mimo coding and relayingewh.ieee.org/r10/xian/com/guan.pdfmimo coding and relaying prof guan, yong...

MIMO Coding and Relaying

Prof Guan, Yong LiangHead – Communication Engineering Division

Director - Positioning and Wireless Technology CentreSchool of Electrical and Electronic Engineering

June 2011

元永亮

Outline

Group-Decodable STBC with Rate 1

Fast-Group-Decodable STBC with Rate > 1

Shift-Orthogonal STBC

2-Step 2-Way MIMO Relaying

Y L Guan, PWTC, EEE, NTU, Singapore

EYLGuanHighlight

Group-Decodable (Quasi-Orthogonal) STBC

Prof Guan Yong LiangDirector - Positioning and Wireless Technology Centre

School of Electrical and Electronic Engineering

Nov 2009

Introduction:MIMO Communications

MIMO: Multiple-Input Multiple-Output Multi-antenna transmitter/receiver, space time processing. Benefits - Increased data rate (spatial multiplexing gain)

- Improved link reliability (transmit diversity gain) Bell Lab Layered Space Time (BLAST) code (1998), Alamouti’s code

(1998), Golden code (2005) …


hi,2

h1,1

ML

Dec

odin

g

Mod

ulat

i on

Dem

o dul

atio

n

.. bp .. b2 b1 2 1ˆ ˆ ˆ.. .. pb b b

rP(k) … rp(k) … r1(k)

CSI Estimator

DataBit

Stream

ReconstructedBit

Stream

cK .. cp .. c2 c1

ModulatedData

Symbols

STB

C C

( ) ( ) ( )1 .. .. R R R

N N NP pr r r

EstimiatedData

Symbols

2 1ˆ ˆ ˆ ˆ.. .. K pc c c c

,T RN Nh

rP(2) … rp(2) … r1(2)

rP(1) … rp(1) … r1(1)xP(1) … xp(1) … x1(1)

xP(2) … xp(2) … x1(2)

xP(i) … xp(i) … x1(i)

( ) ( ) ( )1 .. .. T T T

N N NP px x x

TransmitAntennas

ReceiveAntennas

Introduction:Transmit Diversity

Slow flat fading cause severe reception problems in wireless radio communication FEC decoding fails due to long error bursts Typically mitigated by Rx antenna diversity But Rx diversity may not be effective in mobile phones due to

insufficient antenna separation and cost concern.

Transmit diversity offers an attractive solution because Base stations can afford more antenna separation Cost of additional antennas at BS is amortized over all users


Introduction:Transmit Diversity

h1

h2

c1r1 r2

c2

-c2*

c1*

1 1 2 1 1* *

2 2 1 2 2

r c c hr c c h

1 1 2 1 1* * * *

2 2 1 2 2

r h h cr h h c

Hc1 1H*

2 2

2 2 * *1 2 1 1 1 2 2

* *2 22 2 1 1 21 2

ˆˆ

0

0

c rc r

h h c h hc h hh h

CH1. c1 and c2

decoupled2. MRC effect

achieved using multiple tx antennasHcHHc


Orthogonal STBC

2-antenna tx diversity code on last page = Alamouti’s STBC = an orthogonal space-time block code (O-STBC)

Strength of O-STBC Provide full tx diversity (diversity order = no. of tx ant) Linear symbol-by-symbol ML decoding single symbol decodable

O-STBC code is unitary:T

HNC CH H I


1 2 3* *

1 3 2* *2 3 1

* *3 2 1

00

00

c c cc c c

c c cc c c

GS4

* * * * * *1 1 2 2 3 3 4 4* *1 1 2 2 3 3 4 4

* *2 2 1 1 4 4 3 3

* * * * * *2 2 1 1 4 4 3 3* *3 3 4 4 1 1 2 2

* * * * * *3 3 4 4 1 1 2 2* *4 4 3 3 2 2 1 1

* * * * * *4 4 3 3 2 2 1 1

c c c c c c c cc c c c c c c cc c c c c c c cc c c c c c c cc c c c c c c cc c c c c c c cc c c c c c c cc c c c c c c c

C8

But O-STBC has low code rates Complex O-STBC (MPSK, QAM) have

code rate < 1 for diversity level > 2 Diversity level 3-4: max rate = ¾

Diversity level 5-8 max rate = ½ (for sq O-STBC)

Requires rate adaptation Complicates design of cooperative

relay network etc.

