

  = − =    

Bernoulli distribution

SLIDE 6

Cooperation paradigms Cooperation paradigms

Coordinated multihop networking

Old paradigm source and relays transmit as a time-repetition code, distributed through space

D S R2 R1 R3

New paradigm cooperation between source and relays creates a virtual MIMO system

D S R2 R1 R3

use space-time coding instead of repetition coding

SLIDE 7

Distributed Distributed STC vs. STC vs. conventional conventional STC STC

• 1. Probability of finding a (reliable) relay
• 1. Probability of finding a (reliable) relay
• 2. Synchronization
• 2. Synchronization
• 3. Decision errors at the relay nodes
• 3. Decision errors at the relay nodes

SLIDE 8

Probability of finding reliable relays Probability of finding reliable relays

Average error rate:

( ) ( 1)

e r e k

P p k P k

∞ =

= +

∑

5 10 15 20 25 30 10

• 4

10

• 3

10

• 2

10

• 1

10 SNR (dB) Average BER k = 0 1 2 coop

error probability of a multiple transmit antenna having k+1 Tx antennas

# relays

Coding gain Probability of finding K relays

coding gain may be considerable if relay density is high

SLIDE 9

Cooperation coding gain Cooperation coding gain

0(0)

e r

c P p SNR ∝

At high SNR

2

r c

G e

πρ

=

• diversity gain = 1 (i.e. no gain)
• coding gain :

Gain depends on node density and on power used to discover a relay

Coding gain increases by increasing the transmit power or decreasing the bit rate

SLIDE 10

Cooperation protocol Cooperation protocol

Three main phases: 1. relay discovery ( RD ) 2. transmission from Source to the Relays ( S2R ) 3. coordinated transmission from (Source,Relays) to the Destination ( SR2D )

SLIDE 11

Resource Discovery Resource Discovery

• each source sends orthogonal pseudo-noise codes to verify whether there

are available neighbors

• radio nodes available as relays compute SNIR for each source
• potential relays retransmit an ACK signal back only to those sources

whose SNIR exceeds a given threshold

• each source receives the ACK and SNIR from all potential relays and

it decides which relay to use this operation has to be repeated at least this operation has to be repeated at least

• nce every channel coherence time
• nce every channel coherence time

SLIDE 12

Resource Discovery Resource Discovery

How far should be source and relays ? If source and relays are very close

• less power wasted in S2R time slot
• less synchronization problems in SR2D time slot
• less interference between different source/relay pairs

but …

it is less likely to find sufficiently reliable relays

SLIDE 13

Multi Multi-

• user scenario

user scenario

S S1

1

R R7

7

R R1

1

R R9

9

R R4

4

R R8

8

R R5

5

R R2

2

S S2

2

S S3

3

R R6

6

R R3

3

R R1

1

R R3

3

R R2

2

S2R transmission

SNIR constraint induces a circle of radius SNIR constraint induces a circle of radius D Dimax

imax

around every potential relay around every potential relay Ri Ri There is real cooperation between Si and Ri

• nly if Si is inside the circle of radius Dimax around Ri

All sources transmit simultaneously interference occurs among the active links

SLIDE 14

Multi Multi-

• user scenario

user scenario

R R1

1

R R3

3

R R2

2

S S1

1

R R7

7

R R9

9

R R4

4

R R8

8

R R5

5

S S2

2

S S3

3

R R6

6

D D1

1

D D3

3

D D3

3

SR2D transmission

virtual transmit arrays are induced between virtual transmit arrays are induced between source source-

• relay pairs and destinations

relay pairs and destinations

SLIDE 15

S2R transmission S2R transmission

for each source for each source S Si

i

if if S Si

i has found its own relay

has found its own relay R Ri

i in Relay Discovery phase

in Relay Discovery phase then then S Si

i transmits information to

transmits information to R Ri

i in a dedicated S2R slot

in a dedicated S2R slot

Rate loss resulting from the insertion of S2R slot Increasing constellation order, insertion loss decreases but … the probability of finding a relay decreases

