TELKOM
NIKA
, Vol.13, No
.3, Septembe
r 2015, pp. 8
51~858
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i3.1807
851
Re
cei
v
ed Ma
rch 2
1
, 2015;
Re
vised Ma
y 29, 2015; Accepted June 1
2
, 2015
Distributed Cooperative Multicell Precoding Based on
Local Channel State Information
Jing An*
1
, G
uofen
g He
2
, You
w
u X
u
1
, Bin Wang
1
, Sen Xu
1
1
Yanche
ng Inst
itute of T
e
chnolog
y,
Yanc
he
n
g
224
05
1, Jian
gsu, Chi
n
a
2
Chin
a T
e
lecom Yanche
ng B
r
anch, Yanc
he
ng 22
40
01, Jia
ngsu, Ch
in
a
*
Corresp
on
din
g
author, emai
l: anji
n
g
9
9
198
2
@
16
3.com
A
b
st
r
a
ct
Coo
perativ
e multicel
l prec
odi
ng is an attra
c
tive
w
a
y of i
m
pr
ovin
g the perfor
m
a
n
ce i
n
mu
lticel
l
dow
nli
n
k scen
a
rios esp
e
ci
all
y
for termina
l
s at cell edg
es
. Multipl
e
bas
e stations in a giv
en are
a
serve ea
c
h
termi
nal
after precod
ing, w
h
ic
h can coor
din
a
t
e the in
ter-cel
l
interferenc
e a
nd ach
i
eve h
i
g
her perfor
m
an
ce.
Most previ
ous
w
o
rk in th
e
are
a
h
a
s fo
cus o
n
ce
ntra
li
z
e
d
pr
ecod
in
g w
h
ich r
e
q
u
i
r
es g
a
theri
n
g
al
l
transmitters
’
c
han
nel state in
formati
on (CSI
) at cent
ral station (CS) throu
gh back
hau
l a
nd then pr
eco
d
in
g
at CS. How
e
ve
r, the requ
ire
m
ents on
back
h
aul s
i
gn
ali
ng
a
nd co
mputati
o
nal
pow
er sca
l
e
s rap
i
dly
in l
a
rge
and
de
nse
net
w
o
rks, w
h
ich u
s
ually
mak
e
su
ch fully
ce
ntrali
z
e
d
appr
oac
he
s i
m
practic
a
l. I
n
this
pa
per, w
e
study tw
o pra
c
tical pr
eco
d
in
g strateg
i
es
w
i
th only
loc
a
l CSI
und
er
a re
lative
ly
realistic
scen
a
r
io.
Performanc
e is
finally il
lustrat
ed throu
gh n
u
m
er
ical si
mulat
i
ons.
Ke
y
w
ords
:
co
oper
ative
mul
i
c
e
ll pr
ecod
in
g, distribute
d
pr
ec
odi
ng, virtual S
I
NR, block di
ag
ona
li
z
a
ti
on
Copy
right
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In recent years, with m
u
ltiple-input mult
ip
le-o
utput (MIMO) techni
que
s, the performan
ce
of cell
ular communi
catio
n
sy
stems can b
e
g
r
eatl
y
improved. Many
alg
o
rit
h
ms have b
een
prop
osed for single
-
cell d
o
wnli
nk
scen
ario, w
here
a ba
se stati
on communi
cate
s with m
any
use
r
s. However, in multi-ce
ll downlin
k scenari
o
,
these
single
-
cell al
gorithm
s are oblige
d
to treat
the interferen
ce from adj
a
c
ent cells a
s
noise,
whi
c
h
result
s in a fundame
n
tal limitation on the
system p
e
rfo
r
man
c
e. Recently
, base st
ation co
ordi
n
a
tion (al
s
o known
as Net
w
ork
MIMO
) has
been a
nalyze
d as a mea
n
s of handling i
n
ter-cell inte
rferen
ce.
Ideally, all base
station
s
might sh
are
t
heir chan
nel state
inform
a
t
ion
(
CSI) an
d
data
throug
h ba
ckhaul links, wh
ich wo
uld en
able co
or
dina
ted pre
c
odi
ng
design that can mana
ge the
co-user inte
rf
eren
ce a
s
in t
he sin
g
le-cell
sce
na
rio [1-3
]. In practice,
there a
r
e limit
ations in te
rm
s
of delay and cap
a
city on the bac
kh
aul
and co
mputat
ional po
wer a
t
the transmitters [4-7], which
make
s it ne
cessary to inv
e
stigate di
stri
buted fo
rm
s
of coop
eratio
n that redu
ce the backh
a
u
l
sign
aling
and
pre
c
o
d
ing
co
mplexity, whil
e still b
enef
iti
ng fro
m
a
ro
b
u
st inte
rferen
ce
co
ntrol [8
-9].