Orthogonal STBC


Quasi-Orthogonal STBC

1234

*2

*1

*4

*3

*3

*4

*1

*2

4321

cccccccccccccccc

C

ηcHr c

4

*3

*2

1

4

3

2

1

1234

*2

*1

*4

*3

*3

*4

*1

*2

4321

4

*3

*2

1

cccc

hhhhhhhhhhhh

hhhh

rrrr

rHc cHˆ

4

3

2

1

4

3

2

1

4

3

2

1

000000

00

ˆˆˆˆ

cccc

cccc


Quasi-Orthogonal STBC

4

3

2

1

4

3

2

1

4

3

2

1

000000

00

ˆˆˆˆ

cccc

cccc

4

1

4

1

4

1

ˆˆ

cc

cc

3

2

3

2

3

2

ˆˆ

cc

cc

QO-STBC: HcH Hc is block diagonal

Quasi-orthogonal group-orthogonal group-decodable

Symbols within a group are not orthogonal must be jointly detected higher decoding complexity


QO-STBC with Minimum Decoding Complexity

MDC QO-STBC Joint detection of 2 real symbols single (complex) symbol decodable Smallest group size for any QO-STBC

[Yuen, Guan, Tjhung, IEEE Trans WCOM, Sep 2005], [X G Xia et al], [S Rajan et al] .

Can be systematically constructed using smaller O-STBC:

Rule #1: ; Rule #3: ;

1

Rule #2: ; Rule #4: .

q qq K q

q q

q qq K q

q q

jj

q Kj

j

A 0 B 0A A

0 A 0 B

0 A 0 BB B

A 0 B 0


MDC QO-STBC

C4 =

R I R I R I R I1 3 2 4 3 1 4 2

R I R I R I R I2 4 1 3 4 2 3 1

R I R I R I R I3 1 4 2 1 3 2 4R I R I R I R I4 2 3 1 2 4 1 3

c jc c jc c jc c jcc jc c jc c jc c jc

c jc c jc c jc c jcc jc c jc c jc c jc

R I R I R I R I R I R I1 4 2 5 3 6 4 1 5 2 6 3

R I R I R I R I R I R I2 5 1 4 3 6 5 2 4 1 6 3

R I R I R I R I R I R I3 6 1 4 2 5 6 3 4 1 5 2

R I R I R I R I R3 6 2 5 1 4 6 3 5 2

0 00 0

0 00 0

c jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc c jc

c jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc

I R I4 1

R I R I R I R I R I R I4 1 5 2 6 3 1 4 2 5 3 6R I R I R I R I R I R I5 2 4 1 6 3 2 5 1 4 3 6

R I R I R I R I R I R I6 3 4 1 5 2 3 6 1 4 2 5

R I R I R I R6 3 5 2 4 1 3

0 00 0

0 00 0

c jcc jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc c jc

c jc c jc c jc c jc c jc c jcc jc c jc c jc c jc

I R I R I6 2 5 1 4c jc c jc

C8 =

rate = 4/4 = 1

rate = 6/8 = 3/4


MDC QO-STBC:Single Symbol Decodable

Decoding metrics for C4

4 2 22 R I R I R I1 1 , 1 1 1 1 1 1

1 1

4 2 22 R I R I R I2 2 , 2 2 2 2 2 2

1 1

4 2 22 R I R I R I3 3 , 3 3 3 3 3 3

1

( ) 2Re ( ) ( ) ( )

( ) 2Re ( ) ( ) ( )

( ) 2Re ( ) ( )

R

R

N

i kk i

N

i kk i

i ki

f c h c c c c c c

f c h c c c c c c

f c h c c jc jc c c

1

4 2 22 R I R I R I4 4 , 4 4 4 4 4 4

1 1

( )

( ) 2Re ( ) ( ) ( )