Constellation order as trade Constellation order as trade-

• off rate loss / probability of finding relay
• ff rate loss / probability of finding relay

SLIDE 16

SR2D transmission SR2D transmission

for each pair ( for each pair (S Si

i,R

,Ri

i)

) S Si

i and

and R Ri

i transmit information to

transmit information to D Di

i inducing a virtual array using dedicated resources

inducing a virtual array using dedicated resources

no interference among different SR2D transmissions & Distributed Space Time Coding strategy & D&F: relays decode the received symbols and retransmit them

SLIDE 17

Alternative DSTC strategies Alternative DSTC strategies

(4) Barbarossa [‘04]

Trace-orthogonal STC(4)

potential max diversity gain and max rate, as a function of complexity max rate gain, limited receiver complexity low diversity gain

V-BLAST(1)

(1) Foschini [‘96]

max diversity gain, low rate min receiver complexity

Orthogonal STC(2)

(2) Tarokh et al. [‘96]

Full Rate Full Diversity (3)

max diversity gain and max rate gain, but high receiver complexity

(3) Giannakis et al., El Gamal et al. [‘03]

SLIDE 18

Rate loss Rate loss wrt wrt no no-

• coop system

coop system

• TDMA context
• TS2R = duration of S2R time slots
• TSR2D = duration of SR2D time slots
• Ts= symbol duration in all time slots
• Q = constellation order used in S2R
• M = constellation order used in SR2D
• S2R slot is shared among N source-relay pairs

2 2 2 SR D S R SR D

NT T NT η = +

Rate reduction factor

Rate loss can be reduced: by decreasing TS2R by increasing N

higher SNIR constraint at the relay

the right choice as a trade-off rate / performance

SLIDE 19

Main differences between DSTC and STC:

D S R D

• decoding errors at R node
• lack of synchronization of packets arriving at D
• probability of finding reliable relays

Assumptions:

• block

block transmission with Cyclic Prefix (CP), quasi CP), quasi-

• synchronism

synchronism

• No CSI @ TX,

No CSI @ TX, Perfect CSI @ Rx

• all channels are flat fading
• channel coefficients

2

(0, ),

h

σ Ν ~

2 , 1

2

≥ ∝ α σ

α

d

h

SLIDE 20

Distributed Orthogonal STC & Decode and forward Distributed Orthogonal STC & Decode and forward Protocol

S R D S

, + s(i) s(i 1) ˆ s(i) ˆ , s(i + 1)

R

Time Slot 1

R Tx S Tx time

i i+1 s(i) s(i +1)

S transmits

, +

T T

s (i) s (i 1)

R decodes the received symbols

ˆ , s(i) ˆ s(i +1)

SLIDE 21

Distributed Orthogonal STC & Decode and forward Distributed Orthogonal STC & Decode and forward Protocol

S R D S R D

Fs(i) ˆ Fs(i + 1)

*

ˆ −Gs (i)

*

Gs (i + 1)

R Tx S Tx time

i i+1 s(i) ) s(i 1 +

Time Slot 2

i+2 i+3 Fs(i)

*

Gs (i +1) ˆ Fs(i +1) (i) s G * ˆ −

S,R transmits

, s(i) s(i +1) using block Alamouti

(time reversal)

*

JF G = Bit Rate =

2 2 2

2 log b/s/Hz log 2N+2 log N M R M Q =

SLIDE 22

What about synchronization ? What about synchronization ? Adding a cyclic prefix that incorporates the Adding a cyclic prefix that incorporates the relative delays of packets coming from source relative delays of packets coming from source and relays, the system becomes immune of time and relays, the system becomes immune of time synchronization errors synchronization errors