An informatio
n theoretic a
ppro
a
ch was
prop
osed
in [
10] to dete
r
m
i
ne the d
epe
nden
ce
of m
u
lti-
cell rate
s on
backh
aul cap
a
city. A pract
i
cal iterative
messag
e pa
ssing p
r
o
c
ed
u
r
e wa
s ta
ken
in
[11] to excha
nge informati
on between n
e
ighb
orin
g ce
lls.
Herein,
we a
ddre
s
s the
p
r
oble
m
of di
stribute
d
mul
t
icell MIMO
pre
c
odi
ng
where
th
e
coo
perating
base station
s
sh
ar
e kn
o
w
led
ge of th
e data symb
ols but h
a
ve
only local
CSI,
thereby mu
ch
redu
cin
g
the
feedba
ck loa
d
on the
u
p
li
nk a
nd avoidi
ng cell-to-cell
CSI excha
n
g
e
.
In this
pape
r,
we
provide
two p
r
a
c
tical
pre
c
o
d
ing
st
rategie
s
with
only lo
cal
CSI unde
r virt
ual
SINR frame
w
ork.
One
is
b
eamformi
ng v
e
ctors
ac
hiev
ed by G
ene
ralize
d
Raylei
gh Q
uotient, the
other i
s
di
stributed bl
ock diago
nalization alg
o
ri
thm
with MPR
power all
o
ca
tion. Finally, we
provide
simu
lation
re
sults und
er MISO IC sc
e
n
a
r
ios a
nd m
u
lticell
pre
c
o
d
ing
scen
ari
o
s
respe
c
tively and illustrate the perfo
rma
n
c
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 851 – 858
852
2. Sy
stem Model
2.1. Sy
stem
Con
f
igura
t
ion
We con
s
ide
r
a commu
nication scen
ari
o
with
t
K
transmitters
(e.g., bas
e
s
t
ation in a
cellul
a
r syste
m
) whi
c
h equ
ipped
with
t
N
antenna
s
ea
ch
and
rt
KN
single
-
antenn
a recei
v
ers
(e.g.,
mobil
e
station). The transmitters and re
ceivers
are
de
noted
j
B
S
and
k
M
T
, res
p
ectively,
for
1,
,
t
j
K
and
1,
,
r
kK
. The
j
B
S
only kn
ows the
cha
n
n
e
l between it
self an
d all
receivers
whi
c
h a
r
e
within
its ra
nge,
either th
ro
u
gh
use
r
fee
dba
ck o
r
, in a
TDD sy
stem, fo
rm
those
u
s
e
r
s’
tran
smi
ssi
on i
n
the
uplin
k.
And the
r
e
is
no ex
cha
nge
of CSI b
e
twe
en tran
smitters.
We a
s
sume t
hat the data
symbol
s inte
nded fo
r all
receive
r
s
are
available at b
o
th tran
smitters,
whi
c
h en
able
joint multicell
pre
c
odi
ng.
2.2. Chann
e
l Model
In mobile ce
llular
scena
ri
os, the ra
dio
prop
agation
can b
e
ch
a
r
acte
ri
zed by
three
indep
ende
nt phen
omen
a: path loss vari
ation with
di
stance, la
rge
-
scale shad
o
w
ing, an
d sm
all-
scale fadi
ng.
All of them will be
in
co
rpo
r
ated in thi
s
p
aper.
He
re
we assu
med t
hat the
cha
n
nel
betwe
en
j
B
S
and
k
M
T
is na
rrow-band a
nd fre
quen
cy-flat bl
oc
k fading. T
herefo
r
e it is
can
be
modele
d
as
rt
K
K
random m
a
trix.
,,
,
,
,
1
,,
,
1
,,
j
k
jk
jk
jk
t
r
Hc
d
s
W
j
K
k
K
(1)
Whe
r
e,
-
,
jk
cd
denotes the pat
h loss.
,
jk
d
is the distance (in km) bet
wee
n
the
j
B
S
and
k
M
T
;
is the p
a
th loss exp
o
nent, typically
taking
a valu
e between
3.0 and
5.0; an
d
c
is
the median of
the mean pat
h loss at the referen
c
e di
st
ance of 1 km.