R

R

N

k

N

i kk i

f c h c c jc jc c c

( )* * ( ) ( )* * ( )1, 1 2, 2 3, 3 4, 4

k k k kk k k kh r h r h r h r ,

( )* * ( ) ( )* * ( )3, 1 4, 2 1, 3 2, 4


* *1, 3, 2, 4,2Re( )k k k kh h h h ,

( )* * ( ) ( )* * ( )2, 1 1, 2 4, 3 3, 4


( )* * ( ) ( )* * ( )4, 1 3, 2 2, 3 1, 4


* *1, 3, 2, 4,2Re( )k k k kh h h h


MDC QO-STBC:Benefits

vs. Square O-STBC Rate increased by 1/4 smaller QAM size larger Euclidean distance lower BER/SER

vs. other QO-STBC Min. decoding

complexity little/no performance loss


Code Rate Decoding Complexity Origin 3-4 tx ants 5-8 tx ants

O-STBC 3/4 1/2Symbol-by-

symbol detection

AT&T,Nokia,

UU

QO-STBC 1 3/4

Joint detection of 2 complex

symbols

AT&T, Nokia,

Bell Lab

MDC-QOSTB

C1 3/4

Joint detection of 2 real symbols

NTU

MDC QO-STBC:Performance

5 10 15 2010-7

10-6

10-5

10-4

10-3

10-2

10-1

SNR

BE

R

4 tx antennas

8 tx antennas QO-STBCMDC-QOSTBCCIODQO-STBCMDC-QOSTBCCIOD

< 0.5 dB loss from double-symbol-decodable QO-STBC

BER 10e-3

1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

6

10 15 20 25 30SNR (dB)

Spec

tral E

ffici

ency

(bits

/sec

/Hz)

3/4 Rate O-STBC

MDC-QOSTBC

gain over rate-3/4 O-STBC increases with spectral efficiency


Non-coherent tx diversity can be achieved using Differential Space Time Modulation (DSTM) Start with a known seed codeword Xo Transmit

where Up = unitary info-carrying STBC

Receive assume Hp Hp-1

Decode

no need to know Hp

Non-Coherent Tx Diversity

1p p pX X U

1

p p p p

p p p

R H X NR U N

H 1ˆ arg Max Re Trp

p p p p

UU R R U

U


Non-Coherent Tx Diversity

/ 2 / 2H

/ 2 / 2

t t

t t

N N

N NK K

I 0 0 ICC

0 I I 0

2

1

Kii

c

/ 2 R I R I/ 2 / 212K

i i K i K iic c c c

QO-STBC is not unitary cannot be used for DSTM

But MDC-QOSTBC can be rendered quasi-unitary if its symbol constellation meets following conditions:

= 0 = constant


Single-Symbol-Decodable Differential ST Modulation

Differential MDC QO-STBC Single-symbol-decodable DSTM Unitary criterion met by special constellation design

[Yuen, Guan, Tjhung, IEEE Trans WCOM, Oct 2006]

Same code rate and benefits as coherent MDC QO-STBC

QAM/PSK symbols

constellation mapping

MDC QO-STBC

single-symbol-decodableDSTM


Single-Symbol-Decodable Differential ST Modulation

Single-Symbol-Decodable DSTM constellation design for 4 antennas

-1.5 -1 -0.5 0 0.5 1 1.5-1.5

-1

-0.5

0

0.5

1

1.5

0.577

1.29

1

-1.5 -1 -0.5 0 0.5 1 1.5-1.5

-1

-0.5

0

0.5

1

1.5

0.35

1.05

0.78

1.48

2bps/Hz

spectral efficiency = R logM bps/Hz where R = MDC-QOSTBC code rate, M = constellation size

3bps/Hz


Chau YUEN, Yong Liang GUAN, T T Tjhung (I2R)

R I R I R I R I R I R I1 4 2 5 3 6 4 1 5 2 6 3

R I R I R I R I R I R I2 5 1 4 3 6 5 2 4 1 6 3

R I R I R I R I R I R I3 6 1 4 2 5 6 3 4 1 5 2

R I R I R I R I R3 6 2 5 1 4 6 3 5 2

0 00 0

0 00 0

c jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc c jc

c jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc

I R I4 1

R I R I R I R I R I R I4 1 5 2 6 3 1 4 2 5 3 6R I R I R I R I R I R I5 2 4 1 6 3 2 5 1 4 3 6

R I R I R I R I R I R I6 3 4 1 5 2 3 6 1 4 2 5

R I R I R I R6 3 5 2 4 1 3

0 00 0

0 00 0

c jcc jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc c jcc jc c jc c jc c jc c jc c jc

c jc c jc c jc c jc

I R I R I6 2 5 1 4c jc c jc

MDC QO-STBC

hi,2

h1,1

ML

Dec

odin

g

Mod

ulat

i on

Dem

o dul

atio

n

.. bp .. b2 b1 2 1ˆ ˆ ˆ.. .. pb b b

rP(k) … rp(k) … r1(k)