η = + N N L

Price: rate reduction To reduce the loss, source and relays should be close

SLIDE 23

Distributed FRFD & Decode and forward Distributed FRFD & Decode and forward Protocol

S R D S

( )

3 4

θ ϕ − s s

1 2

ϕ − s s

R

( )

3 4

θ ϕ − s s

1 2

ϕ − s s Time Slot 1

R Tx S Tx time

i i+1 ( )

3 4

θ ϕ − s s

1 2

ϕ − s s

S transmits rotation coefficients *

, θ ϕ

with

( )

3 4

θ ϕ − s s

1 2

ϕ − s s R decodes the received symbols

( )

3 4

θ ϕ − s s

1 2

ϕ − s s * Ma, Giannakis [‘03]

SLIDE 24

Distributed FRFD & Decode and forward Distributed FRFD & Decode and forward

S R D S R

( )

3 4

θ ϕ + s s

( )

3 4

θ ϕ − s s

1 2

ϕ − s s

1 2

ϕ + s s

D

Protocol

R Tx S Tx time

i i+1 ( )

3 4

θ ϕ − s s

1 2

ϕ − s s

Time Slot 2

i+2 i+3

1 2

ϕ + s s

( )

3 4

θ ϕ + s s

1 2

ϕ − s s

( )

3 4

θ ϕ − s s

4 blocks are transmitted in 4 time slots no insertion loss Bit Rate =

2 2 2

log b/s/Hz lo 4 4 g 2N+ log N M R M Q =

SLIDE 25

Distributed BLAST & Decode and forward Distributed BLAST & Decode and forward Protocol

S R D

Time Slot 1

R Tx S Tx time

i i+1

2

s

4

s

S

2

s

4

s

R

ˆ 2 s

4

ˆ s

Assumption: two independent streams of data transmitted from tw Assumption: two independent streams of data transmitted from two antennas

• antennas

S transmits and

2

s

4

s R decodes the received symbols and

4

ˆ s ˆ 2 s

SLIDE 26

Distributed BLAST & Decode and forward Distributed BLAST & Decode and forward Protocol

S R D S R D

3

s

2

s

4

s

1

s

R Tx S Tx time

i i+1

2

s

4

s

Time Slot 2

i+2 i+3

1

s

3

s

4

s

2

s

D-BLAST requires that the relay receives only half of the bits to be transmitted duration reduction of S2R time slot wrt FRFD for a given Q and M, D-BLAST has the highest Bit Rate

Bit Rate =

2 2 2

log b/s/Hz lo 4 2 g 2N+ log N M R M Q =

SLIDE 27

Power allocation Power allocation

If

E= energy radiated by no-coop system

&

= % of energy devoted to S2R link

α

D D S R

E α

E α

= energy radiated by S in S2R TS

(1 ) 2 E α − (1 ) 2 E α −

(1 ) 2 E α − = energy radiated by S and R in SR2D TS

All systems transmit with the same overall radiated energy E

SLIDE 28

Performance Performance

Parameters of Simulation Parameters of Simulation

• N

N = # source = # source-

• relay pairs = 10

relay pairs = 10

• # random independent channels realizations = 16000

# random independent channels realizations = 16000

• SINR

SINRi

i at the useful relay = 20 dB, i=1..N

at the useful relay = 20 dB, i=1..N

• % of energy devoted to the S2R link= 10 %

% of energy devoted to the S2R link= 10 %

• Attenuation factor of transmitted power = 3.5

Attenuation factor of transmitted power = 3.5

• # antennas of sources & relays = 1; # antennas of destinati

# antennas of sources & relays = 1; # antennas of destinations = 1

• ns = 1
• 16

16-

• QAM constellation in S2R link

QAM constellation in S2R link

• relay density high enough to guarantee that each source inte

relay density high enough to guarantee that each source interacts with a relay racts with a relay

SLIDE 29

Comparison of alternative coding strategies: average BER

0(1)

