Mot
ion
Trac
king
for M
obile
, M
otio
n Tr
acki
ng fo
r Mob
ile,
Loca
tion
Loca
tion --
Aw
are
App
licat
ions
Aw
are
App
licat
ions
Can
dida
cy E
xam
Can
dida
cy E
xam
Dre
xel H
alla
way
Dre
xel H
alla
way
Apr
il 5,
200
4A
pril
5, 2
004
Def
initi
ons
and
Sco
peD
efin
ition
s an
d S
cope
Mot
ion
track
ing
Mot
ion
track
ing
––Th
e se
nsor
s an
d co
mpu
tatio
ns b
y w
hich
a s
yste
m
The
sens
ors
and
com
puta
tions
by
whi
ch a
sys
tem
de
term
ines
eith
er ta
rget
’s:
dete
rmin
es e
ither
targ
et’s
:P
ositi
on /
orie
ntat
ion
Pos
ition
/ or
ient
atio
nin
som
e C
arte
sian
fram
e; o
rin
som
e C
arte
sian
fram
e; o
rP
roxi
mity
Pro
xim
ityto
logi
cal e
ntity
or
to lo
gica
l ent
ity o
r res
ourc
e(s
reso
urce
(s)) ——
“in th
e co
py
“in th
e co
py
room
…go
ing
up th
e st
airs
.”ro
om…
goin
g up
the
stai
rs.”
Exc
lusi
ons
Exc
lusi
ons
––M
otio
n C
aptu
re (M
OC
AP
) of e
ntire
bod
y po
se, h
and
Mot
ion
Cap
ture
(MO
CA
P) o
f ent
ire b
ody
pose
, han
d or
eye
trac
king
. . .
.or
eye
trac
king
. . .
.––
Som
e m
echa
nica
l sys
tem
s, w
hich
see
m im
poss
ible
S
ome
mec
hani
cal s
yste
ms,
whi
ch s
eem
impo
ssib
le
to u
se in
a b
road
ly m
obile
con
text
.to
use
in a
bro
adly
mob
ile c
onte
xt.
––P
ure
visi
onP
ure
visi
on-- b
ased
trac
king
.ba
sed
track
ing.
Mot
ivat
ion
Mot
ivat
ion
Why
do
mot
ion
track
ing?
Why
do
mot
ion
track
ing?
––To
orie
nt o
nese
lf w
ith re
fere
nce
to:
To o
rient
one
self
with
refe
renc
e to
:A
virt
ual m
odel
A v
irtua
l mod
el——
view
con
trol i
n V
irtua
l Rea
lity
(VR
).vi
ew c
ontro
l in
Virt
ual R
ealit
y (V
R).
The
real
wor
ld:
The
real
wor
ld:
––N
avig
ate
with
a m
ap.
Nav
igat
e w
ith a
map
.––
Enh
ance
with
virt
ual a
ugm
enta
tions
Enh
ance
with
virt
ual a
ugm
enta
tions
——in
clud
ing
view
con
trol f
or v
irtua
l in
clud
ing
view
con
trol f
or v
irtua
l ov
erla
ys.
over
lays
.
––To
hav
e th
e fre
e ch
oice
to s
hare
one
’s p
ose:
To h
ave
the
free
choi
ce to
sha
re o
ne’s
pos
e:Fo
r rem
ote
colla
bora
tion
For r
emot
e co
llabo
ratio
n ——e.
g., a
vata
r ani
mat
ion.
e.g.
, ava
tar a
nim
atio
n.Fo
r per
sona
l saf
ety
For p
erso
nal s
afet
y ——w
hen
dial
ing
911
on a
w
hen
dial
ing
911
on a
cel
lpho
nece
llpho
ne..
Em
phas
isE
mph
asis
––S
elf
Sel
f -- tra
ckin
gtra
ckin
g ——pr
imar
ily a
ser
vice
for t
he “t
arge
t,” n
ot th
e pr
imar
ily a
ser
vice
for t
he “t
arge
t,” n
ot th
e in
frast
ruct
ure
or o
rgan
izat
ion.
infra
stru
ctur
e or
org
aniz
atio
n.––
Sys
tem
s th
at c
an m
eet (
or c
ould
app
roac
h) re
alS
yste
ms
that
can
mee
t (or
cou
ld a
ppro
ach)
real
-- tim
e la
tenc
ies,
tim
e la
tenc
ies,
an
d su
ban
d su
b --se
cond
upd
ate
frequ
enci
es.
seco
nd u
pdat
e fre
quen
cies
.
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infra
red
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79
Add
lese
e19
97
Priy
anth
a 20
00Fo
xlin
199
8
Wan
t 199
2H
irose
200
3N
ewm
an 2
001
Judd
199
7
Ran
dell
2001
Ran
dell
2003
Foxl
in 1
996
You
199
9W
ang
2000
Dju
knic
2001
Sol
iman
2000
Bah
l 200
0E
nge
1999
Hed
ges
2001
Hal
law
ay 2
003
Wel
ch 1
999
Chi
ntha
mm
it20
03
Teth
ered
?Te
ther
ed?
NO
USU
ALL
YYE
S
Som
e sy
stem
s re
quire
wire
d co
nnec
tions
.S
ome
syst
ems
requ
ire w
ired
conn
ectio
ns.
Mob
ile u
sers
Mob
ile u
sers
——w
ho m
ay w
ant t
o le
ave
dom
ain
who
may
wan
t to
leav
e do
mai
n ——do
n’t w
ant t
o ha
ve to
det
ach!
don’
t wan
t to
have
to d
etac
h!
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79A
ddle
see
1997
Priy
anth
a 20
00Fo
xlin
199
8W
ant 1
992
Hiro
se 2
003
New
man
200
1
Judd
199
7
Ran
dell
2001
Ran
dell
2003
Foxl
in 1
996
You
199
9W
ang
2000
Dju
knic
2001
Sol
iman
2000
Bah
l 200
0
Eng
e19
99
Hed
ges
2001
Hal
law
ay 2
003
Wel
ch 1
999
Chi
ntha
mm
it20
03
Can
Sel
fC
an S
elf -- T
rack
?Tr
ack?
NO
UN
CER
TAIN
YESW
ho “c
onsu
mes
” the
trac
king
dat
a?W
ho “c
onsu
mes
” the
trac
king
dat
a?––
In s
elf
In s
elf -- t
rack
ing,
the
track
ed u
ser i
s m
akin
g us
e of
his
tra
ckin
g, th
e tra
cked
use
r is
mak
ing
use
of h
is
own
posi
tion
/ orie
ntat
ion.
own
posi
tion
/ orie
ntat
ion.
––S
ome
alte
rnat
ives
trac
k pe
ople
for m
anag
eria
l or
Som
e al
tern
ativ
es tr
ack
peop
le fo
r man
ager
ial o
r or
gani
zatio
nal p
urpo
ses.
orga
niza
tiona
l pur
pose
s.––
Use
r priv
acy
is o
ne n
atur
al c
once
rn.
Use
r priv
acy
is o
ne n
atur
al c
once
rn.
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79
Add
lese
e19
97P
riyan
tha
2000
Foxl
in 1
998
Wan
t 199
2
Hiro
se 2
003
New
man
200
1Ju
dd 1
997
R
ande
ll 20
01R
ande
ll 20
03
Foxl
in 1
996
You
199
9W
ang
2000
Dju
knic
2001
Sol
iman
2000
Bah
l 200
0E
nge
1999
Hed
ges
2001
Hal
law
ay 2
003
Wel
ch 1
999
Chi
ntha
mm
it20
03
““ Insi
deIn
side
-- Loo
king
Look
ing --
Out
?”O
ut?”
NO
YES
DEP
END
S
Are
the
sens
ors
mov
ing
with
the
user
?A
re th
e se
nsor
s m
ovin
g w
ith th
e us
er?
––Th
e be
st la
tenc
y ch
arac
teris
tics.
The
best
late
ncy
char
acte
ristic
s.––
Som
e po
ssib
ility
of u
ser
Som
e po
ssib
ility
of u
ser --
loca
l pro
cess
ing.
loca
l pro
cess
ing.
Ano
ther
usa
ge o
f A
noth
er u
sage
of “
Insi
de“In
side
-- Out
.”O
ut.”
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79
Add
lese
e19
97P
riyan
tha
2000
Foxl
in 1
998
Wan
t 199
2
Hiro
se 2
003
New
man
200
1Ju
dd 1
997
Ran
dell
2001
Ran
dell
2003
Foxl
in 1
996
You
199
9W
ang
2000
Dju
knic
2001
Sol
iman
2000
Bah
l 200
0
Eng
e19
99H
edge
s 20
01H
alla
way
200
3W
elch
199
9C
hint
ham
mit
2003
Pre
pare
d E
nviro
nmen
t?P
repa
red
Env
ironm
ent?
NO
YES
Idea
lly, a
trac
king
sys
tem
wou
ld re
quire
no
Idea
lly, a
trac
king
sys
tem
wou
ld re
quire
no
envi
ronm
enta
l sup
port.
envi
ronm
enta
l sup
port.
––S
ome
don’
t.S
ome
don’
t.––
Oth
ers
use
envi
ronm
enta
l res
ourc
es u
ser d
idn’
t set
up.
Oth
ers
use
envi
ronm
enta
l res
ourc
es u
ser d
idn’
t set
up.
––S
till o
ther
s re
quire
sig
nific
ant i
nfra
stru
ctur
al in
vest
men
t S
till o
ther
s re
quire
sig
nific
ant i
nfra
stru
ctur
al in
vest
men
t an
d in
stal
latio
n.an
d in
stal
latio
n.
NA
TUR
E O
R O
THER
Deg
rees
of F
reed
om P
rovi
ded?
Deg
rees
of F
reed
om P
rovi
ded?
3DO
F O
rient
atio
n6D
OF
3DO
F Po
sitio
nO
ther
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79
Add
lese
e19
97
Priy
anth
a 20
00Fo
xlin
199
8
Wan
t 199
2H
irose
200
3
New
man
200
1
Judd
199
7
Ran
dell
2001
Ran
dell
2003
Foxl
in 1
996
You
199
9
Wan
g 20
00
Dju
knic
2001
Sol
iman
2000
Bah
l 200
0E
nge
1999
Hed
ges
2001
Hal
law
ay 2
003
Wel
ch 1
999
Chi
ntha
mm
it20
03
2DO
F Po
sitio
n
Acc
urac
y Ti
er?
Acc
urac
y Ti
er?
cm-/
deg
ree-
leve
l<
m<
5m>
5m
Foxl
in 2
003
Saw
ada
2003
Gol
ding
1999
Raa
b 19
79A
ddle
see
1997
Priy
anth
a 20
00
Foxl
in 1
998
Wan
t 199
2H
irose
200
3
New
man
200
1
Judd
199
7
Ran
dell
2001
Ran
dell
2003
Foxl
in 1
996
You
199
9
Wan
g 20
00D
jukn
ic20
01S
olim
an20
00B
ahl 2
000
Eng
e19
99H
edge
s 20
01
Hal
law
ay 2
003
Wel
ch 1
999
Chi
ntha
mm
it20
03
N/A
Acc
urac
y re
quire
men
ts d
iffer
by
task
.A
ccur
acy
requ
irem
ents
diff
er b
y ta
sk.
Gen
eral
ly, h
ighe
r acc
urac
y co
sts
in:
Gen
eral
ly, h
ighe
r acc
urac
y co
sts
in:
––S
yste
m p
rice,
and
/or
Sys
tem
pric
e, a
nd/o
r––
Ela
bora
te in
frast
ruct
ure,
set
up /
calib
ratio
n.E
labo
rate
infra
stru
ctur
e, s
etup
/ ca
libra
tion.
Coa
rse
syst
ems
are
gene
rally
sim
pler
, che
aper
, C
oars
e sy
stem
s ar
e ge
nera
lly s
impl
er, c
heap
er,
and
som
etim
es s
uffic
ient
. . .
.an
d so
met
imes
suf
ficie
nt .
. . .