CSI Estimator

DataBit

Stream

ReconstructedBit

Stream

cK .. cp .. c2 c1

ModulatedData

Symbols

STB

C C

( ) ( ) ( )1 .. .. R R R

N N NP pr r r

EstimiatedData

Symbols

2 1ˆ ˆ ˆ ˆ.. .. K pc c c c

,T RN Nh

rP(2) … rp(2) … r1(2)

rP(1) … rp(1) … r1(1)xP(1) … xp(1) … x1(1)

xP(2) … xp(2) … x1(2)

xP(i) … xp(i) … x1(i)

( ) ( ) ( )1 .. .. T T T

N N NP px x x

TransmitAntennas

ReceiveAntennas

c1R , c1I

c2R , c2I

: :

Dif

fere

nti

al

Enco

ding

group decoder

group decoder

Single-Symbol-Decodable DSTM:Decoding Complexitybps/Hz DSTM Scheme Decoding Search Space Dimension1.95 Sp2 Code 225

2 Group Code 256

2 Zhu & JafarkhaniWCNC 2004 16

2 Single-Symbol-Decodable DSTM (proposed) 4

3.13 Sp2 Code 5929

3 Zhu & JafarkhaniWCNC 2004 64

3 Single-Symbol-Decodable DSTM (proposed) 8


Single-Symbol-Decodable DSTM:Performance

0 5 10 15 20 25 3010-6

10-5

10-4

10-3

10-2

10-1

100

Sp(2) 1.95 bps/Hz (decoding search space:225) QO-STBC 2 bps/Hz (decoding search space:16) MDC-QOSTBC 2 bps/Hz (decoding search space:4) Sp(2) 3.13 bps/Hz (decoding search space:5929)QOSTBC 3 bps/Hz (decoding search space:64)MDC-QOSTBC 3 bps/Hz (decoding search space:8)

BLE

R

SNR Y L Guan, PWTC, EEE, NTU, Singapore

References (monograph) C Yuen, Y L Guan, T T Tjhung, Quasi-Orthogonal Space Time Block

Code, Imperial College Press, Dec 2007.

C. Yuen; Y. L. Guan; T. T. Tjhung, “Quasi-Orthogonal STBC with Minimum Decoding

Complexity”, IEEE Trans. Wireless Comms., vol. 4, Sept. 2005.

C. Yuen; Y. L. Guan; T. T. Tjhung, “Single-Symbol-Decodable Differential Space-

Time Modulation Based on QO-STBC”, IEEE Trans Wireless Comm, October 2006.


Outline






EYLGuanHighlight

Group-Decodable STBC with Rate > 1

T. P. RenNational University of Defense Technology, China

Y. L. GuanNanyang Technological University, Singapore

C. YuenInstititute for Infocomm Research, Singapore


2

Motivation• To have spatial multiplexing gain, STBC needs code

rate > 1 (e.g. Golden code, Perfect code)

• But decoding complexity of high-rate STBC is very high, unless the code has group-decodable structure.

• Only two group-decodable high-rate STBC reported:

– 2-group-decodable SMC of rate 1.25 for 4 transmit antennas obtained by computer search [C. Yuen, Y. L. Guan, and T. T. Tjhung, 2006];

– 2-group-decodable square SMC for 2m (integer m > 1) transmit antennas [K. Pavan Srinath and B. Sundar Rajan, 2009];

Not available for any tx antenna number, any code rate.