-
,
j
k
s
is a log-n
o
rmal distri
bute
d
sha
d
o
w
ing
variable, i.e.,
2
10
,
10
l
o
g
(
)
(
0
,
)
jk
s
h
sN
,
,
j
k
-
,
j
k
W
rep
r
e
s
e
n
ts
the small
-
scale
fadin
g
. The entrie
s
of
,
j
k
W
ar
e
..
.
ii
d
ci
r
c
ula
r
ly
symmetri
c
co
mplex Gau
ssi
an ran
dom va
riable
s
with
zero me
an an
d unit varian
ce.
The rand
om
variable
s
,
jk
d
,
,
j
k
s
an
d matri
c
e
s
,
j
k
W
are a
s
sume
d t
o
be i
ndep
en
dent of
each other a
n
d
indep
ende
n
t
for all
,
j
k
.
2.3. Do
w
n
link Signal Model
Let
t
N
j
x
C
be the
si
gnal t
r
an
smitted by
j
B
S
an
d th
e corre
s
po
ndi
ng
re
ceived
signal
at
k
M
T
be denote
d
by
k
yC
.
,,
1
r
K
j
jk
j
k
k
k
x
pw
s
(2)
Whe
r
e
(0
,
1
)
k
sC
N
is the data symbol
intende
d for
k
M
T
and is a
s
sum
ed to be avail
able at all
transmitters
.
,
j
k
w
is beamfo
rmi
ng vectors which h
a
ve un
it norms (i.e.
,
,
1
jk
w
) and
,
j
k
p
rep
r
e
s
ent
s the power allo
cated for tran
smissi
on to
k
M
T
fo
rm
j
B
S
. Where
j
B
S
i
s
s
u
bjec
t to an
individual ave
r
age p
o
wer
constraint of
j
P
, that is
2
,
1
r
K
j
jk
j
k
Ex
p
P
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Distri
buted
Coope
rative M
u
lticell Pre
c
o
d
ing Based o
n
Local CSI (Jing An
)
853
,
1
t
K
kj
k
j
k
j
yH
x
n
(3)
Whe
r
e
,
j
k
H
is the chan
nel bet
wee
n
j
B
S
and
k
M
T
,
2
(0
,
)
k
nC
N
is white a
dditi
ve noise.
3. Distribu
te
d Precoding
Algorithms
3.1. Virtual SINR Frame
w
ork
Referen
c
e [1
2] note
s
that
the same
ra
te regi
on m
a
y be a
c
hieve
d
in the
upli
n
k (fo
r
a
reci
procal ch
annel, in the
virtual
uplin
k
otherwise) a
n
d
downlin
k di
rectio
ns
usin
g the sam
e
set of
receive a
nd t
r
an
smit be
am
forming
re
sp
ectively, but
with
different power co
nstraint.
This
is o
ne
form of what i
s
refe
rred to as upli
n
k-do
wnlink d
uality.
In its mo
st g
eneral fo
rm,
a virtual SI
NR at
j
B
S
is defi
ned
as the
ratio bet
wee
n
the
useful
sign
al
power received at its serv
ed user
k
M
T
and
the sum
of noise pl
us th
e interferen
ce
power
whi
c
h
cau
s
e
s
at the
remai
n
ing
users
k
M
T
. For ce
rtain choi
ce
s o
f
param
eters,
the virtual
SINR
can
be
see
n
a
s
the
SINR a
c
hi
eved in th
e u
p
link
(o
r virtual
uplin
k) if th
e
same
filters
were
use
d
.
Thus:
2
,,
,
,
2
2
,,
,
jk
jk
j
k
virt
ual
jk
jk
j
k
jk
kk
pH
w
pH
w
(4)
Her
e
,
,
j
k
w
is use
d
to pro
c
ess t
he re
ceived
signal at
j
B
S
from
k
M
T
and
k
M
T
.
3.2. Beamfo
r
m
ing Vector
s Achiev
ed b
y
Generalized Ra
y
l
eigh
Quotien
t
Con
s
id
er the virtual uplin
k cha
nnel, vi
rtu
a
l SINR is achieved at (4).
Where
2
,,
,
jk
jk
jk
pH
w
is the desi
r
e
d
signal p
o
wer tran
smitte
d from
k
M
T
and
2
,,
,
jk
j
k
jk
kk
pH
w
is the interferen
ce
gene
rated fro
m
k
M
T
at
j
B
S
. In general, the goal o
f
beamformi
n
g is to maxim
i
ze the si
gnal
powe
r
at the inte
nd
ed te
rminal
while
minimi
zing
the
i
n
terferen
ce
caused at
othe
r t
e
rmin
als.