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infra
red
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
Fibe
ropt
icFi
bero
ptic
Gyr
osco
pe (S
awad
a, e
t al.,
200
2)G
yros
cope
(Saw
ada,
et a
l., 2
002)
Fran
cis
Fran
cis
Har
ress
Har
ress
(191
1) a
nd G
eorg
es
(191
1) a
nd G
eorg
es S
agna
cS
agna
c(1
913)
(191
3)——
the
“th
e “ S
agna
cS
agna
cE
ffect
.”E
ffect
.”FO
Gs
FOG
s : p
hase
shi
ft : p
hase
shi
ft ra
te o
f rot
atio
n ar
ound
axi
s.ra
te o
f rot
atio
n ar
ound
axi
s.
Ω=
cLDs
λπφ
2
100m
FO
coi
l
Fibe
ropt
icFi
bero
ptic
Gyr
osco
pe (S
awad
a, e
t al.,
200
2)G
yros
cope
(Saw
ada,
et a
l., 2
002)
Hea
ding
acc
urac
y (d
rift)
spec
H
eadi
ng a
ccur
acy
(drif
t) sp
ec
11 ºº/h
r/h
r ((σσ )
, act
ually
less
.),
actu
ally
less
.PP
itch
and
roll
accu
racy
itc
h an
d ro
ll ac
cura
cy
spec
ified
at +
/sp
ecifi
ed a
t +/ -- 0
.50.5 ºº
, act
ually
, a
ctua
lly
<0.1
<0.1
ºº ——bu
t don
e w
ith a
thre
ebu
t don
e w
ith a
thre
e --ax
is ra
te ta
ble
. . .
no la
tera
l ax
is ra
te ta
ble
. . .
no la
tera
l ac
cele
ratio
ns!!
acce
lera
tions
!!
Dea
d R
ecko
ning
(Ran
dell
2003
)D
ead
Rec
koni
ng (R
ande
ll 20
03)
An
idea
cen
turie
s ol
dA
n id
ea c
entu
ries
old ——
anci
ent
anci
ent
mar
iner
s . .
. .
mar
iner
s . .
. .
Ran
dell
pres
ente
d 4
case
stu
dies
Ran
dell
pres
ente
d 4
case
stu
dies
––O
ptin
g fo
r che
ap s
enso
rs, b
ut:
Opt
ing
for c
heap
sen
sors
, but
:H
opin
g fo
r goo
d st
ep d
etec
tion.
Hop
ing
for g
ood
step
det
ectio
n.H
opin
g fo
r acc
urat
e he
adin
g.H
opin
g fo
r acc
urat
e he
adin
g.––
Bes
t cas
e st
udy
“Gar
den
Pat
h”B
est c
ase
stud
y “G
arde
n P
ath”
22 --ax
is m
agne
tom
eter
tape
d to
sho
ulde
r.ax
is m
agne
tom
eter
tape
d to
sho
ulde
r.Tw
o ac
cele
rom
eter
s, o
ne o
n ea
ch fo
ot.
Two
acce
lero
met
ers,
one
on
each
foot
.M
oder
atel
y so
phis
ticat
ed s
tep
dete
ctio
n.M
oder
atel
y so
phis
ticat
ed s
tep
dete
ctio
n.12
6m ro
ute
track
ed a
t 122
m, σ
= 2
.2m
.12
6m ro
ute
track
ed a
t 122
m, σ
= 2
.2m
.
Dea
d R
ecko
ning
Mod
ule
(Jud
d 19
97)
Dea
d R
ecko
ning
Mod
ule
(Jud
d 19
97)
Foot
Foot
-- sol
dier
nav
igat
ion
sold
ier n
avig
atio
n w
ithou
t hol
ding
com
pass
and
w
ithou
t hol
ding
com
pass
and
m
anua
lly c
ount
ing
pace
s.m
anua
lly c
ount
ing
pace
s.Tr
iTr
i -- axi
al m
agne
tom
eter
s an
d ax
ial m
agne
tom
eter
s an
d ac
cele
rom
eter
sac
cele
rom
eter
s ——co
rrec
t co
rrec
t co
mpa
ss fo
r atti
tude
and
co
mpa
ss fo
r atti
tude
and
de
tect
/ co
unt p
aces
.de
tect
/ co
unt p
aces
.
7x23
m, p
lotte
r fill
ed in
gap
s
Dea
d R
ecko
ning
Mod
ule
(Jud
d 19
97)
Dea
d R
ecko
ning
Mod
ule
(Jud
d 19
97)
Wor
ks b
est i
n co
ncer
t W
orks
bes
t in
conc
ert
with
GP
S:
with
GP
S:
––G
PS
GP
S-- in
por
t.in
por
t.––
GP
S c
orre
cts
whe
n G
PS
cor
rect
s w
hen
avai
labl
e an
d ac
cura
te.
avai
labl
e an
d ac
cura
te.
––D
R fi
lls in
whe
n no
t.D
R fi
lls in
whe
n no
t.––
Kal
man
filte
r (K
F) fu
ses
Kal
man
filte
r (K
F) fu
ses
inpu
ts a
nd a
djus
ts
inpu
ts a
nd a
djus
ts
head
ing
and
step
leng
th.
head
ing
and
step
leng
th.
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
Rud
olph
E. K
alm
anR
udol
ph E
. Kal
man
––bo
rn H
unga
ry 1
930
born
Hun
gary
193
0––
Col
umbi
a C
olum
bia
D.S
ciD
.Sci
.; 19
57.;
1957
––se
min
al p
aper
196
0se
min
al p
aper
196
0R
ecur
sive
sol
utio
n he
pro
ved
optim
al to
R
ecur
sive
sol
utio
n he
pro
ved
optim
al to
th
e pr
oble
m o
f dis
cret
eth
e pr
oble
m o
f dis
cret
e --da
ta li
near
da
ta li
near
fil
terin
g, a
ssum
ing
filte
ring,
ass
umin
g pr
oces
s an
d pr
oces
s an
d m
easu
rem
ent n
oise
are
:m
easu
rem
ent n
oise
are
:––
Inde
pend
ent o
f one
ano
ther
Inde
pend
ent o
f one
ano
ther
––““ w
hite
whi
te””
––G
auss
ian
/ nor
mal
(zer
oG
auss
ian
/ nor
mal
(zer
o --m
ean)
mea
n)
()n
nn
n
nn
n
xz
nx
x
zn
xnn
x
zx
−+
+=
++
+=
=
+++
111
11
11
11
1
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
Est
imat
e a
cons
tant
you
E
stim
ate
a co
nsta
nt y
ou
can
mea
sure
dire
ctly
?ca
n m
easu
re d
irect
ly?
Equ
al c
erta
inty
for e
ach
Equ
al c
erta
inty
for e
ach
mea
sure
men
tm
easu
rem
ent i
s ju
st a
is
just
a
recu
rsiv
e av
erag
e.re
curs
ive
aver
age.
“res
idua
l” or
“inn
ovat
ion”
“gai
n”
(]
()
221
11
22
2
22
21
22 1
11
ˆ )1(
ˆ
ˆˆ
ˆ
1,0,
ˆˆ
1ˆ
1ˆ
1
ˆ ,ˆ
1
1 kk
kk
kk
zk
k
zk
k
z
k
xz
kx
x
kk
zx
k
k
σσ
σσ
σ
σσ
σ
σσ −
=
−+
=
∈+
=
+=
== +
++
+
+
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)D
iffer
ent c
erta
intie
s fo
r eac
h m
easu
rem
ent
Diff
eren
t cer
tain
ties
for e
ach
mea
sure
men
t ??N
eed
a N
eed
a ““ K
alm
an g
ain
Kal
man
gai
n ””to
wei
ght i
nnov
atio
n an
d ex
pres
s to
wei
ght i
nnov
atio
n an
d ex
pres
s fil
ter
filte
r ’’ s v
aria
nce.
s va
rianc
e.
Kal
man
pro
ved
this
sor
t of ‘
k’
min
imiz
ed e
rror
in th
e es
timat
es—
the
scal
ar “K
alm
an g
ain.
”
Sam
e re
sidu
al .
. .
Kal
man
gai
n!!
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
()
()
− ++
+
− ++
+− +
+
−− +
− ++
− +− ++
− +− +
−=
−+
=
+=
+=
+==
11
1
11
11
1
11
11
1
1111
)(
)ˆ
(ˆ
ˆ
ˆˆ
kk
k
kk
kk
k
Tk
Tk
k
Tk
Tk
k
Tk
k
kk
PH
KI
P
xH
zK
xx
RH
HP
HP
K
RH
HP
HP
K
QA
APP
xAx
Pre
dict
sta
te a
nd c
ovar
ianc
e—su
per-
min
us m
eans
a
prio
ri—us
ing
stat
e-tra
nsiti
on m
atrix
A.
Rem
embe
r cov
(p) =
Pco
v(A
p) =
AP
AT ?
We
prop
agat
e bo
th th
e st
ate
and
its e
rror
by
A.
We
add
Q, t
he p
roce
ss n
oise
cov
aria
nce
mat
rix.
Htra
nsfo
rms
stat
e to
mea
sure
-m
ent“
spac
e”—
in w
hich
, thi
s lo
oks
a lo
t lik
e th
e sc
alar
σve
rsio
n be
fore
!
Sin
ce w
e do
n’t r
eally
div
ide
mat
rices
. . .
!
Sam
e as
bef
ore,
exc
ept o
n pr
e-di
ctio
n, n
ot la
st v
alue
…no
w
mat
rix K
,and
Hto
get
xin
to
mea
sure
men
t spa
ce.
Rat
her l
ike
befo
re .
. . .
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
Filte
r Fi
lter ““
tuni
ngtu
ning
”” is
mak
ing
good
cho
ices
for p
roce
ss a
nd
is m
akin
g go
od c
hoic
es fo
r pro
cess
and
m
easu
rem
ent n
oise
m
easu
rem
ent n
oise
cov
aria
nces
cova
rianc
es, , QQ
and
and
RR. .
Sca
lar
Sca
lar RR
is 1
on
left,
and
0.0
001
on ri
ght .
. . .
is 1
on
left,
and
0.0
001
on ri
ght .
. . .
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
Som
etim
es w
e ca
nS
omet
imes
we
can ’’
t mea
sure
eno
ugh
to d
o a
com
plet
e t m
easu
re e
noug
h to
do
a co
mpl
ete
stat
e es
timat
e in
a g
iven
mom
ent.
stat
e es
timat
e in
a g
iven
mom
ent.
Wai
ting
for t
he
Wai
ting
for t
he re
stre
stfa
lse
fals
e““ s
imul
tane
ity a
ssum
ptio
n.si
mul
tane
ity a
ssum
ptio
n.””
Gre
g W
elch
coi
ned
the
term
G
reg
Wel
ch c
oine
d th
e te
rm ““ S
CA
AT
SC
AA
T ””(s
ingl
e co
nstra
int
(sin
gle
cons
train
t at
a ti
me)
to d
escr
ibe
a us
ing
a K
F w
ith:
at a
tim
e) to
des
crib
e a
usin
g a
KF
with
:––
Diff
eren
t obs
erva
tion
mat
rix
Diff
eren
t obs
erva
tion
mat
rix HH
at e
ach
step
to
at e
ach
step
to ““ s
trip
dow
nst
rip d
own ””
the
the
stat
e to
the
stat
e to
the
elem
ent(s
elem
ent(s
) mea
sure
d no
w.
) mea
sure
d no
w.
––HH
also
mas
ks th
e pa
rts o
f the
cov
aria
nce
and
gain
mat
rices
the
also
mas
ks th
e pa
rts o
f the
cov
aria
nce
and
gain
mat
rices
the
mea
sure
men
t(sm
easu
rem
ent(s
) sho
uld
effe
ct.
) sho
uld
effe
ct.
Ass
umin
g al
l nee
ded
mea
sure
men
t A
ssum
ing
all n
eede
d m
easu
rem
ent ““
type
sty
pes ””
arriv
e of
ten
arriv
e of
ten
enou
gh, t
he fi
lter s
till c
onve
rges
, and
arg
uabl
y m
ore
enou
gh, t
he fi
lter s
till c
onve
rges
, and
arg
uabl
y m
ore
rigor
ousl
y.rig
orou
sly.
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
Can
we
use
Can
we
use
KFs
KFs
on n
onon
non
-- line
ar p
roce
sses
?lin
ear p
roce
sses
?