3

Group Decodable STBC with Rate > 1

block diagonal property of HcHHc

4

Signal Model

5

Code Construction

• Step 1: Pick a full-rank matrix of size 2x2

• Step 2: Write (i) as

where1 0 1 0 0 0 0 00 0 0 0 1 0 -1 0

1 0 -1 0 1 0 1 00 -1 0 1 0 1 0 1

C

0y C

1

1 11 1

C

11 11 21 21 12 12 22 22 TR I R I R I R Iy y y y y y y y y

T. P. Ren, Y. L. Guan, C. Yuen, E. Gunawan and E. Y. Zhang, "Unbalanced and Balanced 2-Group Decodable Spatial Multiplexing Code,“ IEEE VTC-Fall, Sep 2009.

over-determined

6

• Step 3: Obtain the solution space of (i):

• Step 4: Transform to form dispersion matrices of 2nd group:

1 1 2 2 3 3 4 4

1 2 3 4

0 0 0 11 1 1 00 0 0 11 1 1 0 0 0 0 11 1 1 00 0 0 11 1 1 0

y k y k y k y k y

k k k k

1 2 3 4

1 1, , ,

1 1j j j j j j

Y Y Y Yj j j j j j

iy

iy

7

Achievable Code RatesUnbalanced Balanced

T < Nt

T Nt

22 12

t tTN NT

2 12

TT

2 1t tTN NT

2 44

TT

Group-Decodable STBC with Rate > 1

T. P. RenNational University of Defense Technology, China

Y. L. GuanNanyang Technological University, Singapore

C. YuenInstititute for Infocomm Research, Singapore


9

Fast Decoding of STBC

LL

L

L

f

ffff

00

00

222

111

R

r=Hs+z r’=Rs+z’

The code is fast-decodable if

Fast decoding:

MLL

MLssTN

MLTN

MLTN

ssrr

srrr

srrr

MLML

r

r

r

,,decodingjoint ML ',,' )3

decodingjoint ML ',,',' )2

decodingjoint ML ',,',' )1

3 ,

3

232

131

21

where s=[s1 s2 ... sL]T, H=QR is QR decomposition,

r’=[r’1, r’2, … ]T=QTr, z’=QTz.

10

Motivation

• Fast-decodable (FD) STBC was proposed by Biglieri et

al in 2009 to reduce the QRM decoding complexity of

high-rate STBC. But complexity reduction is limited.

• Group-Decodable (GD) STBC has more complexity

reduction, but code rate is low.

Nice to have STBC with both FD and GD advantages

11

Relationship between FD, GD and FGD

12

Outline

• Motivation

• Code Construction

• Performance

• Conclusions

13

Group-Decodable STBC

14

Fast-Group-Decodable STBC

Ref: T. P. Ren, Y. L. Guan, C. Yuen and R. J. Shen, "Fast-Group-Decodable Space-Time Block Code,“ IEEE ITW'10, 6-8 Jan. 2010.

15

Code Construction (max complexity reduction)

16

Code ExampleFGD-STBC of size 4x4 can be constructed based on O-STBC of size 4x4.

Such an O-STBC [Ganesan and Stoica, 2001] has dispersion matrices:

17

Code ExampleIn GD-STBC construction, let C1 be the seed matrix in the first group. There will be 2NtT – NtNt = 2x4x4 - 4x4 = 16 dispersion matrices in the second group, hence the code rate is 17/8. Use C2-C6 to obtain the other 11 dispersion matrices in the second group:

18

Code Example

19

Complexity Comparison

Note: For FGD-STBC proposed, only C1-C16 are applied for a rate-2 code.

20

BER Performance

-6 -4 -2 0 2 4 6 8 10 12

10-5

10-4

10-3

10-2

10-1

SNR per bit/dB

BE

R

BLAST[2], O=24

Perfect code[4], O=216

FD-STBC best known[5], O=4x213

Proposed rate-2 FGD-STBC, O=5x211

21

Remarks

•FD and GD code structure can be integrated to generate FGD-STBC

•A scalable code construction for FGD-STBC with max complexity reduction is presented

•The proposed FGD-STBC achieves•Full transmit diversity gain •Best known rate-complexity trade off

22

References • Tian Peng Ren; Yong Liang Guan; Chau Yuen, Gunawan, E.; Er Yang Zhang;,

"Group-Decodable Space-Time Block Codes with Code Rate > 1", IEEE Transactions

on Communications, accepted, 2010.

• C. Yuen; Y. L. Guan; T. T. Tjhung, “On the Search for High-Rate Quasi-Orthogonal

Space-Time Block Code”, International Journal of Wireless Information Network

(IJWIN), May 2006, Pages 1 – 12. http://dx.doi.org/10.1007/s10776-006-0033-2.