Th
ese
ambition
s a
r
e
cou
n
teractin
g and
re
pre
s
ented by m
a
ximum ratio t
r
an
smi
ssi
on
(MRT
) an
d ze
ro-
forc
ing (ZF), res
p
ec
tively. MRT s
t
rategy
foc
u
s
on
ma
ximizing the
useful
sign
al received at o
ne’s
own
re
ceive
r
and the
ge
n
e
rated
interfe
r
en
ce i
s
com
p
letely igno
re
d, while
ZF
strategy m
a
i
n
ly
focu
s on
re
d
u
cin
g
the int
e
rferen
ce
ca
use
d
to
othe
rs.
Rema
rka
b
ly, both the
strate
gie
s
a
r
e
con
s
i
s
tent wit
h
the distrib
u
ted ch
ann
el st
ate informatio
n at transmitt
er (CSIT).
,
,
,
M
RT
j
k
jk
jk
H
w
H
(5)
,
,
,
,
,
jk
jk
H
jk
ZF
jk
H
jk
H
w
H
(6)
Whe
r
e
,
jk
H
is the proje
c
tion m
a
trix onto the null sp
ace of
,
jk
H
.
In [11], the
a
u
thors
stated
that rate tupl
es
on th
e Pa
reto bo
und
ary
of two
u
s
e
r
MISO IC
can
be
achie
v
ed by be
am
forming
vecto
r
s th
at a
r
e lin
ear
co
mbinati
ons of di
strib
u
ted M
R
T
an
d
ZF.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 851 – 858
854
(1
)
()
,
1
,
2
(1
)
MR
T
Z
F
ii
i
i
ii
MR
T
Z
F
ii
i
i
ww
wi
ww
(7)
Whe
r
e
01
i
are o
p
timization
coe
fficients. [14] has
worke
d
o
u
t the optimization co
effici
ent
i
and sho
w
n
to attain the Pareto bou
ndary of MISO interfere
n
ce
cha
nnel
s (IC) with
this
beamfo
rming strategy.
A more gen
erali
z
ed met
hod is the g
eneralized Rayleigh quoti
ent who
s
e solution i
s
actually a line
a
r co
mbin
atio
n of the MRT and ZF vecto
r
s.
Strateg
y
1.
The virtual S
I
NR frame
w
o
r
k
ca
n be
a
pplied to
bal
ance the
sig
nal an
d
interferen
ce
powers. As the o
b
je
ctive is to
have a
distrib
u
ted
al
gorithm
which reli
es only
on
informatio
n local to each b
a
se
station, we propo
se t
hat each tran
smitter solve
a virtual SINR
maximizatio
n
probl
em, whi
c
h can be
sta
t
ed as follo
ws [12]:
,
,
,
,
2
,,
,
,
2
1
2
,,
,
2
,,
2
2
1
,
,
,
,,
,
,
2
1
,,
,
,
,,
,
,,
,
,
2
1
,
arg
m
ax
arg
m
ax
arg
m
ax
arg
m
ax
jk
jk
jk
jk
jk
jk
j
k
jk
w
jk
jk
jk
kk
jk
j
k
w
jk
jk
kk
jk
HH
j
k
jk
jk
j
k
w
HH
H
j
kj
k
j
k
j
k
jk
j
k
kk
jk
HH
jk
j
k
jk
jk
w
H
jk
pH
w
w
pH
w
Hw
Hw
p
wH
H
w
ww
w
H
H
w
p
wH
H
w
w
p
,
,
,,
,
,,
1
,,
arg
m
ax
jk
H
jk
jk
j
k
kk
jk
H
jk
jk
H
w
jk
j
k
HH
w
wA
w
wB
w
(8)
Whe
r
e
,,
H
jk
j
k
A
HH
and
2
,,
,
H
j
kj
k
kk
jk
BH
H
p
.
