[PPT] - Optimal Planning of Digital Cor d less T ele c ommunic PowerPoint Presentation

SLIDE 1 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 1 Optimal Planning

f

Digital Cordless T elecomm unication Systems a T. Fr

uhwir

th Ludwig-Maximilians-Univ ersit at M

unc

hen, German y P . Brisset

Ecole

Nationale de l'Aviation Civile, F rance a W

rk

w as done at ecr c, Munic h, German y p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 2 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 2

BS BS BS PABX PSTN

Using a Mobile Phone in a Building p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 3 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 3 The popular protot yp e Data :

A

blue-prin t

f

the building

Information

ab

ut

the materials used for w alls and ceilings The problem :

Placing

senders to co v er all the ro

ms

in the building

Computing

the minim um n um b er

f

senders needed The solution :

Using

constrain t tec hnology p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 4 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 4 Propagation mo del: loss/distance

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 92 dB

p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 5 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 5 Propagation mo del (con t.) L = L 1m + 10n log 10 d + X i k i F i + X j p j W j L T

tal

path loss in dB L 1m path loss in 1m distance from the sender n propagation factor d distance b et w een transmitter and receiv er k i n um b er

f
rs
f

kind i in the propagation path F i atten uation factor

f
ne
r
f

kind i p j n um b er

f

w alls

f

kind j in the propagation path W j atten uation factor

f
ne

w all

f

kind j p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 6 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 6 Direct Enco ding A naiv e solution w

uld

b e to

Discretize

the space in grid p

in

ts P i

Express

the relation (constrain t) b et w een senders S j p

sitions

and signal lev el at eac h p

in

t P i : S ig nal (P i ) = max j (S ig nal (S j )

Loss(S

j ; P i ))

Express

that the signal m ust b e ab

v

e a threshold at eac h p

in

t : S ig nal (P i )

T

hr eshol d It do es not w

rk

b ecause the relations are to

complex

to constrain senders p

sitions.

p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 7 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 7 Dual Problem Since the propagation

f

a signal is not directional, sender and receiv er can b e exc hanged. Therefore the t w

follo

wing prop erties are equiv ale n t : Eac h grid p

in

t is reac hed b y the signal

f
ne

sender : 8P i 9S j P i 2 C

v

er ed(S j ) There is a sender in the neigh b

urho
d
f

eac h grid p

in

t : 8P i 9S j S j 2 C

v

er ed(P i ) The dual problem is easier to solv e b ecause the C

v

er ed(P i ) zones can b e staticall y computed. p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 8 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 8 Grid

f

test p

in

ts

step

step/x

H1 H1 1,7m test point H2 H2

step/x

p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 9 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 9 Represen tation

f

Co v ered Surfaces In

rder

to express the constrain t S j 2 C

v

er ed(P i ), the C

v

er ed(P i ) m ust b e simple enough. It can b e appro ximated b y

A

rectangle

A

list

f

rectangles Algorithm 1. Compute the C

v

er ed(P i ) zone b y ra y tracing for eac h P i 2. Appro ximate C

v

er ed(P i ) 3. Set the constrain ts S j 2 C

v

er ed(P i ) 4. Do clev er lab eli ng p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 10 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 10 Ra y T racing

sender (xs,ys,zs)

f loor2 f loor1

path (xf,yf,zf) (xw,yw,zw)

wall wall

p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 11 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 11 Appro ximation b y a Union

f

Rectangles p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 12 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 12 Constrain t Handling Rules What? : A declarativ e language designed for writing user-dened constrain ts : a commited-c hoice language with m ulti-headed rules for rewriting the constrain ts in to simple

nes.

Ho w ? : A library for the Prolog ECL i PS e system including

a

translator from constrain t handling rules to Prolog co de,

a

run time for handling the constrain t store. p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 13 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 13 CHR inside constrain t Rules for the inside constrain t stating that a p

in

t is inside a rectangle % inside((X0, Y0), (XLeftLow, YLeftLow)-(X Rig htU p, YRightUp)) inside(_, (Xm, Ym)-(XM, YM)) ==> Xm < XM, Ym < YM. inside((X, Y), (Xm, Ym)-(XM, YM)) ==> Xm < X, X < XM, Ym < Y, Y < YM. inside(XY, (Xm1,Ym1)-(X M1, YM1 )), inside(XY, (Xm2,Ym2)- (XM 2,Y M2) ) <=> Xm is max(Xm1,Xm2) , Ym is max(Ym1,Ym 2), XM is min(XM1,XM2) , YM is min(YM1,YM 2), inside(XY, (Xm,Ym)-(X M,Y M)) . p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 14 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 14 Extension to Union

f

Rectangles Rules for the inside constrain t stating that a p

in

t is within a list

f

rectangles (a GEOMetrical

b

ject) inside(S, L1), inside(S, L2) <=> intersect_geoms(L1 , L2, L3), inside(S, L3). intersect_geoms(L1, L2, L3) <=> setof(Rect, intersect_geom(L 1, L2, Rect), L3). intersect_geom(L1, L2, Rect) <=> member(Rect1, L1), member(Rect2, L2), intersect_rectangl es(Re ct1, Rect2, Rect). p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 15 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 15 Lab eling The constrain t phase asso ciates a sender to eac h C

v

er ed(P i ) zone. The lab eling phase has to c ho

se

the n um b er and the p

sitions
f

the senders. It is expressed b y stating that as man y senders as p

ssible

are equal. equate_senders([]) <=> true. equate_senders([S|L] ) <=> ( member(S, L)

r

true ), % Try to equate a sender with

thers

equate_senders(L). p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 16 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 16 A T rue Example p a ct97 T. Fr

uhwir

th & P. Brisset

SLIDE 17 Optimal Planning

f

Digital Cor d less T ele c

mmunic

ation Systems 17 Conclusion On this application, constrain t tec hnology (CHR) pro v es to

ha

v e big expression p

w

er: the whole program for solving the problem is

nly

a couple

f

h undred lines and required few man-mon ths to b e implemen ted.

b

e exible: the rst protot yp e w as easily extended from rectangles to union

f

rectangles, from 2-D to 3-D, ...

b

e extensible: for example, restricting allo w ed senders lo cations to w alls needs

nly
ne

more inside constrain t.

b

e ecien t: for a t ypical

ce

building, an

ptimal

placemen t is found within a few min utes (up to 25 base stations). p a ct97 T. Fr

uhwir

th & P. Brisset