• T. P. Ren, Y. L. Guan, C. Yuen, E. Gunawan and E. Y. Zhang, "Unbalanced and

Balanced 2-Group Decodable Spatial Multiplexing Code,“ IEEE VTC-Fall, Sep 2009.

• T. P. Ren, Y. L. Guan, C. Yuen and R. J. Shen, "Fast-Group-Decodable Space-Time

Block Code,“ IEEE ITW'10, Cairo, Egypt, 6-8 Jan. 2010.


Outline






EYLGuanHighlight

MIMO RELAYING (1)

Async Cooperative Diversity

Relaying

Ren Tian Peng 1, Guan Yong Liang2, Yuen Chau3,

Erry Gunawan2, and Rong Jun Shen1

1 National University of Defense Technology, China 2 Nanyang Technological University, Singapore3 Institute of Infocomm Research, Singapore

Dec 2010

元永亮

2

• Introduction

• Shift-Orthogonal STBC (SO-STBC)

• Delay-Tolerant Cooperative Diversity Routing

• Performance

• Conclusion

Outline

MIMO Relaying

• Benefits– Diversity via single-antenna base stations/relays

(cooperative diversity)

– Range extension

– Relay (router) failure tolerance

Distributed Alamouti code

Distributed Alamouti Coding

h1

h2

c1r1 r2

c2

-c2*

c1*

1 1 2 1 1* *

2 2 1 2 2

r c c hr c c h

1 1 2 1 1* * * *

2 2 1 2 2

r h h cr h h c

Hc is unitary1 1H

*2 2

2 2 * *1 2 1 1 1 2 2

* *2 22 2 1 1 21 2

ˆˆ

0

0

c rc r

h h c h hc h hh h

CH

1. c1 and c2decoupled

2. MRC effect achievedHcHHc = I

R1

R2

Distributed Alamouti Coding

h1

h2

c1r1 r2

c2

-c2*

c1*

1 1 2 1 1* *

2 2 1 2 2

r c c hr c c h

1 1 2 1 1* * * *

2 2 1 2 2

r h h cr h h c

Hc is unitary

What if– Received STBC signals are not synchronized in time?– OFDM (high PAPR) or joint detection (high complexity)

cannot be afforded?

R1

R2

6

Asynchronous MIMO Relaying

STBC matrix X of length L:

Asynchronous version of X received:

Design objectives on X A sub-matrix of Xa (an async version of X) should have full row-rank and be unitary. Rows of X should have aperiodic correlation = 0

7


is a shift-orthogonal STBC

:

AMM can also be formed using zero correlation sequence set.

[T Ren, Y L Guan, E Gunawan, E Y Zhang, IEEE Trans Comm, June 2010]

8


is a shift-orthogonal STBC

:

p

code rate of X = Ls / (M Ls+Lp) 1/M when Ls Lp


9

the shift-orthogonal STBC

Code example for 2 relays

2 relays

differential delay < 2

10

Async received signal:

SO-STBC


Equivalent channel matrix:

to obtain:

Decode SO-STBC by: H


Despite shifts in 2nd and 3rd rows, sub-matrix still orthogonal and rank 3

3

1

2

3

13/4

2

23/2

33/4

1

13/2

33/2

23/4

23/4

13/2

33/2

321

13/4

33/4

2

23/2

13

321

13

2

321

sss

sse

s

sese

s

sesesesesese

sss

sesessess

sss

ssssss

j

j

j

jjj

jjj

jj

j

1

3

2

3

2

1

23/2

13/2

33/2

13/4

33/4

23/4

321

23/4

13/4

33/4

13/2

33/2

23/2

321

213

132

321

:relay3:relay2:relay1

sss

sss

sesesesesese

sss

sesesesesese

sss

sssssssss

jjj

jjj

jjj

jjj

postfix

12

An Application

S

I12

I11

D

I32

I31I21

I22

In a SPR (single-path routing) MANET, a link from a source node to a destination node is typically accomplished by finding the best single path through some relay nodes.

Fig. 1. Single-path routing (SPR) MANET forinter-vehicle communications

However, when a node fails, re-routing is necessary, leading to delays.

-- Async MANET

13

An ApplicationIn the DR (dispersity routing) MANET, several parallel

routing paths are permitted in order to transmit information with redundancy.

Fig. 3. Dispersity routing (DR) MANET

S

I12

I11

D

I32

I31I21

I22

-- Async MANET

However, DR requires additional bandwidth to prevent self-interference among the parallel paths.