The
Ext
ende
d K
alm
an F
ilter
(EK
F) d
oes
just
that
The
Ext
ende
d K
alm
an F
ilter
(EK
F) d
oes
just
that
——in
an
in a
n ap
prox
imat
e, n
ot th
eore
tical
ly o
ptim
al, b
ut o
ften
very
ap
prox
imat
e, n
ot th
eore
tical
ly o
ptim
al, b
ut o
ften
very
us
eful
way
:us
eful
way
:––
We
use
our n
onW
e us
e ou
r non
-- line
ar p
roce
sslin
ear p
roce
ss-- tr
ansi
tion
func
tion
trans
ition
func
tion
ff , w
here
bef
ore
, whe
re b
efor
e w
e ha
d m
atrix
w
e ha
d m
atrix
AA..
––A
pos
sibl
y no
nA
pos
sibl
y no
n --lin
ear o
bser
vatio
n fu
nctio
n lin
ear o
bser
vatio
n fu
nctio
n h h
repl
aces
mat
rixre
plac
es m
atrix
H.
H.
––W
e W
e ““ li
near
ize
linea
rize ””
f f and
an
d h h
into
in
to J
acob
ian
Jaco
bian
mat
rices
.m
atric
es.
––Ja
cobi
ans
Jaco
bian
spr
opag
ate
the
cova
rianc
e m
atric
es o
f the
filte
r lik
e pr
opag
ate
the
cova
rianc
e m
atric
es o
f the
filte
r lik
e A
A
an
d an
d HH
befo
re.
befo
re.
––Ja
cobi
ans
Jaco
bian
sar
e de
fined
by
the
deve
lope
r, bu
t get
eva
luat
ed a
t ar
e de
fined
by
the
deve
lope
r, bu
t get
eva
luat
ed a
t ea
ch s
tep
as a
loca
l, lin
ear a
ppro
xim
atio
n of
the
effe
cts
of th
eea
ch s
tep
as a
loca
l, lin
ear a
ppro
xim
atio
n of
the
effe
cts
of th
e ir ir
rela
ted
func
tions
.re
late
d fu
nctio
ns.
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
[]
[] []
[]
[] []
[]
[] []
[]
[] []
)0,~ (
)0,~ (
)0,ˆ (
)0,ˆ (
,,
1,
1,
kji
ji
kji
ji
kji
ji
kji
ji
xvh
V
xxh
H
xwf
W
xxf
A
∂∂=
∂∂=
∂∂=
∂∂=
−−
)0,~ (
~)0,
ˆ (~
),
()
,(
11
kk
kk
kk
k
kk
k
xh
zx
fx
vx
hz
wx
fx
====−
−
Vec
tors
w a
nd v
are
pro
cess
and
m
easu
rem
ent n
oise
vec
tors
, who
se
cova
rianc
e m
atric
es w
e sa
w b
efor
e as
, re
spec
tivel
y, Q
and
R.
Kal
man
Filt
er (W
elch
, Bis
hop,
K
alm
an F
ilter
(Wel
ch, B
isho
p, 1
995)
1995
)
()
−
−−
−−
−
−−
−
−−
−=
−+
=
+=
+=
==
kk
kk
kk
kk
k
T kk
kT k
kk
T kk
k
T kk
kT k
kk
k
kk
k
PH
KI
P
xh
zK
xx
VR
VH
PH
HP
K
WQ
WA
PA
P
xf
xx
)(
)]0,ˆ (
[ˆ
ˆ
)0,ˆ (
~ˆ
11
1
1
Not
e th
at th
e co
varia
nces
of n
oise
are
now
bra
cket
ed in
Jac
obia
nsto
pro
paga
te
thei
r effe
cts—
sinc
e fa
nd h
depe
nd o
n w
and
v, re
spec
tivel
y, b
ut th
ose
valu
es w
ere
set t
o th
eir z
ero
expe
ctat
ions
. . .
.
Iner
tial H
ead
Iner
tial H
ead --
Trac
king
(Fox
lin 1
996)
Trac
king
(Fox
lin 1
996)
Pro
blem
: D
rifty
gyr
os
Pro
blem
: D
rifty
gyr
os
and
and
slos
hysl
oshy
, flu
id
, flu
id
incl
inom
eter
. . .
how
to
incl
inom
eter
. . .
how
to
get b
ette
r res
ults
?ge
t bet
ter r
esul
ts?
Ans
wer
: In
add
ition
to
Ans
wer
: In
add
ition
to
the
3D o
rient
atio
n th
e 3D
orie
ntat
ion
stat
e, m
odel
the
gyro
st
ate,
mod
el th
e gy
ro
drift
/bia
s as
“sta
te” .
. .
drift
/bia
s as
“sta
te” .
. .
figur
e ou
t how
to
figur
e ou
t how
to
corr
ect d
rift w
ith
corr
ect d
rift w
ith
incl
inom
eter
. . .
.in
clin
omet
er .
. . .
Iner
tial H
ead
Iner
tial H
ead --
Trac
king
(Fox
lin 1
996)
Trac
king
(Fox
lin 1
996)
Doe
sn’t
do th
e co
nven
tiona
l, di
rect
Kal
man
filte
r (K
F)
Rat
her,
a C
ompl
emen
tary
KF
Iner
tial H
ead
Iner
tial H
ead --
Trac
king
(Fox
lin 1
996)
Trac
king
(Fox
lin 1
996)
Filte
r mea
sure
men
t is
the
Filte
r mea
sure
men
t is
the
slos
hysl
oshy
incl
inom
eter
and
in
clin
omet
er a
nd
com
pass
inpu
ts.
com
pass
inpu
ts.
Mea
sure
men
t noi
se c
ovar
ianc
e M
easu
rem
ent n
oise
cov
aria
nce
RR––
Run
ning
val
ue fo
rR
unni
ng v
alue
for ““
still
tim
est
ill ti
me ””
ττ ..––σσ vv
= 1
/ (1
+ 40
0=
1 / (
1 +
400 ττ
)) but
if
but i
f σσvv
< 0.
01
< 0.
01 th
enth
enσσ vv
= 0.
0=
0.0 1
.1.––
Com
pass
get
s hi
gher
var
ianc
e.C
ompa
ss g
ets
high
er v
aria
nce.
=
σσ
σ
v
v
vR
00
00
002
2
Iner
tial H
ead
Iner
tial H
ead --
Trac
king
(Fox
lin 1
996)
Trac
king
(Fox
lin 1
996)
KF
disa
bled
(no
resi
dual
gai
n)K
F en
able
d (g
ain
driv
en b
y “s
tilln
ess”
)
Indo
ors
with
Che
ap S
enso
rs (
Indo
ors
with
Che
ap S
enso
rs ( G
oldi
ngG
oldi
ng19
99)
1999
)
Indo
or, u
npre
pare
d In
door
, unp
repa
red
envi
ronm
ent.
envi
ronm
ent.
Che
ap s
enso
rs w
orn
by u
ser:
Che
ap s
enso
rs w
orn
by u
ser:
––Tr
iTr
i -- axi
al a
ccel
erom
eter
and
ax
ial a
ccel
erom
eter
and
m
agne
tom
eter
.m
agne
tom
eter
.––
Fluo
resc
ent l
ight
det
ecto
r.Fl
uore
scen
t lig
ht d
etec
tor.
––Te
mpe
ratu
re s
enso
r.Te
mpe
ratu
re s
enso
r.B
ayes
ian
infe
rent
ial a
ppro
ach
Bay
esia
n in
fere
ntia
l app
roac
h––
Trai
ning
on
“loca
tions
” bui
lds
Trai
ning
on
“loca
tions
” bui
lds
sens
or d
istri
butio
ns.
sens
or d
istri
butio
ns.
Indo
ors
with
Che
ap S
enso
rs (
Indo
ors
with
Che
ap S
enso
rs ( G
oldi
ngG
oldi
ng19
99)
1999
)
Trac
king
sam
ples
sen
sors
at 2
0Hz:
Trac
king
sam
ples
sen
sors
at 2
0Hz:
––S
enso
r sta
tes,
pre
viou
s pr
obab
ility
+
Sen
sor s
tate
s, p
revi
ous
prob
abili
ty +
Bay
esB
ayes
’ Rul
e’ R
ule
Res
ults
:R
esul
ts:
––B
efor
e “c
ooki
ng”
Bef
ore
“coo
king
” zz-- a
ccel
acce
l -- var
ianc
eva
rianc
e , 5
2% e
rror
s!, 5
2% e
rror
s!––
Afte
r, 2%
err
or ra
te, t
houg
h va
rianc
e ta
kes
long
er to
A
fter,
2% e
rror
rate
, tho
ugh
varia
nce
take
s lo
nger
to
resp
ond.
resp
ond.
Iner
tial /
Dea
d R
ecko
ning
Sum
mar
yIn
ertia
l / D
ead
Rec
koni
ng S
umm
ary
Hug
e ad
vant
ages
:H
uge
adva
ntag
es:
––N
o en
viro
nmen
tal p
repa
ratio
n.N
o en
viro
nmen
tal p
repa
ratio
n.––
Mob
ile, s
elf
Mob
ile, s
elf -- c
onta
ined
.co
ntai
ned.
––O
ften
high
upd
ate
rate
s.O
ften
high
upd
ate
rate
s.––
Clo
sest
thin
g to
a “S
ilver
Bul
let”
we’
ll lik
ely
find!
Clo
sest
thin
g to
a “S
ilver
Bul
let”
we’
ll lik
ely
find!
––S
elf
Sel
f -- tra
ckin
gtra
ckin
g ——no
no
priv
acy
priv
acy
conc
erns
.co
ncer
ns.
Hug
e pr
oble
m:
With
out s
omet
hing
abs
olut
e to
H
uge
prob
lem
: W
ithou
t som
ethi
ng a
bsol
ute
to
corr
ect t
hem
, all
truly
iner
tial s
yste
ms
drift
.co
rrec
t the
m, a
ll tru
ly in
ertia
l sys
tem
s dr
ift.
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infra
red
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
Act
ive
Floo
r (A
ctiv
e Fl
oor (
Add
lese
eA
ddle
see
1997
)19
97)
Mec
hani
cal /
gra
vim
etric
M
echa
nica
l / g
ravi
met
ric
track
ing
of w
alke
rs.
track
ing
of w
alke
rs.
Load
cel
ls a
t cor
ners
of f
loor
Lo
ad c
ells
at c
orne
rs o
f flo
or
tiles
.til
es.
Mea
sure
ver
tical
com
pone
nt o
f M
easu
re v
ertic
al c
ompo
nent
of
Gro
und
Rea
ctiv
e Fo
rce
(GR
F).
Gro
und
Rea
ctiv
e Fo
rce
(GR
F).
Hid
den
Mar
kov
Mod
els
try to
H
idde
n M
arko
v M
odel
s try
to
““ rec
ogni
zere
cogn
ize ””
wal
ker.
wal
ker.
Act
ive
Floo
r (A
ctiv
e Fl
oor (
Add
lese
eA
ddle
see
1997
)19
97)
B
A,=
λ
15 s
ubje
cts:
15 s
ubje
cts:
––20
sig
natu
re s
teps
eac
h.20
sig
natu
re s
teps
eac
h.––
10 fo
r tra
inin
g, 1
0 fo
r tes
ting
10 fo
r tra
inin
g, 1
0 fo
r tes
ting
HM
Ms
HM
Ms ..
Bes
t res
ults
(You
ngB
est r
esul
ts (Y
oung
’’ s H
TK: H
MM
Too
lkit)
:s
HTK
: HM
M T
oolk
it):
––U
sing
7 s
tate
s, 2
5 sa
mpl
es /
obse
rvat
ion,
2U
sing
7 s
tate
s, 2
5 sa
mpl
es /
obse
rvat
ion,
2-- s
ampl
e ov
erla
p.sa
mpl
e ov
erla
p.––
Mis
sed
only
13
out o
f 150
M
isse
d on
ly 1
3 ou
t of 1
50 te
st id
entif
icat
ions
.te
st id
entif
icat
ions
.B
ut, n
orm
aliz
ing
sign
atur
es fo
r wei
ght i
ncre
ased
the
But
, nor
mal
izin
g si
gnat
ures
for w
eigh
t inc
reas
ed th
e er
rors
to 5
0%+!
er
rors
to 5
0%+!