Expressio
n
(8) is the Ge
nerali
z
e
d
Ra
yleigh
Quotie
nt if matrix
A and B are Hermitian
matrix, and B is positive de
finite. We can
find that:
,,
,,
()
HH
jk
jk
jk
jk
HH
HH
HH
A
A
(9)
2
,,
,
2
,,
,
2
,,
,
H
HH
jk
j
k
kk
jk
H
H
H
jk
jk
kk
jk
H
jk
jk
kk
jk
BH
H
p
HH
p
HH
p
B
(10
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Distri
buted
Coope
rative M
u
lticell Pre
c
o
d
ing Based o
n
Local CSI (Jing An
)
855
2
,,
,
()
(
)
H
jk
j
j
k
kk
k
HH
BI
p
(
1
1
)
And
,,
H
jk
j
k
HH
is non
-n
egative defin
e matrix,
,,
0
H
jk
jk
HH
,
then:
,,
()
0
H
jk
jk
kk
HH
(12)
At the s
a
me time,
2
,
0
jk
p
. So,
()
0
B
(
1
3
)
No
w we
can
dra
w
a co
ncl
u
sio
n
that expre
ssi
on (6)
satisfie
s the
con
d
ition
s
mentione
d
above a
nd i
s
Gen
e
rali
ze
d
Rayleig
h
Qu
otient. It can
be
solved
by eigen
value
techni
que
s. T
h
e
widely known s
o
lution to this
problem is
s
u
c
h
that:
,,
,
j
kj
k
j
k
A
wB
w
(
1
4
)
Accordi
ng to
this
specific solution, the expr
essio
n
(7)
is maximize
d
when
the column of
,
j
k
w
are the domi
nating eig
env
ectors of
1
B
A
corresp
ondi
ng to the highe
st ei
gen value
s
.
Strateg
y
2.
Since th
e si
gn
als transmitted from diffe
re
nt tran
smitters expe
rien
ce
different
macro
s
copi
c fading, efficie
n
t powe
r
allo
cation ov
e
r
the transmitters will enhan
ce
signifi
cantly the
SNR at the receive
r
s a
nd
increa
se
the
cap
a
city or th
e diversity ga
in of the coop
erative multicell
system. Thi
s
pape
r we int
r
odu
ce the ma
ximal path loss
ratio (MP
R
) app
roa
c
h, which follo
ws the
intuition of allocatin
g
more
power to st
ro
ng term
in
als,
sin
c
e wea
k
te
rminal
s ho
pef
ully are serve
d
more effe
ctively by other base
station
s
.
,,
,
,,
1
r
H
jk
jk
j
kj
K
jk
j
k
k
tr
H
H
pP
tr
H
H
(
1
5
)
Whe
r
e
j
P
is the maximal tran
smitting po
we
r at
j
B
S
.
3.3. Distribu
ted Block
Dia
gonaliza
t
ion
Algorithm
w
i
th MPR Po
w
e
r Alloca
tion
Define
,1
,
1
,
,
,1
jk
jk
j
K
r
H
HH
H
H
jk
j
HH
H
H
H
. We
can elimi
nate all
multi-u
s
er
interferen
ce t
h
rou
gh fo
rci
n
g
,
j
k
w
to lie in the null
space
of
,
j
k
H
. Data
c
a
n be trans
m
itt
ed to
k
M
T
if the null
spa
c
e
of
,
j
k
H
has
a dim
e
n
s
ion g
r
eate
r
t
han 0.
Thi
s
is
satisfie
d
wh
e
n
,
()
jk
t
r
ank
H
N
. Let
,,
()
j
kj
k
r
r
ank
H
, and def
ine the sin
gul
ar value de
co
mpositio
n (S
VD).
10
,,
,,
,
H
jk
jk
jk
jk
jk
HU
V
V
(
1
6
)
Whe
r
e
0
,
j
k
V
hold
s
the la
st
,
()
tj
k
Nr
rig
h
t sin
gula
r
v
e
ctors, a
nd
1
,
j
k
V
holds
the first
,
j
k
r
right
sin
gula
r
vecto
r
s.
0
,
j
k
V
forms a
n
o
r
tho
g
onal
ba
sis fo
r the
null
sp
a
c
e
of
,
j
k
H
, and it
s
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 851 – 858
856
colum
n
s a
r
e
candi
date
s
for the bamfo
rming ve
ctors
,
j
k
w
. Assumi
n
g
that the indepe
nden
ce
con
d
ition is satisfied fo
r all mobile
stations,
we define th
e matrix
(0
)
,
,
j
k
sj
k
HH
V
an
d
,
()
j
ks
rr
a
n
k
H
, now the sy
stem ca
pa
city unde
r the zero
-inte
r
fere
nce con
s
traint can be written
as:
2
,
2,
,
1
ma
x
l
o
g
jk
H
H
s
jk
j
k
s
w
n
CI
H
w
w
H
(
1
7
)
The problem
is now to fi
nd a matrix
,
j
k
w
that maximize
s the dete
r
mina
nt, and
the
solution is to let
,
j
k
w
be the
ri
ght sin
gula
r
vectors of
s
H
, weig
hted by
MPR po
we
r
allocation
approa
ch
on
the corre
s
po
nding
si
ngul
a
r
valu
es.