[N F Maxemchuk, IEEE MILCOM 2007]

14

An ApplicationIn a cooperative diversity routing (CDR) MANET, data is

transported from relay cluster to relay cluster. Any two adjacent relay clusters form a virtual MIMO system and apply distributed STBC to obtain cooperative diversity.

Fig. 4. Synchronous cooperative diversity routing (S-CDR) MANET

-- Async MANET

1x

2x

1x

2x

However, cooperative diversity works only if there is timing alignment among the routing nodes, this requires complex MAC protocol and may be impossible if node mobility is high.

[M. A. Tope, J. C. McEachen, and A. C. Kinney, ISCC, June 2006]

15

Delay-Tolerant Cooperative Diversity Routing MANETUsing SO-STBC, delay-tolerant cooperative diversity routing (DT-CDR) with simple symbol-wise decoding is possible.

Fig. 5. Delay-tolerant cooperative diversity routing (DT-CDR) MANET

1x

2x

1ax

2ax

[T Ren, Y L Guan, et. al., VTC-Fall, 2010]

16

Simulation

All nodes (source, destination, relays/routers) use only 1 antenna

• DR-MANET (all radio links synchronous) loses diversity when routing node I21 fails

• CDR-MANET (all radio links asynchronous) maintains diversity

Simulation – DF Example

SO-STBC

SO-STBCSO-STBC

[T Ren, Y L Guan, et. al., VTC-Fall, 2010]

In DF (decode and forward) relaying, every relay detects (decode) the symbols and uses only the “cleaned” symbols in the next hop.

In AF (amplify and forward) relaying, the relays use the noisy symbols to form the next-hop signal.

1 2s k s s

1g

2g

1h

2h

1r k

2r k

1x k

2x k

1 2y k y k y k

Asynchronous AF Relaying

Due to its row-wise design, SO-STBC supports async AF relaying + full diversity + linear decoding

Simulation: AF Example

ML decoding BER of SO-STBC and O-STBC for an asynchronous AF

wireless relay network with 3 relays and maximum time difference 2


Comparison: Delay-Tolerant STC

The other delay-tolerant STC’s require trellis decoding

21

• Shift-Orthogonal STBC (SO-STBC) provides distributed, open-loop

spatial diversity for async MANET or async Network MIMO.

• SO-STBC can be systematically constructed based on Fourier Matrix or

Zero Periodic Correlation sequence set.

• Symbol-wise ML decoding of SO-STBC can be done using very simple

Alamouti-like linear decoding.

• SO-STBC supports both async DF relaying and async AF relaying.

• SO-STBC simplifies the routing protocol and reduces the network

latency of async MANET.

Conclusion

Outline






EYLGuanHighlight

MIMO RELAYING (2)

2-Step Bi-Directional Relaying

M. Eslamifar (Nanyang Technological University, Singapore)C. Yuen (Institute for Infocomm Research, Singapore)

W. H. Chin (Toshiba Research Lab, UK)Y. L. Guan (Nanyang Technological University, Singapore)

Dec 2010

2

• 2-Step Bi-Directional Relaying• Uplink • Downlink

– Antenna Selection – Antenna Selection with Binary Network Coding– STBC with Binary Network Coding– Transmit Beamforming– Max-Min AS-BNC

• Performance• Conclusion

Outline

3

Introduction Bi-directional relaying:

Conventional routing:

4 time slots A

A

A

A

B

B

B

B

R

R

R

R1st step

2nd step

3rd step

4th step

sB

sA

sA

sB

node A

Relay

node B

4

Network coding with single-antenna relay: 3 time slots

A

A

A

B

B

B

R

R

R

1st step

2nd step

3rd step sXOR sXOR

sA

sB

BAXOR sss ˆˆ

Network coding with multi-antenna relay: 2 time slots

Multi-Antenna Downlink options:

- Spatial diversity gain?

- Binary/Analog network coding?

- Open/Closed loop (need CSI)?

- Single/Multiple relays?