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infra
red
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
The
“Cric
ket”
(Priy
anth
a 20
00)
The
“Cric
ket”
(Priy
anth
a 20
00)
Wor
ldW
orld
-- affi
xed
beac
ons
affix
ed b
eaco
ns––
sim
ulta
neou
s R
F an
d ul
traso
nic
chirp
. si
mul
tane
ous
RF
and
ultra
soni
c ch
irp.
––N
o ce
ntra
l sch
edul
ing
No
cent
ral s
ched
ulin
g ——ra
ndom
ized
eith
er s
ide
of 0
.25s
.ra
ndom
ized
eith
er s
ide
of 0
.25s
.––
Und
er $
10 e
ach,
rang
e of
~30
Und
er $
10 e
ach,
rang
e of
~30
’’ ..R
F ca
rrie
s st
ring
for n
etw
ork
look
up in
Inte
ntio
nal
RF
carr
ies
strin
g fo
r net
wor
k lo
okup
in In
tent
iona
l N
amin
g S
ervi
ce (I
NS
)N
amin
g S
ervi
ce (I
NS
)––
Map
Map
––R
esou
rces
Res
ourc
esN
o ce
ntra
l adm
in, e
xcep
t the
INS
in th
is im
plem
enta
tion.
No
cent
ral a
dmin
, exc
ept t
he IN
S in
this
impl
emen
tatio
n.
The
“Cric
ket”
(Priy
anth
a 20
00)
The
“Cric
ket”
(Priy
anth
a 20
00)
List
ener
s w
ith R
SLi
sten
ers
with
RS
-- 232
to m
obile
.23
2 to
mob
ile.
––al
so <
$10
also
<$1
0––
sing
lesi
ngle
-- chi
p R
F re
ceiv
erch
ip R
F re
ceiv
er––
sing
lesi
ngle
-- chi
p to
nech
ip to
ne-- d
etec
tor c
ircui
tde
tect
or c
ircui
tR
ange
from
aco
ustic
TO
F.R
ange
from
aco
ustic
TO
F.To
det
erm
ine
clos
est C
ricke
tTo
det
erm
ine
clos
est C
ricke
t ——pr
oxim
ity.
prox
imity
.H
ow to
get
clo
sest
pro
xim
ity?
How
to g
et c
lose
st p
roxi
mity
?––
Mai
ntai
ned
traili
ng b
uffe
r of c
hirp
s.M
aint
aine
d tra
iling
buf
fer o
f chi
rps.
––S
tatic
cas
e:
Sta
tic c
ase:
Min
Mea
nM
inM
ean
and
and
Min
Mod
eM
inM
ode
both
bo
th
bette
r tha
n M
ajor
ity.
bette
r tha
n M
ajor
ity.
––D
ynam
ic c
ase:
D
ynam
ic c
ase:
Min
Mod
eM
inM
ode
som
ewha
t so
mew
hat
bette
r tha
n be
tter t
han
Min
Mea
nM
inM
ean ..
Can
dis
tingu
ish
~1ft.
pas
t bas
elin
e C
an d
istin
guis
h ~1
ft. p
ast b
asel
ine
betw
een
beac
ons
as n
ear a
s 4
betw
een
beac
ons
as n
ear a
s 4 ’’
apar
t.ap
art.
The
“Bat
” (N
ewm
an 2
001)
The
“Bat
” (N
ewm
an 2
001)
Infra
stru
ctur
e po
lls e
ach
bat v
ia
Infra
stru
ctur
e po
lls e
ach
bat v
ia
Wi
Wi -- F
iFi(2
0ms
slot
s).
(20m
s sl
ots)
.M
obile
Bat
repl
ies
with
ultr
ason
ic
Mob
ile B
at re
plie
s w
ith u
ltras
onic
ch
irp.
chirp
.S
yste
mS
yste
m––
mea
sure
s TO
F to
cei
ling
rece
iver
s.m
easu
res
TOF
to c
eilin
g re
ceiv
ers.
––Tr
ilate
rate
sTr
ilate
rate
sa
posi
tion
a po
sitio
n ——w
ithin
w
ithin
3cm
, 3c
m,
95%
95%
of th
e tim
e.of
the
time.
Mob
iles
get s
tate
thro
ugh
CO
RB
A
Mob
iles
get s
tate
thro
ugh
CO
RB
A
obje
ct o
ver n
etw
ork.
obje
ct o
ver n
etw
ork.
HM
D /
AR
and
PD
A v
ersi
ons
. . .
.H
MD
/ A
R a
nd P
DA
ver
sion
s . .
. .
The
“Bat
” (N
ewm
an 2
001)
The
“Bat
” (N
ewm
an 2
001)
HM
D /
AR
ver
sion
HM
D /
AR
ver
sion
––H
elm
et, 3
Bat
s 1
iner
tial o
rient
atio
n tra
cker
.H
elm
et, 3
Bat
s 1
iner
tial o
rient
atio
n tra
cker
.––
gets
con
secu
tive
slot
s fo
r 3 c
hirp
s.ge
ts c
onse
cutiv
e sl
ots
for 3
chi
rps.
––O
rient
atio
n fro
m 3
pos
ition
s if
cohe
rent
.O
rient
atio
n fro
m 3
pos
ition
s if
cohe
rent
.––
Onl
y 2
Onl
y 2 --
3 up
date
s pe
r sec
ond
3 up
date
s pe
r sec
ond
from
sys
tem
from
sys
tem
..––
Iner
tial t
rack
er c
an g
ive
100H
z.In
ertia
l tra
cker
can
giv
e 10
0Hz.
––Fu
ses
iner
tial t
rack
erFu
ses
iner
tial t
rack
er——
and
all u
pdat
esan
d al
l upd
ates
——us
ing
SLE
RP
.us
ing
SLE
RP
.P
DA
P
DA
““ Bat
Por
tal
Bat
Por
tal ””
––O
ne B
at o
n C
ompa
q O
ne B
at o
n C
ompa
q iP
AQ
iPA
Q..
––A
noth
er B
at a
roun
d ne
ck.
Ano
ther
Bat
aro
und
neck
.––
Hea
dH
ead --
toto-- iP
AQ
iPA
Qra
y dr
ives
dis
play
.ra
y dr
ives
dis
play
.
““ Con
stel
latio
n” (F
oxlin
199
8)C
onst
ella
tion”
(Fox
lin 1
998)
Idea
beh
ind
ISId
ea b
ehin
d IS
-- 600
and
IS60
0 an
d IS
-- 900
.90
0.E
xten
ds in
ertia
l tra
cker
(199
6) w
ith
Ext
ends
iner
tial t
rack
er (1
996)
with
ul
traso
nic
ultra
soni
c la
tera
tion
late
ratio
n ..Tr
ansp
onde
r ultr
ason
ic b
eaco
ns a
t Tr
ansp
onde
r ultr
ason
ic b
eaco
ns a
t m
odel
ed p
oses
(typ
ical
ly c
eilin
g).
mod
eled
pos
es (t
ypic
ally
cei
ling)
.U
ltras
onic
Rec
eive
r Mod
ules
U
ltras
onic
Rec
eive
r Mod
ules
(( U
RM
sU
RM
s ) w
orn
by m
obile
use
r) w
orn
by m
obile
use
r––
Exc
ite c
eilin
g tra
nspo
nder
s w
ith u
niqu
e E
xcite
cei
ling
trans
pond
ers
with
uni
que
IR c
odes
.IR
cod
es.
––Ti
me
trans
pond
erTi
me
trans
pond
er’’ s
chi
rp
s ch
irp T
OFs
TOFs
for
for
rang
es.
rang
es.
––ar
rival
s at
mob
ile
arriv
als
at m
obile
0
late
ncy.
0 la
tenc
y.
““ Con
stel
latio
n” (F
oxlin
199
8)C
onst
ella
tion”
(Fox
lin 1
998)
SC
AA
T E
KF
fuse
s in
ertia
l inp
uts
with
rang
esS
CA
AT
EK
F fu
ses
iner
tial i
nput
s w
ith ra
nges
––P
redi
cts
each
rang
e.P
redi
cts
each
rang
e.––
Rec
eive
r onl
y ga
ted
Rec
eive
r onl
y ga
ted
““ ope
nop
en”” a
roun
d pr
edic
tion.
arou
nd p
redi
ctio
n.––
Thro
ws
out s
uspi
ciou
s de
viat
ions
.Th
row
s ou
t sus
pici
ous
devi
atio
ns.
––sc
alar
sc
alar
rang
era
nge
nonom
atrix
inve
rsio
n.m
atrix
inve
rsio
n.E
rror s
ourc
es:
Erro
r sou
rces
:––
Bea
cons
Bea
cons
’’ pos
ition
s po
sitio
ns m
ism
is-- m
odel
ed.
mod
eled
.––
Air
tem
pera
ture
alte
rs s
ound
pro
paga
tion
vs.
Air
tem
pera
ture
alte
rs s
ound
pro
paga
tion
vs.
mod
el.
mod
el.
––C
onst
ant
Con
stan
t -- tim
e de
lays
in b
eaco
ns/
time
dela
ys in
bea
cons
/ UR
Ms
UR
Ms
(par
t(p
art --
toto-- p
art v
aria
tions
).pa
rt va
riatio
ns).
––Tr
ansd
ucer
s in
bea
cons
/Tr
ansd
ucer
s in
bea
cons
/ UR
Ms
UR
Ms
off a
xis.
off a
xis.
Sim
ulat
ion
accu
racy
(2m
bel
ow):
S
imul
atio
n ac
cura
cy (2
m b
elow
): 2
.5m
m;
2.5m
m;
mor
e lik
e 1.
5mm
mor
e lik
e 1.
5mm
with
cor
rect
ions
for
with
cor
rect
ions
for
syst
emic
sou
rces
abo
ve.
syst
emic
sou
rces
abo
ve.
Aco
ustic
Tra
ckin
g S
umm
ary
Aco
ustic
Tra
ckin
g S
umm
ary
Adv
anta
ge:
An
abso
lute
refe
renc
e w
ith o
utsi
de
Adv
anta
ge:
An
abso
lute
refe
renc
e w
ith o
utsi
de
wor
ld, w
ithou
t phy
sica
l con
nect
ion
to it
.w
orld
, with
out p
hysi
cal c
onne
ctio
n to
it.
Req
uire
men
ts:
Req
uire
men
ts:
––S
omet
hing
to s
how
whe
n so
und
left
sour
ce.
Som
ethi
ng to
sho
w w
hen
soun
d le
ft so
urce
.InIn
-- Out
: Mob
ile B
ats
chirp
upo
n R
F po
ll (m
ore
late
ncy)
.O
ut: M
obile
Bat
s ch
irp u
pon
RF
poll
(mor
e la
tenc
y).
Out
Out
-- In, o
r tim
e re
fere
nce
(less
late
ncy)
:In
, or t
ime
refe
renc
e (le
ss la
tenc
y):
––C
onst
ella
tion:
sta
tiona
ry b
eaco
ns c
hirp
upo
n m
obile
IR p
oll.
Con
stel
latio
n: s
tatio
nary
bea
cons
chi
rp u
pon
mob
ile IR
pol
l. ––
Ran
dell’
sR
ande
ll’s
low
low
-- cos
t sys
tem
had
4 s
tatio
nary
bea
cons
chi
rp a
t co
st s
yste
m h
ad 4
sta
tiona
ry b
eaco
ns c
hirp
at
equa
l int
erva
ls a
fter R
F “c
lock
tick
s.”
equa
l int
erva
ls a
fter R
F “c
lock
tick
s.”
––C
ricke
t chi
rps
and
emits
RF
sim
ulta
neou
sly.
Cric
ket c
hirp
s an
d em
its R
F si
mul
tane
ousl
y.
––A
line
of s
ight
from
bea
con
to s
enso
r.A
line
of s
ight
from
bea
con
to s
enso
r.––
No
jang
ling
keys
! ;
No
jang
ling
keys
! ; -- ))
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infra
red
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
Indo
or M
agne
tic T
rack
ing
(Raa
b 19
79)
Indo
or M
agne
tic T
rack
ing
(Raa
b 19
79)
Impo
sed
mag
netic
fiel
ds:
Impo
sed
mag
netic
fiel
ds:
––S
eque
ntia
lly.