He
re we
ch
oo
se
MPR rath
er
than
wate
r-fill
ing
becau
se it’s
more reali
s
tic. Define the SVD:
(0
)
10
,
,
,,
,
,
0
00
H
jk
jk
jk
jk
jk
jk
HV
U
V
V
(
1
8
)
Whe
r
e
,
j
k
is
,,
j
kj
k
rr
and
1
,
jk
V
repres
ents
the firs
t
,
j
k
r
si
ngul
ar vectors. T
h
e
n
0
1
,
,,
,
jk
j
kj
k
j
k
wV
V
p
is the be
amfo
rming ve
ctors that can m
a
ximize the i
n
fo
rmation
rate
subje
c
t
to produc
i
ng z
e
ro interferenc
e
.
4. Simulation Resul
t
s
4.1. MISO IC Scenarios
We
con
s
id
er the MISO in
terfere
n
ce ch
annel
whe
r
e
2
t
K
transmitters
with
2
t
N
antenn
as ea
ch and
2
r
K
singl
e antenn
a re
cei
v
ers. Each transmitte
r, wh
ich ha
s only the data
informatio
n i
n
tende
d for its own re
ceivers
a
n
d
the CSI b
e
twee
n itself
and all
users,
comm
uni
cate
s with a si
ngl
e receiver [7, 8].
Figure 1. Illustrates the ava
ilabl
e chann
el
capa
city of the different b
eamformi
ng strategie
s
whi
c
h
are M
R
T, ZF, Zakho
u
r p
r
o
posed ap
pro
a
c
h,
Raylei
gh
quotient an
d distrib
u
ted B
D
0
2
4
6
8
10
12
14
16
18
20
0.
5
1
1.
5
2
2.
5
3
3.
5
4
4.
5
5
5.
5
N
o
m
a
liz
e
d
s
y
m
b
o
l S
N
R
c
*
P
/
D
a
(d
B
)
C
a
p
a
c
i
ty
(
b
i
t
s
/
H
z
/s
e
c
)
MR
T
ZF
Z
a
k
hour
P
r
op
os
ed
R
a
y
l
ei
gh quo
t
i
ent
D
i
s
t
r
i
but
ed B
D
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Distri
buted
Coope
rative M
u
lticell Pre
c
o
d
ing Based o
n
Local CSI (Jing An
)
857
From
the fig
u
r
es,
we
can
see that i
n
thi
s
scen
ario
di
st
ributed
BD e
quivalent to
Z
F
, and
Rayleig
h
q
uot
ient a
c
qui
re
s
identical a
nd
best
pe
rform
ance
with Z
a
kho
u
r Prop
osed a
pproa
ch.
So
we can d
r
a
w
the con
c
lu
si
on that Rayle
i
gh quotie
nt approa
ch ca
n get the opti
m
al perfo
rma
n
ce
Since ha
s p
r
oved that Zakh
our p
r
op
ose
d
app
ro
a
c
h arrived the rate tupl
e on the Paret
o
boun
dary.
4.2. Multicell Precoding Scenarios
In this
se
ctio
n, we ill
ustrate the
p
r
e
c
odi
ng pe
rforman
c
e in
a
scena
rio
with
2
t
K
base
station with
2
t
N
antenn
as ea
ch an
d
2
r
K
sin
g
le
anten
na m
o
bile terminal
s. Each
ba
se
station
kno
w
s the data i
n
fo
rmation t
r
an
smitted fo
r all
mobile te
rmin
als a
nd
ha
s
CSI that can
be
obtaine
d loca
lly.
The availabl
e cap
a
city o
f
Rayleigh q
uot
ient and
Distri
buted B
D
are given.
As a
comp
ari
s
o
n
, we give the capa
city of MRT,
ZF and Centralized BD. From Figu
re 2 we can se
e
that the perfo
rman
ce of Di
stribute
d
BD and Ra
ylei
gh
quotient are
simila
r and b
e
tter than MRT
and ZF. Interestingly, the
perfo
rman
ce
loss when
compa
r
ed
with Ce
ntrali
zed
BD whi
c
h h
a
s
global
CSI is only no more
than 2 bits/Hz/se
c. At
the same time, we redu
ce the
comp
utationa
l
deman
ds
sin
c
e th
ere i
s
n
o
exchan
ge
o
f
CSI bet
wee
n
ba
se
statio
ns. So
they a
r
e m
o
re p
r
a
c
tica
l
scheme
s
.