1

r

p =

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

• 10
• 5

5 10 15 20 25 10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1

no coop 16-QAM V-BLAST 4-QAM OSTC 16-QAM FRFD 4-QAM no coop 4-QAM OSTC 4-QAM

R=2

D-OSTC outperforms no-coop system D-FDFR with 4-QAM achieves the lowest average BER

R=4

SLIDE 30

Comparison of alternative coding strategies: average Rate Comparison of alternative coding strategies: average Rate

0(1)

1

r

p =

no coop 16-QAM V-BLAST 4-QAM OSTC 16-QAM FRFD 4-QAM no coop 4-QAM OSTC 4-QAM

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 Average SNR at the Destination

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 40/21 Average SNR at the Destination

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 40/21 80/20 Average SNR at the Destination

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 40/21 80/20 80/22 Average SNR at the Destination

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 40/21 80/20 80/22 Average SNR at the Destination

• 10
• 5

5 10 15 20 25 0.5 1 1.5 2 2.5 3 3.5 4 Average Information Rate Averages over: 16000 channel realizations Block Length: Nb = 10 Number of Active Sources = 10 SNIR = 20 dB α = 0.1 40/20 40/21 80/20 80/22 80/21 Average SNR at the Destination

D-VBLAST is little better because it suffers from less insertion losses

SLIDE 31

Performance Performance

Parameters of Simulation Parameters of Simulation

• N

N = # sources = 10 ; N = # sources = 10 ; NR

R = 190 potential relays all scattered uniformly and

= 190 potential relays all scattered uniformly and independently of each other within a circular cell independently of each other within a circular cell

• circular cell of radius R = 200 m

circular cell of radius R = 200 m

• SINR

SINRi

i at the useful relay =

at the useful relay =

• % of energy devoted to the S2R link= 10 %

% of energy devoted to the S2R link= 10 %

• Attenuation factor of transmitted power = 3.5

Attenuation factor of transmitted power = 3.5

• # antennas of sources & relays = 1; # antennas of destinatio

# antennas of sources & relays = 1; # antennas of destinations = 1 ns = 1

• 16

16-

• QAM constellation in S2R link

QAM constellation in S2R link

r

12.5 , p (1) 0.72 dB =

r

15 dB, p (1) 0.65 =

SLIDE 32

Comparison of alternative coding strategies: average BER Comparison of alternative coding strategies: average BER

0(1)

1

r

p <

it depends from SNIR constraint

no coop 16-QAM SNIR = 12.5 dB V-BLAST 4-QAM SNIR = 12.5 dB OSTC 16-QAM SNIR = 12.5 dB FRFD 4-QAM SNIR = 12.5 dB FRFD 4-QAM SNIR = 15 dB V_BLAST 4-QAM SNIR = 15 dB OSTC 16-QAM SNIR = 15 dB

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

• 10
• 5

5 10 15 20 25 30 35 40 10

• 6

10

• 5

10

• 4

10

• 3

10

• 2

10

• 1

10 Average SNR at the Destination Average BER Averages over: 5 realizations of the Terminals' deployment 10( x 5 ) realizations of the Active Sources locations 20( x 10 x 5 ) channels realizations (for each Source) Block Length: Nb = 10 Number of Terminals = 200 Number of Active Sources = 10 α = 0.1

Increasing the SNIR constraint, the floor on BER decreases Increasing the SNIR constraint, the floor on BER decreases but it is less likely to find a reliable relay but it is less likely to find a reliable relay

SLIDE 33

Conclusion

Distributed space Distributed space-

• time coding

time coding provides provides

• efficient use of available radio resources

efficient use of available radio resources

• design flexibility: diversity / rate trade

design flexibility: diversity / rate trade-

• off
• ff
• cooperation should occur between nearby nodes

cooperation should occur between nearby nodes

• impact of additional signaling for coordinated

impact of additional signaling for coordinated transmission should be properly assessed transmission should be properly assessed