Seq
uent
ially
.––
Aro
und
sour
ce a
xes
Aro
und
sour
ce a
xes
xx , , yy
and
and
zz ..––
One
tim
eO
ne ti
me --
slot
un
slot
un --
impo
sed,
for r
oom
impo
sed,
for r
oom
’’ s s
ambi
ent f
ield
.am
bien
t fie
ld.
Eac
h fie
ld e
voke
s cu
rren
t in
the
3E
ach
field
evo
kes
curr
ent i
n th
e 3 --
coil
coil
sens
ors.
sens
ors.
Cur
rent
s m
odel
ed a
s se
ries
of fr
ame
Cur
rent
s m
odel
ed a
s se
ries
of fr
ame
rota
tions
on
a ra
nge
rota
tions
on
a ra
nge --
atte
nuat
ed
atte
nuat
ed
dipo
le v
ecto
r.di
pole
vec
tor.
3 co
il cu
rren
ts (
3 co
il cu
rren
ts (
-- am
bien
ce) X
3
ambi
ence
) X 3
fie
lds,
sol
ves
6DO
F se
nsor
pos
e.fie
lds,
sol
ves
6DO
F se
nsor
pos
e.S
olut
ion
redu
ces
to 9
equ
atio
ns, 6
S
olut
ion
redu
ces
to 9
equ
atio
ns, 6
un
know
ns, u
sing
:un
know
ns, u
sing
:––
Last
est
imat
e an
d de
ltas
pres
umed
sm
all.
Last
est
imat
e an
d de
ltas
pres
umed
sm
all.
––S
mal
lS
mal
l -- ang
le th
eore
m a
ppro
xim
atio
ns.
angl
e th
eore
m a
ppro
xim
atio
ns.
ζπρ
ρcos
23
MH
=
ζπρ
sin
43
MHt=
Indo
or M
agne
tic T
rack
ing
(Raa
b 19
79)
Indo
or M
agne
tic T
rack
ing
(Raa
b 19
79)
Hug
e S
TRE
NG
TH:
Una
ffect
ed b
y ph
ysic
al o
cclu
sion
s lik
e th
e H
uge
STR
EN
GTH
: U
naffe
cted
by
phys
ical
occ
lusi
ons
like
the
hum
an b
ody!
No
line
of s
ight
requ
irem
ent!
hum
an b
ody!
No
line
of s
ight
requ
irem
ent!
WE
AK
NE
SS
ES
:W
EA
KN
ES
SE
S:
––Fi
eld
stre
ngth
dim
inis
hes
with
the
CU
BE
of r
ange
Fiel
d st
reng
th d
imin
ishe
s w
ith th
e C
UB
E o
f ran
ge——
thes
e te
nd n
ot to
th
ese
tend
not
to
wor
k be
yond
10
wor
k be
yond
10 ’’
, if t
hat.
, if t
hat.
––GG
radi
ent
radi
ent o
f fie
ld s
treng
th d
imin
ishe
s w
ith th
e FO
UR
TH p
ower
of r
ange
of fi
eld
stre
ngth
dim
inis
hes
with
the
FOU
RTH
pow
er o
f ran
gere
solu
tion
/ gra
nula
rity
dim
inis
hes
as a
resu
lt: a
1cm
mot
ion
nre
solu
tion
/ gra
nula
rity
dim
inis
hes
as a
resu
lt: a
1cm
mot
ion
n ear
the
ear t
he
sour
ce h
as m
uch
mor
e ra
tio e
ffect
than
the
sam
e m
otio
n fu
rther
aso
urce
has
muc
h m
ore
ratio
effe
ct th
an th
e sa
me
mot
ion
furth
er a
way
.w
ay .
A/D
con
verte
rs o
nly
have
so
man
y bi
ts .
. . .
A/D
con
verte
rs o
nly
have
so
man
y bi
ts .
. . .
––M
etal
or o
ther
non
Met
al o
r oth
er n
on-- e
arth
mag
netic
sou
rces
intro
duce
dis
torti
ons.
DC
ea
rth m
agne
tic s
ourc
es in
trodu
ce d
isto
rtion
s. D
C
vers
ions
hel
p in
som
e ca
ses,
but
not
in o
ther
s.ve
rsio
ns h
elp
in s
ome
case
s, b
ut n
ot in
oth
ers.
Am
usin
g C
AV
EA
T: S
ymm
etry
of f
ield
s cr
eate
s a
hem
isph
eric
al
Am
usin
g C
AV
EA
T: S
ymm
etry
of f
ield
s cr
eate
s a
hem
isph
eric
al
ambi
guity
ambi
guity
——an
othe
r pos
e, e
xact
ly o
ppos
ite th
e cu
rrent
one
anot
her p
ose,
exa
ctly
opp
osite
the
curre
nt o
ne——
whe
re th
e se
nsor
wou
ld h
ave
iden
tical
read
ings
.w
here
the
sens
or w
ould
hav
e id
entic
al re
adin
gs.
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infr
ared
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
Act
ive
Bad
ge (W
ant 1
992)
Act
ive
Bad
ge (W
ant 1
992)
Bad
ges
emit
IR e
very
15s
Bad
ges
emit
IR e
very
15s
––en
codi
ng a
uni
que
IDen
codi
ng a
uni
que
ID––
unsy
nchr
oniz
edun
sync
hron
ized
——ch
eap
chea
p co
mpo
nent
sco
mpo
nent
scl
ock
cloc
kdr
ift
drift
an
d fe
w c
ollis
ions
.an
d fe
w c
ollis
ions
.C
usto
m, i
nexp
ensi
ve n
etw
ork
Cus
tom
, ine
xpen
sive
net
wor
k––
cont
rolle
r can
pol
l up
to 1
28 s
enso
rsco
ntro
ller c
an p
oll u
p to
128
sen
sors
––co
ntro
llers
can
be
aggr
egat
ed th
roug
h co
mpu
ter
cont
rolle
rs c
an b
e ag
greg
ated
thro
ugh
com
pute
r ne
twor
ks fo
r sca
lene
twor
ks fo
r sca
le-- u
p.up.
Sta
tiona
ry s
enso
rs b
uffe
r up
to 2
0 m
ost
Sta
tiona
ry s
enso
rs b
uffe
r up
to 2
0 m
ost
rece
nt
rece
nt ““ s
ight
ings
sigh
tings
”” bet
wee
n co
ntro
ller p
olls
.be
twee
n co
ntro
ller p
olls
.S
yste
m in
fers
coa
rse
Sys
tem
infe
rs c
oars
e pr
oxim
itypr
oxim
ityfro
m B
adge
fro
m B
adge
““ h
its,
hits
, ””an
d ca
ches
rece
nt h
isto
ry.
and
cach
es re
cent
his
tory
.
Act
ive
Bad
ge (W
ant 1
992)
Act
ive
Bad
ge (W
ant 1
992)
Rec
eptio
nist
imm
edia
tely
love
d sy
stem
.R
ecep
tioni
st im
med
iate
ly lo
ved
syst
em.
Sta
ff gr
udgi
ngly
trie
d it
2 w
eeks
then
vol
unta
rily
Sta
ff gr
udgi
ngly
trie
d it
2 w
eeks
then
vol
unta
rily
cont
inue
dco
ntin
ued ——
rais
e pr
ivac
y co
ncer
ns!
rais
e pr
ivac
y co
ncer
ns!
At l
east
thre
e m
ore
inst
alla
tions
by
the
time
of th
e pa
per.
At l
east
thre
e m
ore
inst
alla
tions
by
the
time
of th
e pa
per.
Gre
en:
“Inte
rsec
tions
” Fr
agm
ent
Red
: “In
ters
ectio
ns
and
Sub
tract
ions
” Fr
agm
ent
Blu
e: “
Elli
pse
of
Con
fiden
ce”
arou
nd e
stim
ate
Coa
rse,
Inex
pens
ive
IR (H
alla
way
200
3)C
oars
e, In
expe
nsiv
e IR
(Hal
law
ay 2
003)
Red
: IR
trac
ker
Red
: IR
trac
ker
Blu
e: In
terS
ense
ISB
lue:
Inte
rSen
se IS
-- 900
“gro
und
truth
”90
0 “g
roun
d tru
th”
3m
4.5m
Ca.
1 m
pos
ition
acc
urac
y
Ca.
1 s
tem
pora
l lag
Coa
rse,
Inex
pens
ive
IR (H
alla
way
200
3)C
oars
e, In
expe
nsiv
e IR
(Hal
law
ay 2
003)
IR B
eaco
n S
umm
ary
IR B
eaco
n S
umm
ary
Wan
tW
ant
––S
enso
rs o
utsi
de:
Sen
sors
out
side
:S
yste
m tr
acks
use
rs.
Sys
tem
trac
ks u
sers
.U
sers
initi
ally
U
sers
initi
ally
wor
ry a
bout
wor
ry a
bout
priv
acy.
pr
ivac
y.
––Ta
rget
Ta
rget
pro
xim
itypr
oxim
ity(w
hich
room
)(w
hich
room
) ..
Hal
law
ay
Hal
law
ay
––S
enso
rs in
side
:S
enso
rs in
side
:M
obile
use
r tra
cks
him
self.
Mob
ile u
ser t
rack
s hi
mse
lf.C
ould
C
ould
pro
tect
prot
ect p
rivac
y.pr
ivac
y.
––C
oars
e C
oars
e C
arte
sian
Car
tesi
anta
rget
pos
ition
(2D
) and
orie
ntat
ion
targ
et p
ositi
on (2
D) a
nd o
rient
atio
n (1
D).
(1D
).
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infr
ared
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
Con
stel
latio
nC
onst
ella
tion3
Di
3Di(H
edge
s 20
01)
(Hed
ges
2001
)
Tran
smitt
ers
spin
~40
rps
Tran
smitt
ers
spin
~40
rps
––S
wee
ping
vol
ume
with
two
angl
ed,
Sw
eepi
ng v
olum
e w
ith tw
o an
gled
, pl
anar
fans
of I
R la
ser
plan
ar fa
ns o
f IR
lase
r––
Em
ittin
g an
E
mitt
ing
an o
mni
dire
ctio
nal
omni
dire
ctio
nal s
trobe
st
robe
tim
ing
refe
renc
e.tim
ing
refe
renc
e.R
ecei
vers
:R
ecei
vers
:––
Sen
sors
are
IRS
enso
rs a
re IR
-- filte
red
diod
es th
at
filte
red
diod
es th
at
“see
” the
fan
and
stro
be “h
its.”
“see
” the
fan
and
stro
be “h
its.”
––B
oard
firm
war
e so
rts “h
its”
Boa
rd fi
rmw
are
sorts
“hits
”W
hat t
rans
mitt
er.
Wha
t tra
nsm
itter
.W
hich
fan
or s
trobe
.W
hich
fan
or s
trobe
.Te
mpo
ral b
asis
.Te
mpo
ral b
asis
.––
Boa
rd o
utpu
ts tr
ansm
itter
Boa
rd o
utpu
ts tr
ansm
itter
-- fram
e fra
me
angl
e pa
ir at
whi
ch fa
n hi
ts o
ccur
red.
angl
e pa
ir at
whi
ch fa
n hi
ts o
ccur
red.
Con
stel
latio
nC
onst
ella
tion3
Di
3Di(H
edge
s 20
01)
(Hed
ges
2001
)
Sof
twar
e:S
oftw
are:
––In
itial
ly c
alib
rate
s tra
nsm
itter
sIn
itial
ly c
alib
rate
s tra
nsm
itter
sIn
to g
loba
l ref
eren
ce fr
ame.
Into
glo
bal r
efer
ence
fram
e.––
Dur
ing
posi
tioni
ng, c
alcu
late
s ra
ys
Dur
ing
posi
tioni
ng, c
alcu
late
s ra
ys
alon
g w
hich
al
ong
whi
ch s
enso
r(s
sens
or(s
) mus
t lie
) mus
t lie
Usi
ng k
now
n tra
nsm
itter
par
amet
ers
Usi
ng k
now
n tra
nsm
itter
par
amet
ers -
---fa
n fa
n an
gles
, etc
.an
gles
, etc
.––
Res
olve
s bu
ndle
s of
rays
into
R
esol
ves
bund
les
of ra
ys in
to
clos
est p
ositi
on in
glo
bal f
ram
e.cl
oses
t pos
ition
in g
loba
l fra
me.
