[PDF] - Optimal Placemen t of Base Stations in Wireless Indo or T PDF Document

SLIDE 1 Optimal Placemen t

f

Base Stations in Wireless Indo

r

T elecomm unication

Thom

F r

uh

wirth LudwigMaximiliansUniv ersit y Munic h frueh wirinformatikunim uenc hende h ttpwwwinformatikunim uenc hendefrueh wir P ascal Brisset Ecole Nationale de lAviation Civile P ascalBrissetrec herc heenacfr Octob er

W
rk

w as done while at ECR C Munic h German y

SLIDE 2 Remarks ABSTRA CT Planning

f

lo cal wireless comm unication net w

rks

is ab

ut

installing base stations small radio transmitters to pro vide wireless devices with strong enough signals POPULAR is an adv anced industrial protot yp e that allo ws to compute the minimal n um b er

f

base stations and their lo cation giv en a blueprin t

f

the installation site and information ab

ut

the materials used for w alls and ceilings It do es so b y sim ulating the propagation

f

radiow a v es using ra y tracing and b y subsequen t

ptimization
f

the n um b er

f

base stations needed to co v er the whole building T aking adv an tage

f

stateoftheart tec hniques for programmable applicationorien ted constrain t solving POPULAR is among the rst practical to

ls

that can

ptimally

plan wireless comm unication net w

rks

SLIDE 3 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Wireless T elecomm unication

Reac

hable ev erywhere

Flexible
no

cabling

Impro

v ed qualit y and securit y ECR C Siemens ICL Bull Siemens Priv ate Net w

rks

Siemens Researc h and Dev elopmen t R WTH Aac hen Comm unication Net w

rks

Dept

Indo
r

digital telecomm unication

Need

sender stations

Ha

v e to plan their lo cation

SLIDE 4 Remarks Pro ject leader Aac hen Prof Dr W alk e Hr Humann

Mill

Email addresses

Mill

F ax mac hines

Mill

Mobile phones

W

estern Europ e

Mill

wireless phones Ger man y

Mill

Flexible

no

cabling required Impro v ed transmis sion qualit y and securit y

Propagation
f

radio w a v es inside buildings T

da

y

the

n um b er and p

sitioning
f

base stations is estimated b y an exp erienced sales p erson W as Op en Problem

W
rldwide

rst together with A TT WISE

IEEE

Exp ert Magazine USA Best Application in T elecomm unications

Winner

T elecom Application Con test

f

T elecom Italia at CP Conference

SLIDE 5 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C The Challenge Optimal placemen t

f

senders

Complete

co v erage

f

site

Minimal

n um b er

f

senders

BS BS BS PABX PSTN

SLIDE 6 Remarks Propagation

f

radio w a v es inside buildings Curren t systems are cellular in that a base station sender transmitter con trols the links to the trancei v ers A radio cell is the space that is co v ered b y a sin gle base station F

r

buildings m ulticellular systems are required b ecause w alls and

rs

absorb part

f

the radio signal

SLIDE 7 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Radio W a v e Propagation Radio signal suers from

atten

uation w eak ening due to distance

shado

wing absorption through

bstacles
m

ultipath propagation due to reection P ath loss w alls at m and m log scale

30 40 50 60 70 80 90 100 110 1 2 3 4 5 6 7 8 9 10 15 20 25 30 path loss / dB distance / m 38 dB 107 dB

SLIDE 8 Remarks The mo del is based

n

the p

w

er balance

f

wireless transmission It com bines a distance dep enden t term with correction factors for extra path loss due to

rs

and w alls

f

the building in the propagation path T

tak

e reection and m ultipath eects in to ac coun t a fading r eserve fade margin is in tro duced W e also extended the mo del to tak e the directional ef fect

f

an an tenna in to accoun t since an tennas do not b eam with the same energy in ev ery direction

SLIDE 9 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C The Constrain t Approac h