A

A B

B

R

R

2nd stepsR sR

sA sB1st step

5

Uplink In the 1st step:

The relay performs joint ML decoding:

Diversity order of uplink = N (number of relay antennas)

A BR

6

The best antenna channel to node A is selected to transmit

Likewise, the best antenna channel to node B is selected to tx

e.g.

node A receives:

After canceling off ,

Downlink 1: Antenna Selection

Bŝ

|h||h| and |h||h| B1B2A2A1 if

Aŝ

AABBA

A nshshy

2)ˆ( 21

AA sh ˆ1 ABAAA nshyy ˆˆ 1

R

1Ah

2Bh

BŝAŝ

7

Same as “Antenna Selection” scheme, except that both selected antennas transmit

Node A receives , cancels hA1 sXOR ,

decodes hB2 sXOR , then obtain .

Downlink 2: Antenna Selection with Binary Network Coding

AXORBA

A nshhy

2)( 21

R

1Ah

2Bh

XORsXORs

BAXOR sss ˆˆ

BAXOR sss ˆˆ

8

Downlink 3: Alamouti STBC with Binary Network Coding

Since both nodes A and B are able to extract the other node’s data from sXOR , it may be transmitted using STBC such as Alamouti Code.

Note: 2 time slots are used to transmit 2 symbols rate = 1

RRs Rs

9

Downlink 4: Tx Beamforming Relay transmits:

where (maximal ratio transmission)

Node A receives:

After canceling off , yA

10

Downlink 5: Max-Min AS-BNC 1 out of N relay antennas is selected to transmit

Antenna selection is determined in 2 steps.

Step 1:

Step 2:

E.g. with N = 2 relay antennas: ABR

hA1 hB1

hA2 hB2

11

Moment Generating Function:

Diversity Order:

Average Symbol Error Probability (SEP):

, = instantaneous received SNR ,= moment generating function (MGF) of and .

Max-Min AS-BNC: Analysis

12

0 5 10 15 20 2510-4

10-3

10-2

10-1

100

SNR(dB)

Sym

bol E

rror P

roba

bilit

y(S

EP

)

Simulation Results for AS-BNCSimulation Results for AlamoutiSimulation Results for Max-Min AS-BNCSimulation Results for ASSimulation Results for TB

PerformanceComparing all 5 schemes At high SNR,

• TB best• Max-Min AS- BNC• Alamouti-BNC• AS• AS-BNC worst

• AS-BNC has no spatial diversity gain

13

0 5 10 15 20 2510-4

10-3

10-2

10-1

100

101

SNR (dB)

Sym

bol E

rror P

roba

bilit

y

Simulation Results for TBAnalytical Results for TBSimulation Results for Max-Min-AS-BNCAnalytical Results for Max-Min-AS-BNC

Comparing best 2 schemes (TB and Max-Min AS BNC)

SEP upper bounds verified by simulation

Both have diversity order N

TB outperforms Max-Min AS-BNC

Performance

14

Remarks Downlink CSI (channel state info) knowledge

• TB needs full and accurate CSI at the relay.• Max-Min AS-BNC only needs coarse CSI (which

channel is better) at the relay. No phase info needed.

Max-Min AS-BNC can be implemented across multiple relays using a distributed protocol: • Each relay waits for a delay inversely proportional to its

own downlink channel gain before transmitting• While waiting, listen to other relays. If another relay

transmit first, abort own transmission.

15

Conclusions Multi-antenna relaying with binary/analog network coding

achieves bi-directional communication in 2 time slots(shortest possible) + spatial diversity.

TB has the best performance but requires full downlinkCSI (channel state information) at the relay

Max-Min AS-BNC has comparable performance as TBbut only requires course CSI and can also be extendedto multiple relays with distributed antennas

STBC-BNC comes next in performance and does notrequire CSI, but suffers rate loss or more complexdecoding if the number of relay antennas is more than 2.

16

References • Mahshad Eslamifar, Woon Hau Chin, Chau Yuen, Y L Guan, “Performance

Analysis of Bi-Directional Relaying with Multiple Antennas”, IEEE Trans

Wireless Comms, 2nd review.

• --, “Performance Analysis of Two-Way Multiple-Antenna Relaying with

Network Coding”, VTC-Fall 2009

• --, “Max-Min Antenna Selection for Bi-Directional Multi-Antenna Relaying”,

VTC-Spring 2010



Thank You

mimo coding and relayingewh.ieee.org/r10/xian/com/guan.pdfmimo coding and relaying prof guan, yong...

Documents