Per
form
ance
: P
erfo
rman
ce:
mill
imet
erm
illim
eter
-- leve
l le
vel
posi
tiona
l acc
urac
ypo
sitio
nal a
ccur
acy ——
in th
e st
atic
in
the
stat
ic
mod
el,
mod
el, a
vera
ging
out
noi
se o
ver 2
av
erag
ing
out n
oise
ove
r 2
seco
nds.
seco
nds.
SA
RS
DT
(S
AR
SD
T ( C
hint
ham
mit
Chi
ntha
mm
it20
03)
2003
)
Virt
ual R
etin
al D
ispl
ay
Virt
ual R
etin
al D
ispl
ay
shar
es “a
pertu
re” w
ith
shar
es “a
pertu
re” w
ith
track
er:
track
er:
––C
onst
ant I
R “d
ot” a
dded
to
Con
stan
t IR
“dot
” add
ed to
vi
sibl
e sc
an.
visi
ble
scan
.––
Bea
m s
plitt
er re
flect
s IR
B
eam
spl
itter
refle
cts
IR
com
pone
nt o
ut, a
nd v
isib
le
com
pone
nt o
ut, a
nd v
isib
le
light
into
the
eye.
light
into
the
eye.
––IR
is a
lway
s co
lline
ar w
ith
IR is
alw
ays
colli
near
with
vi
sibl
e ra
y sc
anni
ng th
e vi
sibl
e ra
y sc
anni
ng th
e re
tina.
retin
a.––
Sen
sors
in e
nviro
nmen
t.S
enso
rs in
env
ironm
ent.
Stru
ctur
ed L
ight
Sum
mar
yS
truct
ured
Lig
ht S
umm
ary
Con
stel
latio
nC
onst
ella
tion3
Di
3Di
––S
enso
rs “i
nsid
e” o
n tra
cked
use
r:S
enso
rs “i
nsid
e” o
n tra
cked
use
r:C
an b
e un
teth
ered
for m
obilit
y.C
an b
e un
teth
ered
for m
obilit
y.C
an p
rote
ct p
rivac
y of
the
track
ed.
Can
pro
tect
priv
acy
of th
e tra
cked
.––
Vol
ume
of tr
acki
ng is
larg
e an
d co
nsta
nt, g
iven
set
up.
Vol
ume
of tr
acki
ng is
larg
e an
d co
nsta
nt, g
iven
set
up.
SA
RS
DT
SA
RS
DT
––S
enso
rs “o
utsi
de” i
n th
e en
viro
nmen
t:S
enso
rs “o
utsi
de” i
n th
e en
viro
nmen
t:D
ispl
ay m
ust b
e te
ther
ed to
sys
tem
for t
ight
syn
chro
niza
tion.
Dis
play
mus
t be
teth
ered
to s
yste
m fo
r tig
ht s
ynch
roni
zatio
n.––
Vol
ume
whe
re tr
acki
ng o
ccur
s at
any
giv
en m
omen
t V
olum
e w
here
trac
king
occ
urs
at a
ny g
iven
mom
ent
is th
e 16
.5x6
.5is
the
16.5
x6.5
ººFO
V o
f the
VR
D!
FOV
of t
he V
RD
!Fo
r dec
ent c
over
age,
wal
ls w
ould
hav
e to
be
For d
ecen
t cov
erag
e, w
alls
wou
ld h
ave
to b
e ““ c
arpe
ted
carp
eted
”” with
w
ith
sens
ors.
sens
ors.
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infr
ared
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
HiB
all (
Wel
ch, B
isho
p, e
t al.,
199
9)H
iBal
l (W
elch
, Bis
hop,
et a
l., 1
999)
Dod
ecah
edra
l dev
ice
Dod
ecah
edra
l dev
ice
––6
heav
ily IR
6 he
avily
IR-- fi
ltere
d, fi
xed
filte
red,
fixe
d --fo
cus
lens
es.
focu
s le
nses
.––
6 6 LE
PD
sLE
PD
s ——fa
st!
fast
!
Cei
ling
inst
rum
ente
d w
ith
Cei
ling
inst
rum
ente
d w
ith
IR L
ED
s.IR
LE
Ds.
2kH
z up
date
s; <
1ms
2kH
z up
date
s; <
1ms
late
ncy;
<0.
2mm
and
0.0
2 la
tenc
y; <
0.2m
m a
nd 0
.02
degr
ees
nois
ede
gree
s no
ise ..
HiB
all (
Wel
ch, B
isho
p, e
t al.,
199
9)H
iBal
l (W
elch
, Bis
hop,
et a
l., 1
999)
Est
imat
ion
cycl
e:E
stim
atio
n cy
cle:
––R
ando
m c
hoic
e of
one
of 2
6 R
ando
m c
hoic
e of
one
of 2
6 FO
Vs
FOV
s ..––
Flas
h le
ast
Flas
h le
ast -- r
ecen
tly u
sed
LED
in th
e ce
iling
re
cent
ly u
sed
LED
in th
e ce
iling
pr
ojec
tion
of it
s LE
PD
.pr
ojec
tion
of it
s LE
PD
.
SC
AA
T E
KF
SC
AA
T E
KF
––S
epar
ate
KF
stat
e fo
r eac
h LE
D a
ppen
ded
to
Sep
arat
e K
F st
ate
for e
ach
LED
app
ende
d to
tra
ckin
g K
F.tra
ckin
g K
F.––
Pre
dict
imag
eP
redi
ct im
age --
plan
e pr
ojec
tion
of L
ED
.pl
ane
proj
ectio
n of
LE
D.
––A
pply
gai
n to
resi
dual
of m
easu
rem
ent.
App
ly g
ain
to re
sidu
al o
f mea
sure
men
t.––
Upd
ate
LED
cal
ibra
tion
alon
g w
ith tr
acki
ng
Upd
ate
LED
cal
ibra
tion
alon
g w
ith tr
acki
ng
estim
ate.
estim
ate.
––S
ave
LED
cor
rect
ions
to it
s ow
n fil
ter.
Sav
e LE
D c
orre
ctio
ns to
its
own
filte
r.
VIS
VIS
-- Tra
cker
(Fox
lin 2
003)
Trac
ker (
Foxl
in 2
003)
VIS
VIS
-- Tra
cker
ext
ends
iner
tial s
olut
ion
Trac
ker e
xten
ds in
ertia
l sol
utio
n (1
993)
with
(1
993)
with
fidu
cial
fiduc
ial -- b
ased
vis
ion
base
d vi
sion
tra
ckin
g.tra
ckin
g.In
ertia
Cub
eIn
ertia
Cub
ew
ith 1
/3” C
CD
.w
ith 1
/3” C
CD
.C
CD
exp
osur
e an
d tra
nsfe
r tak
es
CC
D e
xpos
ure
and
trans
fer t
akes
40
+ms,
but
allo
ws
for a
bsol
ute
40+m
s, b
ut a
llow
s fo
r abs
olut
e co
rrec
tion.
corr
ectio
n.In
ertia
Cub
eIn
ertia
Cub
eup
date
s ev
ery
8ms,
up
date
s ev
ery
8ms,
ba
sed
on 1
920H
z sa
mpl
ing.
base
d on
192
0Hz
sam
plin
g.A
ccur
acy
on th
e or
der o
f 1cm
Acc
urac
y on
the
orde
r of 1
cmdu
ring
durin
g w
alki
ng.
wal
king
.
Out
door
Orie
ntat
ion
(You
199
9)O
utdo
or O
rient
atio
n (Y
ou 1
999)
Cal
ibra
ting,
con
trolli
ng o
r C
alib
ratin
g, c
ontro
lling
or
mod
ifyin
g th
e ou
tdoo
rs is
m
odify
ing
the
outd
oors
is
unre
alis
tic.
unre
alis
tic.
Out
door
vie
ws
are
ofte
n at
O
utdo
or v
iew
s ar
e of
ten
at
long
rang
es, a
ffect
ed:
long
rang
es, a
ffect
ed:
––gr
eatly
by
orie
ntat
ion.
grea
tly b
y or
ient
atio
n.––
little
by
smal
l cha
nges
in
little
by
smal
l cha
nges
in
posi
tion.
posi
tion.
Out
door
orie
ntat
ion
track
ing
Out
door
orie
ntat
ion
track
ing
in u
npre
pare
d en
viro
nmen
t?in
unp
repa
red
envi
ronm
ent?
––3
rate
gyr
os3
rate
gyr
os——
sam
pled
at
sam
pled
at
1kH
z.1k
Hz.
––C
ompa
ss a
nd ti
lt se
nsor
.C
ompa
ss a
nd ti
lt se
nsor
.––
Son
y ca
mer
a.S
ony
cam
era.
Out
door
Orie
ntat
ion
(You
199
9)O
utdo
or O
rient
atio
n (Y
ou 1
999)
Com
pass
fuse
d w
ith g
yros
:C
ompa
ss fu
sed
with
gyr
os:
––16
Hz
vs. 1
kHz
upda
tes
16H
z vs
. 1kH
z up
date
s co
mpa
ss c
orre
ctio
ns 9
2ms
into
pas
t.co
mpa
ss c
orre
ctio
ns 9
2ms
into
pas
t.C
amer
aC
amer
a --in
ertia
l fus
ion:
iner
tial f
usio
n:––
Cam
era
mai
ntai
ns s
et o
f wor
ld p
oint
s / f
eatu
res.
Cam
era
mai
ntai
ns s
et o
f wor
ld p
oint
s / f
eatu
res.
––In
ter
Inte
r --fra
me
pixe
l vel
ociti
es c
orre
ct in
ertia
l sys
tem
, bas
ed o
n fra
me
pixe
l vel
ociti
es c
orre
ct in
ertia
l sys
tem
, bas
ed o
n in
ertia
llyin
ertia
llypr
ojec
ted
pixe
l vel
ocity
:pr
ojec
ted
pixe
l vel
ocity
:)
(1C u
I ux
xA
−=
∆−
ωR
egis
tratio
n er
ror a
vera
ged
4.27
pi
xels
ove
r seq
uenc
e, o
r ~0.
4º!
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Iner
tial &
Gra
vim
etric
Saw
ada
2003
Mec
hani
cal
Add
lese
e19
97
Foxl
in 2
003
Vis
ible
Infr
ared
You
199
9W
ant 1
992
H
alla
way
200
3W
elch
199
9
H
edge
s 20
01C
hint
ham
mit
2003
Aco
ustic
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8N
ewm
an 2
001
Elec
trom
agne
tic S
pect
rum
Rad
io F
requ
ency
(RF)
Hiro
se 2
003
Bah
l 200
0
Enge
1999
Wan
g 20
00
Dju
knic
2001
Solim
an20
00
Ligh
t Spe
ctru
m
Raa
b 19
79M
agne
tic F
ield
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Gol
ding
1999
(+te
mp.
)
RFI
D fo
r Dea
fR
FID
for D
eaf -- B
lind
(Hiro
se 2
003)
Blin
d (H
irose
200
3)
Pro
blem
: Le
adin
g an
d P
robl
em:
Lead
ing
and
track
ing
the
deaf
track
ing
the
deaf
-- blin
d.bl
ind.
1349
act
ive,
bat
tery
1349
act
ive,
bat
tery
--po
wer
ed R
FID
tags
: po
wer
ed R
FID
tags
: ––
embe
dded
in 1
700m
embe
dded
in 1
700m
22flo
or.
floor
.––
emitt
ing
uniq
ue ID
s at
5H
z.em
ittin
g un
ique
IDs
at 5
Hz.
––~1
.2m
rang
e to
ant
enna
e.~1
.2m
rang
e to
ant
enna
e.
Wor
n an
tenn
a ve
st.
Wor
n an
tenn
a ve
st.
RFI
D fo
r Dea
fR
FID
for D
eaf -- B
lind
(Hiro
se 2
003)
Blin
d (H
irose
200
3)
N n
ew re
adin
gs
N n
ew re
adin
gs
σσ==
~0.4
m, r
ecor
ding
~0
.4m
, rec
ordi
ng
posi
tions
at 1
0Hz.
posi
tions
at 1
0Hz.