Sim

ulation

Mo

delling

f

site W alls Materials

Propagation
f

radio w a v es ra y tracing

Optimization
Constrain

t solving in tersection

f

radio cells

Searc

h try equating senders

Optimization

branc h and b

und

SLIDE 10 No Remarks

SLIDE 11 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Sim ulation

f

Radio W a v e Propagation Ra y T racing

Sender (xs,ys,zs)

Floor2 Floor1

Pfad (xf,yf,zf) (xw,yw,zw) Wall Wall

SLIDE 12 Remarks Ra y tracing sim ulates the propagation

f

radio w a v es through the w alls and ceilings

f

the building T

get

to the p

in

t

f

minimal sensitivit y ie maxi mal p ermissible path loss eac h path m ust b e follo w ed through the whole building The v alues

f

an tenna atten uation in the direction

f

the path the path loss due to the distance and the insertion losses due to in tersections

f

the path with w alls and

rs

are added up to the maximal p ermis sible path loss

SLIDE 13 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C T est P

in

t Grid Co v ers the site

step

step/x

H1 H1 1,7m test point H2 H2

step/x

SLIDE 14 Remarks In the sim ulation phase the c haracteristics

f

the building are computed using

f

test p

in

ts Eac h test p

in

t represen ts a p

ssible

receiv er p

sition

The test p

in

ts are placed

n

a dimensional grid inside the v

lume

that should b e co v ered F

r

eac h test p

in

t the space where a base station can b e put to co v er the test p

in

t the radio cell is calculated The end p

in

ts resulting from ra y tracing are used to describ e the h ull

f

the radio cell If the test grid is sucien tly small sev eral p er squaremeter w e can exp ect that if t w

neigh

b

uring

test p

in

ts are co v ered the space in b et w een

hence

the whole building

can

also b e co v ered

SLIDE 15 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Radio Cells Mo dern Monks Monastery

SLIDE 16 Remarks Note that the radio cell will usually b e a rather

ddshap

ed

b

ject since the receiv ed p

w

er ma y exhi bit discon tin uities b ecause

f

tin y c hanges in the lo ca tion

suc

h as a mo v e around the corner

SLIDE 17 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Constrain t Solv er In tersection

f

radio cells

Sender

in rectangle notempty

S

in ABCD

AC

BD intersection

S

in ABCD S in ABCD

A
maxAA

B

maxBB

C

minCC

D

minDD

S in ABCD

SLIDE 18 Remarks F

r

eac h

f

the resulting radio cells a constrain t is set up that there m ust b e a lo cation

f

a base station geometrically sp eaking a p

in

t somewhere in that space Then w e try to nd lo cations that are in as man y cells at the same time as p

ssible

This means that a base station at

ne
f

these lo cations will co v er sev eral test p

in

ts at

nce

Th us the p

ssible

lo cations are constrained to b e in the in tersections

f

the cells co v ered In a rst attempt restricted to t w

dimensions

w e appro ximated a cell b y a single rectangle The D co

rdinates

are

f

the form XY rectangles are

r

thogonal to the co

rdinate

system and are represen ted b y a pair comp

sed
f

their left lo w er and righ t up p er corner co

rdinates

F

r

eac h cell simply a cons train t insideSender Rectangle is imp

sed

whe re Sender refers to a p

in

t that m ust b e inside the Rectangle The rst rule named not empty sa ys that the constrain t S in ABCD is

nly

v alid if also the condition ACBD is fullled so that the rectangle has a nonempt y area The intersect rule sa ys that if a base station lo cation S is constrained b y t w

inside

constrain ts to b e in t w

rectangles

at

nce

w e can replace these t w

constrain

ts b y a single inside cons train t whose rectangle is computed as the in tersection