Figure 2. Illustrate the pre
c
oding p
e
rfo
r
mance in a scen
ario
with
2
t
N
base station
wi
th
2
t
N
antenn
as e
a
ch and
2
r
K
single
antenn
a mobi
le terminal
s
5. Conclusio
n
In this paper,
we have ad
dre
s
sed the probl
em
of distribute
d
multicell MIMO preco
d
in
g
whe
r
e the co
operative base statio
n
s
do not sha
r
e kno
w
led
ge of
the data symbol
s but have on
ly
local
CSI. Under virtual SINR fra
m
e
w
ork, we p
r
ovid
ed two practi
cal precodin
g
strategie
s
wi
th
only local CS
I. One can b
e
obtained b
y
generali
z
ed
Rayleigh qu
otient, and the other wa
s
an
appli
c
ation o
f
block dia
g
onali
z
ation with MPR
power all
o
catio
n
. The distri
buted preco
d
ing
algorith
m
s
re
duced the fe
edba
ck load
on the upl
i
n
k and avoid
e
d
cell
-to-cell
CSI excha
n
g
e
.
Simulation
result
s
sho
w
that the
propo
sed
tw
o
distrib
u
ted
al
gorithm
s ca
n
achieve si
milar
available
rate
perfo
rman
ce
whi
c
h i
s
mu
ch b
e
tter tha
n
MRT
and
Z
F
. Although t
here
is
a limited
perfo
rman
ce
l
o
ss
comp
are
d
with
centrali
zed
al
g
o
rithm
s
, the
propo
sed two di
strib
u
ted al
gorith
m
s
are mo
re p
r
a
c
tical
scheme
s
sin
c
e the
r
e
is no exchan
ge of CSI betwee
n
ba
se st
ations.
Ackn
o
w
l
e
dg
ements
This
wo
rk
wa
s supp
ort
ed by the f
i
nan
cial
sup
port of
Nati
onal
Natural
Scien
c
e
Found
ation o
f
China (G
ra
nt No.6110
50
57) an
d Ji
an
gsu Provincia
l
Natural Sci
e
nce Fo
und
ation
unde
r (G
rant
No.13KJB
5
2002
4, BK201312
21) a
n
d
“Six Talent Peak” Prog
ram of Jia
n
g
su
Province (Chi
na) un
de
r Grant No.ZBZZ
-
043.
0
5
10
15
20
25
30
35
0
5
10
15
20
25
N
o
m
a
l
i
z
ed s
y
m
bol
S
N
R
c
*
P
/
D
a
(d
B
)
C
a
p
a
c
i
t
y(
b
i
ts/
H
z/
se
c)
MR
T
ZF
R
a
y
l
ei
gh quot
i
ent
D
i
s
t
ri
but
e
d
-B
D
C
ent
r
a
l
i
z
ed B
D
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 851 – 858
858
Referen
ces
[1]
M Karaka
ya
li,
GJ F
o
schini, RA Valen
z
uel
a. Net
w
or
k coordi
natio
n for spectrall
y
effici
en
t
communic
a
tio
n
s
in cell
ular s
y
s
t
ems.
IEEE Wireless
Com
m
unications.
200
6; 13(4): 56-
61.
[2]
H Z
han
g, H
D
a
i. Coc
h
a
n
n
e
li
nterferenc
e mit
i
gat
i
on
an
d co
oper
ative
proc
essin
g
i
n
d
o
w
n
link
multic
el
l
multius
e
r MIMO net
w
o
rks.
EURASIP J. Wir
e
less Cmm
u
n. Network
. 2001;
2: 222-23
5.
[3]
F
Boccard
i, H
Hu
ang.
Opti
mum p
o
w
e
r al
lo
cation
for th
e
MIMO-BC
z
e
r
o
-forcin
g
pr
ec
oder
w
i
th p
e
r-
anten
na p
o
w
e
r constraints
. Proc. Conf. Inf. Scienc
es S
y
ste
m
s (CISS). 2006; 3: 504.
[4]
A W
i
esel, Y E
l
dar, S S
hama
i
. Z
e
ro-forcin
g
preco
d
i
ng
an
d ge
ner
aliz
ed
invers
es.
IEEE Trans. Sig.
Processi
ng.
20
08; 55(9): 4
409
-441
8.