+
−+
=∑ =n i
iut
xn
tx
1)1
(11
)(
““ RA
DA
R”
RA
DA
R” W
iW
i -- FiFiLo
catio
n (B
ahl 2
000)
Loca
tion
(Bah
l 200
0)
Wire
less
NIC
driv
ers
expo
se:
Wire
less
NIC
driv
ers
expo
se:
––S
igna
l stre
ngth
(SS
)S
igna
l stre
ngth
(SS
) ——co
rrel
ated
wel
l co
rrel
ated
wel
l w
ith ra
nge.
with
rang
e.––
Sig
nal
Sig
nal -- t
oto-- n
oise
ratio
n (S
NR
)no
ise
ratio
n (S
NR
) ——di
d no
t.di
d no
t.M
ultip
le b
ase
stat
ions
M
ultip
le b
ase
stat
ions
tri
late
ratio
ntri
late
ratio
non
on
ra
nges
.ra
nges
.E
mpi
rical
stra
tegy
:E
mpi
rical
stra
tegy
:––
Sam
ple
envi
ronm
ent,
build
ing
tabl
e of
S
ampl
e en
viro
nmen
t, bu
ildin
g ta
ble
of
(mea
n)
(mea
n) tu
ples
tupl
es: p
ositi
on, d
irect
ion,
sta
tion,
: p
ositi
on, d
irect
ion,
sta
tion,
S
S, S
NR
.S
S, S
NR
.––
Trac
k by
usi
ng ta
ble
Trac
k by
usi
ng ta
ble
mat
ch(e
sm
atch
(es )
clo
sest
in
) clo
sest
in
SS
spa
ce.
SS
spa
ce.
––G
ot b
est r
esul
ts [
Got
bes
t res
ults
[ med
ian
erro
r m
edia
n er
ror 2
.5m
2.5m
] ] us
ing
2us
ing
2 --4
mul
tiple
4 m
ultip
le-- n
eare
stne
ares
t -- nei
ghbo
r ne
ighb
or
tupl
estu
ples
, and
the
max
imum
of t
he fo
ur
, and
the
max
imum
of t
he fo
ur
dire
ctio
ns’ S
S (v
s. s
ingl
e ne
ares
t di
rect
ions
’ SS
(vs.
sin
gle
near
est
neig
hbor
and
wor
stne
ighb
or a
nd w
orst
-- cas
e di
rect
ion)
.ca
se d
irect
ion)
.
““ RA
DA
R”
RA
DA
R” W
iW
i -- FiFiLo
catio
n (B
ahl 2
000)
Loca
tion
(Bah
l 200
0)R
adio
Pro
paga
tion
Mod
el:
Rad
io P
ropa
gatio
n M
odel
:
––Te
sts
foun
d C
= 4
, WA
F =
Test
s fo
und
C =
4, W
AF
= 3.
1dB
m, a
nd
3.1d
Bm
, and
––
Med
ian
resu
lts w
ere
not t
hat
Med
ian
resu
lts w
ere
not t
hat
muc
h w
orse
than
em
piric
al,
muc
h w
orse
than
em
piric
al,
4.3m
med
ian
erro
r.4.
3m m
edia
n er
ror.
––S
aves
hug
ely
on
Sav
es h
ugel
y on
pr
epar
atio
n, if
you
hav
e a
prep
arat
ion,
if y
ou h
ave
a m
ap a
nd a
n al
gorit
hm to
m
ap a
nd a
n al
gorit
hm to
co
mpu
te w
all i
nter
sect
ions
. co
mpu
te w
all i
nter
sect
ions
.
≥<
−
−=
CnW
WAF
CC
nWWAF
nWdd
nd
Pd
P:
*:
*log
10)
()
(0
0
Glo
bal P
ositi
onin
g S
yste
m (
Glo
bal P
ositi
onin
g S
yste
m ( E
nge
Eng
e19
99)
1999
)
Sim
ple
Sim
ple
Pos
ition
Pos
ition
::––
Dat
a m
essa
ge in
clud
es s
atel
lite
time.
Dat
a m
essa
ge in
clud
es s
atel
lite
time.
––4+
sat
ellit
es4+
sat
ellit
es4+
equ
atio
ns, 3
unk
now
n 4+
equ
atio
ns, 3
unk
now
n ra
nges
, 1 u
nkno
wn
cloc
k of
fset
. S
olve
d ra
nges
, 1 u
nkno
wn
cloc
k of
fset
. S
olve
d cl
ock
offs
et g
ives
app
roxi
mat
e tra
nsit
cloc
k of
fset
giv
es a
ppro
xim
ate
trans
it tim
e.tim
e.––
Rec
eive
r kno
ws
pseu
dora
ndom
noi
se
Rec
eive
r kno
ws
pseu
dora
ndom
noi
se
(PR
N) C
oars
e/A
cqui
sitio
n (C
/A)
(PR
N) C
oars
e/A
cqui
sitio
n (C
/A)
spre
adsp
read
-- spe
ctru
m c
ode
mod
ulat
ed o
n sp
ectru
m c
ode
mod
ulat
ed o
n si
gnal
.si
gnal
.––
Aut
ocor
rela
tion
of it
s ow
n C
/A c
opy
has
Aut
ocor
rela
tion
of it
s ow
n C
/A c
opy
has
a sh
arp
peak
.a
shar
p pe
ak.
Ref
ines
tran
sit t
ime
mod
1m
s.R
efin
es tr
ansi
t tim
e m
od 1
ms.
Idea
lly y
ield
s m
eter
Idea
lly y
ield
s m
eter
-- leve
l ran
ging
. le
vel r
angi
ng.
Glo
bal P
ositi
onin
g S
yste
m (
Glo
bal P
ositi
onin
g S
yste
m ( E
nge
Eng
e19
99)
1999
)
Sim
ple
Sim
ple
Vel
ocity
Vel
ocity
::––
Ano
ther
loop
mon
itors
car
rier p
hase
and
Dop
pler
A
noth
er lo
op m
onito
rs c
arrie
r pha
se a
nd D
oppl
er
shift
.sh
ift.
––D
oppl
er s
hift
is s
atel
lite
Dop
pler
shi
ft is
sat
ellit
e --re
ceiv
er v
eloc
ity p
roje
cted
on
rece
iver
vel
ocity
pro
ject
ed o
n lin
e of
sig
ht.
line
of s
ight
.––
Four
suc
h ve
loci
ties,
and
kno
wn
sate
llite
pos
ition
s,
Four
suc
h ve
loci
ties,
and
kno
wn
sate
llite
pos
ition
s,
allo
w g
roun
d ve
loci
ty c
alcu
latio
n.al
low
gro
und
velo
city
cal
cula
tion.
Pre
cise
UTC
P
reci
se U
TC ti
me
time
is ju
st th
e bi
asis
just
the
bias
-- cor
rect
ed
corr
ecte
d re
ceiv
er ti
me
plus
leap
rece
iver
tim
e pl
us le
ap-- s
econ
ds th
at
seco
nds
that
di
ffere
ntia
te G
PS
tim
e fro
m U
TC.
diffe
rent
iate
GP
S ti
me
from
UTC
.
Glo
bal P
ositi
onin
g S
yste
m (
Glo
bal P
ositi
onin
g S
yste
m ( E
nge
Eng
e19
99)
1999
)
GP
S p
reci
sion
dep
ends
on:
GP
S p
reci
sion
dep
ends
on:
––G
eom
etry
(GD
OP
)G
eom
etry
(GD
OP
)Th
e m
ore
orth
ogon
al th
e ra
nges
use
d, th
e m
ore
accu
rate
Th
e m
ore
orth
ogon
al th
e ra
nges
use
d, th
e m
ore
accu
rate
th
e fix
.th
e fix
.
––N
umbe
r of s
atel
lite
rang
es a
vaila
ble.
Num
ber o
f sat
ellit
e ra
nges
ava
ilabl
e.––
Iono
sphe
ricIo
nosp
heric
and
and
tropo
sphe
rictro
posp
heric
dist
ortio
ns o
f the
di
stor
tions
of t
he
velo
city
of l
ight
.ve
loci
ty o
f lig
ht.
Glo
bal P
ositi
onin
g S
yste
m (
Glo
bal P
ositi
onin
g S
yste
m ( E
nge
Eng
e19
99)
1999
)
Aug
men
tatio
ns to
GP
S re
duce
err
or fu
rther
.A
ugm
enta
tions
to G
PS
redu
ce e
rror
furth
er.
––D
iffer
entia
l (D
GP
S):
land
Diff
eren
tial (
DG
PS
): la
nd-- b
ased
radi
o st
ream
of e
rror
ba
sed
radi
o st
ream
of e
rror
co
rrec
tions
giv
es ~
1m a
ccur
acy.
corr
ectio
ns g
ives
~1m
acc
urac
y.––
Rea
lR
eal -- T
ime
Kin
emat
ic (R
TK)
Tim
e K
inem
atic
(RTK
)sp
ecia
lized
rece
iver
mon
itors
sp
ecia
lized
rece
iver
mon
itors
car
rier
carr
ier p
hase
.ph
ase.
Pha
se o
ffset
is g
ood,
but
whi
ch w
ave
itera
tion?
Pha
se o
ffset
is g
ood,
but
whi
ch w
ave
itera
tion?
Land
Land
-- bas
ed ra
dio
stre
am h
elps
reso
lve
inte
ger a
mbi
guity
.ba
sed
radi
o st
ream
hel
ps re
solv
e in
tege
r am
bigu
ity.
Upo
n co
nver
genc
e, y
ield
s pr
ecis
ion
of ~
5cm
.U
pon
conv
erge
nce,
yie
lds
prec
isio
n of
~5c
m.
––W
AA
S: G
WA
AS
: Geo
sync
hron
ous
eosy
nchr
onou
ssa
telli
tes
over
US
:sa
telli
tes
over
US
:A
re th
emse
lves
ser
ving
as
addi
tiona
l A
re th
emse
lves
ser
ving
as
addi
tiona
l pse
udor
ange
sps
eudo
rang
es..
Bro
adca
stin
g di
ffere
ntia
l cor
rect
ions
.B
road
cast
ing
diffe
rent
ial c
orre
ctio
ns.
Yie
ldin
g ~1
Yie
ldin
g ~1
-- 3m
acc
urac
y3m
acc
urac
y ——FA
A a
ppro
ved
for m
any
land
ings
FA
A a
ppro
ved
for m
any
land
ings
no
w.
now
.
Cel
lula
r / P
CS
Cel
lula
r / P
CS
EE-- 9
11 L
egis
latio
n91
1 Le
gisl
atio
n––
Sin
ce O
ctob
er 1
, 200
1:S
ince
Oct
ober
1, 2
001:
Net
wor
ks m
ust b
e ab
le to
loca
te a
ll N
etw
orks
mus
t be
able
to lo
cate
all
cellp
hone
sce
llpho
nes ::
––95
% w
ithin
300
m.
95%
with
in 3
00m
.––
67%
with
in 1
00m
.67
% w
ithin
100
m.
Cel
lpho
nes
Cel
lpho
nes
with
ALI
(aut
omat
ic lo
catio
n id
entif
icat
ion)
:w
ith A
LI (a
utom
atic
loca
tion
iden
tific
atio
n):
––M
ust b
e av
aila
ble,
and
50%
+ of
act
ivat
ions
.M
ust b
e av
aila
ble,
and
50%
+ of
act
ivat
ions
.––
95%
mus
t be
loca
tabl
e w
ithin
150
m.
95%
mus
t be
loca
tabl
e w
ithin
150
m.
––67
% m
ust b
e lo
cata
ble
with
in 5
0m.
67%
mus
t be
loca
tabl
e w
ithin
50m
.
––S
ince
Oct
ober
1, 2
002,
95%
+ of
S
ince
Oct
ober
1, 2
002,
95%
+ of
cel
lpho
nece
llpho
neac
tivat
ions
mus
t be
ALI
activ
atio
ns m
ust b
e A
LI-- c
apab
le u
nits
.ca
pabl
e un
its.
Cel
lula
r / P
CS
Cel
lula
r / P
CS
How
are
H
ow a
re n
etw
orks
netw
orks
acco
mpl
ishi
ng th
is?
acco
mpl
ishi
ng th
is?