f

the t w

initial

rectangles

SLIDE 19

Extend

to union

f

geometric

bjects

empty

S

in nil

false

intersection

S

in L S in L

intersectallL

L L S in L intersectallL L L

setofRchooseR

L chooseR L intersectoneRRR L chooseXconsYL XY

chooseXL
Prolog

Search labeling

try

equating senders equatesendersnil true equatesendersconsSL chooseSL

true

equatesendersL

Branch

and Bound

ptimization

in Prolog bbCML constraintsCS maxsenderSM equatesendersS numbersendersSN bbCNL

LS

SLIDE 20 Remarks It to

k

just

min

utes to extend this solv er so that it w

rks

with union

f

rectangles that can describ e the cell more accurately

actually

to an y desired degree

f

precision The union corresp

nds

to a disjunctiv e cons train t

f

the form insideSR

r

insideSR

r
r

insideSRn whic h is more compactly im plemen ted as insideSRR Rn

The

subse quen t lifting to

dimensions

just amoun ted to adding a third co

rdinate

and co de analogous to the

ne

for the

ther

dimensions T

compute

a solution after w e ha v e set up all the in constrain ts w e try to equate as man y base sta tions as p

ssible

Equating base stations causes the intersect rule to re with the constrain ts asso ciated with the base stations As a result

f

this lab eling pro cedure a base sta tions lo cation will b e constrained more and more and th us the intersect rule will b e applied again and again un til the rectangle b ecomes v ery small and nally em pt y

Then

the not empty rule applies causes failure and so initiates c hronological bac ktrac king that will lead to another c hoice A rst solution is computed in this w a y

Next

to minimize the n um b er

f

base stations w e use a br anch andb

und

metho d It consists in rep eatedly searc hing for a solution with a smaller n um b er

f

base stations un til the minimal n um b er is found

SLIDE 21 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Complete Co v erage

f

Monastery

SLIDE 22 Remarks The result

f

co v ering a mediev al monastery is sho wn where four base stations are needed If more than

ne

base station co v ers a region it is attributed to the base station that pro vides the strongest signal

SLIDE 23 Thom F r

uh

wirth

LMU

P ascal Brisset

ENA

C Conclusions Constrain t programming with CHR pro v ed to b e

rapid

protot yp e a few manmon ths eort

expressiv

e ev erything in CLP

including

ra y tracing and a graphical user in terface

exible

from single rectangles to union

f

rectangles from D to D within min utes

extensible

restricting senders to b e

n

w alls

nly
just

another constrain t

n

eac h sender

ecien

t for a t ypical

ce

building

ptimal

placemen t is found within a few min utes

SLIDE 24 Remarks F

r

a t ypical

ce

building an

ptimal

placemen t is found b y POPULAR within a few min utes This is impressiv e since ev erything including ra y tracing and a graphical user in terface w as implemen ted in a CLP language The CLP co de is just ab

ut
lines

with more than half

f

it for graphics and user in ter face The

v

erall qualit y

f

the placemen ts pro duced is comparable to that

f

a h uman exp ert The preci sion is inuenced b y the underlying path loss mo del with its the fading reserv e the n um b er

f

ra ys used in the sim ulation and the appro ximation

f

radio cells b y unions

f

rectangles While w e w

rk

ed

n

POPULAR without kno wing from eac h

ther

WiSE uses an adaptation

f

the Nelder Mead direct searc h metho d that

ptimizes

the p ercen tage

f

the building co v ered Ab

ut
lines
f

C WiSE has b een paten ted and is in commercial use b y Lucen t T ec hnologies since

to

plan their DEFINI TY Wireless Business System

PWT

SLIDE 25 Constrain t Programming with Constrain t Handling Rules Optimal Placemen t

f

Senders for digital wireless T elecomm unication IEEE Exp ert Magazine USA Best Application in T elecomm unications

Winner

T elecom Application Con test

f

T elecom Italia at CP Conference

SLIDE 26 Remarks The result

f

co v ering a mediev al monastery is sho wn where four base stations are needed If more than

ne

base station co v ers a region it is attributed to the base station that pro vides the strongest signal