[5]
H
Hu
h,
H Pa
pad
op
oul
os, G
Caire.
MIM
O
broa
dcast
chan
nel
o
p
ti
mi
z
a
ti
on
un
der
gen
eral
li
ne
ar
constrai
nts
. Proc. IEEE Int. S
y
mp. Inf.
T
heor
y
(ISIT
)
. 2009; 7: 2664-
26
68.
[6]
S Pramonn
o, E
T
r
i
y
o
no. Performanc
e of
ch
ann
el estim
a
tio
n
in MIMO-OFDM s
y
stems.
TEL
K
OMNIKA
T
e
leco
mmunic
a
ton, Co
mp
utin
g, Electronics
and C
ontrol
. 2
013; 11(
2): 355
-362.
[7]
H Setia
w
a
n
, Y
Nag
ao, M Ku
rosaki, H Och
i
. IEEE 802.11
n Ph
y
s
ic
al L
a
y
er Implem
entat
ion
on Fie
l
d
Programm
abl
e
Gate Arra
y.
TE
L
K
O
M
N
I
K
A
T
e
leco
mmunic
a
ton, C
o
mputi
ng,
El
ectronics
and
Co
ntrol
.
201
2; 10(1): 67
-74.
[8]
O Simeone,
O Somekh,
H Poor
, S
Sh
a
m
ai. Do
w
n
l
i
nk
multice
ll
proc
essin
g
w
i
t
h
l
i
m
ited-b
a
ckha
u
l
capac
it
y
.
EUR
ASIP Journal
o
n
Advanc
es in
Sign
al Process
i
ng
. 20
09.
[9]
M Koba
ya
sh
i, M Deb
b
a
h
, J B
e
lfior
e
.
Outage
efficient strate
gies
netw
o
rk M
I
MO w
i
th partia
l
CSIT
.
Proc.
IEEE ISIT. 2009.
[10]
BL Ng, J Ev
an
s, S Hanl
y,
D
Aktas. Distribut
ed
d
o
w
n
li
nk b
e
a
mformin
g
w
i
t
h
coo
per
ative
base stati
ons.
IEEE Trans. Inf. Theory.
2008;
54(12): 54
91-
549
9.
[11]
O Simeon
e, O Somekh, HV P
oor, S Sham
ai.
Distribut
ed MI
MO in multi-ce
ll wireless system
s via finit
e
-
capac
ity links.
Proc. IEEE IS
CCSP. 2008.
[12]
Ran
da Z
a
kh
ou
r, David Ges
b
e
r
t.
Coordi
natio
n on th
e MISO interfere
n
ce ch
ann
el us
in
g the
virtual SI
N
R
framew
ork
. Internati
ona
l IT
G
W
o
rkshop o
n
Smart Antenn
a
s
. Berlin, German
y
. 200
9; 2: 1-5.
[13]
EA Jors
w
i
eck,
EG Larsson.
T
he MISO interferenc
e cha
nne
l from
a g
a
me-the
oretic
persp
ective: A
combi
natio
n of
selfishn
ess a
n
d
altruis
m
achi
eves Par
e
to o
p
timality
. IEEE International
Coference o
n
Acoustics, Spe
e
ch an
d Sig
nal
Processin
g
. 2008.
[14]
Xi
ao-H
u
Y
ou,
Don
g
-Min
g
W
ang. C
o
oper
ativ
e
Dis
tributed
Ante
nna
S
y
stems
for Mob
ile
Commun
i
cati
o
n
s [Coor
din
a
te
d an
d Distri
but
ed MIMO].
IEEE W
i
reless C
o
mmu
n
icati
ons.
201
0; 17(
3):
35-4
3
.
[15]
EA Jors
w
i
eck,
EG Larsson.
T
he MISO inter-ference ch
ann
el from
a
ga
me-th
eoretic
perspectiv
e
:
Acombi
natio
n of
selfish
ness and altr
uis
m
a
c
hiev
es Pareto
opti
m
a
lity.
In Proc. Int’l Conf. Acoustics,
Speec
h an
d Si
g. Proc. (ICASSP). Las Vega
s. 2008.
[16]
EA Jors
w
i
eck, EG Larsson. C
o
mpl
e
te char
a
c
ter-iza
tio
n
of the Par
e
to b
o
u
ndar
y for th
e
MISO interfer-
ence ch
an
nel.
IEEE Transactions on
Signal Processing
. 20
08; 56(1
0
): 529
2–5
29
6.
Evaluation Warning : The document was created with Spire.PDF for Python.