––C
an’t
forb
id u
se o
f old
han
dset
s.C
an’t
forb
id u
se o
f old
han
dset
s.––
Ang
le o
f arr
ival
(AO
A) t
o to
wer
s.A
ngle
of a
rriv
al (A
OA
) to
tow
ers.
Req
uire
s ca
refu
lly c
alib
rate
d an
tenn
a ar
ray.
Req
uire
s ca
refu
lly c
alib
rate
d an
tenn
a ar
ray.
Eve
n th
en le
ss d
efin
ite a
sol
utio
n.E
ven
then
less
def
inite
a s
olut
ion.
––Ti
me
of a
rriv
al (T
OA
) or T
ime
diffe
renc
e of
arr
ival
Ti
me
of a
rriv
al (T
OA
) or T
ime
diffe
renc
e of
arr
ival
(T
DO
A) t
o to
wer
s.(T
DO
A) t
o to
wer
s.R
equi
res
tight
tim
e sy
nchr
oniz
atio
n.R
equi
res
tight
tim
e sy
nchr
oniz
atio
n.C
DM
A a
lread
y ha
s su
ch s
ynch
roni
zatio
n.C
DM
A a
lread
y ha
s su
ch s
ynch
roni
zatio
n.G
SM
/ TD
MA
net
wor
ks c
an u
se v
ario
us s
trate
gies
.G
SM
/ TD
MA
net
wor
ks c
an u
se v
ario
us s
trate
gies
.––
Mul
tipat
hM
ultip
ath
/ RF
“fing
erpr
intin
g” o
f loc
atio
ns in
/ R
F “fi
nger
prin
ting”
of l
ocat
ions
in
data
base
.da
taba
se.
––P
rovi
ding
GP
SP
rovi
ding
GP
S-- c
apab
le h
ands
ets
with
Ass
iste
d G
PS
.ca
pabl
e ha
ndse
ts w
ith A
ssis
ted
GP
S.
Cel
lula
r / P
CS
Cel
lula
r / P
CS
Ass
iste
d G
PS
(AG
PS
)A
ssis
ted
GP
S (A
GP
S)
––Th
ese
hand
sets
hav
e a
full
or p
artia
l GP
S re
ceiv
er.
Thes
e ha
ndse
ts h
ave
a fu
ll or
par
tial G
PS
rece
iver
.––
Con
vent
iona
l rec
eive
rs ta
ke m
inut
es to
est
ablis
h fix
on
wea
k sa
tC
onve
ntio
nal r
ecei
vers
take
min
utes
to e
stab
lish
fix o
n w
eak
sat e
llite
el
lite
sign
als:
sign
als:
Unc
erta
in fr
eque
ncie
s (D
oppl
er)
Unc
erta
in fr
eque
ncie
s (D
oppl
er)
Unc
erta
in P
RN
C/A
cod
e ph
ase.
Unc
erta
in P
RN
C/A
cod
e ph
ase.
––A
GP
S n
etw
orks
sen
d G
PS
acq
uisi
tion
data
ove
r the
air
AG
PS
net
wor
ks s
end
GP
S a
cqui
sitio
n da
ta o
ver t
he a
irlo
wer
s G
PS
acq
uisi
tion
time
to s
econ
dslo
wer
s G
PS
acq
uisi
tion
time
to s
econ
ds——
or le
ss.
or le
ss.
Allo
ws
acqu
isiti
on o
f low
erAl
low
s ac
quis
ition
of l
ower
-- pow
ered
sig
nals
that
with
a c
onve
ntio
nal
pow
ered
sig
nals
that
with
a c
onve
ntio
nal
rece
iver
rece
iver
——so
me
are
usin
g la
ngua
ge a
bout
indo
or G
PS
acq
uisi
tion.
som
e ar
e us
ing
lang
uage
abo
ut in
door
GP
S a
cqui
sitio
n.
gpsO
negp
sOne
TMTM(( S
olim
anS
olim
an20
00)
2000
)
How
els
e ca
n ne
twor
ks h
elp
GP
S?
By
treat
ing
How
els
e ca
n ne
twor
ks h
elp
GP
S?
By
treat
ing
tow
ers
as
tow
ers
as p
seud
olite
sps
eudo
lites
..Q
ualc
omm
’s C
DM
A s
trate
gy is
alre
ady:
Qua
lcom
m’s
CD
MA
stra
tegy
is a
lread
y:––
Pre
cise
ly ti
me
Pre
cise
ly ti
me --
sync
hron
ized
.sy
nchr
oniz
ed.
––O
n G
PS
tim
e.O
n G
PS
tim
e.
gpsO
negp
sOne
TMTM(( S
olim
anS
olim
an20
00)
2000
)
σ=
19m
, jus
t res
olvi
ng c
lock
σ=
88m
, clo
ck fi
x, p
lus
extra
rang
e
σ=
139m
, clo
ck fi
x an
d tw
o ex
tra ra
nges
Rad
io F
requ
ency
Ana
lysi
sR
adio
Fre
quen
cy A
naly
sis
RFI
Ds
RFI
Ds
prov
ide
clos
epr
ovid
e cl
ose --
rang
e pr
oxim
ity d
etec
tion:
rang
e pr
oxim
ity d
etec
tion:
––In
suf
ficie
nt d
ensi
ties
even
Car
tesi
an tr
acki
ng c
an h
appe
n.In
suf
ficie
nt d
ensi
ties
even
Car
tesi
an tr
acki
ng c
an h
appe
n.
Wi
Wi -- F
iFila
tera
tion
late
ratio
nus
eful
but
coa
rse:
usef
ul b
ut c
oars
e:––
Mig
ht b
e in
tere
stin
g to
hyb
ridiz
e it
with
, say
, ine
rtial
tech
niq
Mig
ht b
e in
tere
stin
g to
hyb
ridiz
e it
with
, say
, ine
rtial
tech
niq u
es.
ues.
GP
S is
pro
ven
and
GP
S is
pro
ven
and
impr
ovin
gim
prov
ing
tech
nolo
gy:
tech
nolo
gy:
––R
emov
al o
f SA
, and
the
emer
genc
e of
WA
AS
are
sig
nific
ant!
Rem
oval
of S
A, a
nd th
e em
erge
nce
of W
AA
S a
re s
igni
fican
t!––
Poo
r rec
eptio
n ar
eas
will
be
ever
with
us,
thou
gh.
Poo
r rec
eptio
n ar
eas
will
be
ever
with
us,
thou
gh.
Giv
en c
onve
rgen
ce tr
ends
, G
iven
con
verg
ence
tren
ds, c
ellp
hone
cellp
hone
track
ing
offe
rs
track
ing
offe
rs
inte
rest
ing
poss
ibili
ties,
but
:in
tere
stin
g po
ssib
ilitie
s, b
ut:
––W
ill po
sitio
ning
dat
a be
por
ted
to e
xper
t use
rs o
f pho
nes?
Will
posi
tioni
ng d
ata
be p
orte
d to
exp
ert u
sers
of p
hone
s?––
Will
net
wor
k op
erat
ors
mer
ely
sell
loca
tion
Will
net
wor
k op
erat
ors
mer
ely
sell
loca
tion --
spec
ific
serv
ices
spec
ific
serv
ices
——ev
en a
dver
tisem
ents
?ev
en a
dver
tisem
ents
?
Clo
sing
Pon
derin
gsC
losi
ng P
onde
rings
The
futu
re o
f tra
ckin
g in
the
mar
ketp
lace
of t
he
The
futu
re o
f tra
ckin
g in
the
mar
ketp
lace
of t
he
“com
mon
man
” will
like
ly:
“com
mon
man
” will
like
ly:
––Le
vera
ge te
chni
ques
that
:Le
vera
ge te
chni
ques
that
:R
equi
re n
o en
viro
nmen
tal p
repa
ratio
n; o
rR
equi
re n
o en
viro
nmen
tal p
repa
ratio
n; o
rTa
ke a
dvan
tage
of i
nfra
stru
ctur
es a
lread
y in
pla
ce.
Take
adv
anta
ge o
f inf
rast
ruct
ures
alre
ady
in p
lace
.––
Not
offe
r any
thin
g lik
e th
e “s
tate
of t
he a
rt” w
e lik
e to
N
ot o
ffer a
nyth
ing
like
the
“sta
te o
f the
art”
we
like
to
expe
ct in
our
pro
fess
iona
l lab
orat
ory
setti
ngs.
expe
ct in
our
pro
fess
iona
l lab
orat
ory
setti
ngs.
––R
equi
re u
sing
mul
tiple
stra
tegi
es in
par
alle
l or
Req
uire
usi
ng m
ultip
le s
trate
gies
in p
aral
lel o
r se
quen
ce fo
r:se
quen
ce fo
r:B
road
mob
ility
; and
Bro
ad m
obili
ty; a
ndC
ontin
uous
cov
erag
e.C
ontin
uous
cov
erag
e.––
Ben
efit
huge
ly fr
om th
e in
crea
sed
qual
ity o
f hyb
rid
Ben
efit
huge
ly fr
om th
e in
crea
sed
qual
ity o
f hyb
rid
com
bina
tions
ove
r sta
ndal
one
tech
niqu
es.
com
bina
tions
ove
r sta
ndal
one
tech
niqu
es.
Clo
sing
Pon
derin
gsC
losi
ng P
onde
rings
As
a re
sult,
sev
eral
line
s of
rese
arch
look
pro
mis
ing
and
As
a re
sult,
sev
eral
line
s of
rese
arch
look
pro
mis
ing
and
usef
ul:
usef
ul:
––In
vest
igat
ing
hybr
id s
trate
gies
not
alre
ady
“don
e to
dea
th.”
Inve
stig
atin
g hy
brid
stra
tegi
es n
ot a
lread
y “d
one
to d
eath
.”––
“Met
a“M
eta --
track
ing”
stra
tegi
es to
track
ing”
stra
tegi
es to
Man
age
mul
tiple
trac
king
tech
nolo
gies
;M
anag
e m
ultip
le tr
acki
ng te
chno
logi
es;
Coo
rdin
ate
“han
doffs
” bet
wee
n th
em o
ver t
rans
ition
s; o
rC
oord
inat
e “h
ando
ffs” b
etw
een
them
ove
r tra
nsiti
ons;
or
Filte
r the
ir si
mul
tane
ous
usag
e; a
ndFi
lter t
heir
sim
ulta
neou
s us
age;
and
Mai
ntai
n so
me
tech
nolo
gyM
aint
ain
som
e te
chno
logy
-- glo
bal s
ense
of v
aria
nce.
glob
al s
ense
of v
aria
nce.
––A
dapt
ive
UIs
that
Ada
ptiv
e U
Is th
atM
ake
the
mos
t pro
duct
ive
choi
ces
abou
t how
to p
rese
nt
Mak
e th
e m
ost p
rodu
ctiv
e ch
oice
s ab
out h
ow to
pre
sent
in
form
atio
n, g
iven
cur
rent
ser
vice
s an
d th
eir a
ccur
acie
s;in
form
atio
n, g
iven
cur
rent
ser
vice
s an
d th
eir a
ccur
acie
s;D
on’t
“pro
mis
e m
ore
than
they
are
del
iver
ing”
in re
al ti
me,
in t
Don
’t “p
rom
ise
mor
e th
an th
ey a
re d
eliv
erin
g” in
real
tim
e, in
t erm
s er
ms
of a
ccur
acy
of lo
catio
n aw
aren
ess.
of a
ccur
acy
of lo
catio
n aw
aren
ess.
Do
give
intu
itive
cue
s to
use
rs a
bout
the
exte
nt th
at th
e cu
rren
Do
give
intu
itive
cue
s to
use
rs a
bout
the
exte
nt th
at th
e cu
rren
t t sy
stem
’s s
tate
can
be
trust
ed.
syst
em’s
sta
te c
an b
e tru
sted
.
Foxl
in 2
003
Sur
veys
and
Filt
erin
gW
elch
200
2
H
ight
ower
, 200
1
W
elch
, 199
5
Vis
ible
Infra
red
Rad
io F
requ
ency
(RF)
Iner
tial &
Gra
vim
etric
Mec
hani
cal
Mag
netic
Fie
ld
Aco
ustic
Saw
ada
2003
Gol
ding
1999
(+te
mp.
)
You
199
9
Raa
b 19
79
Add
lese
e19
97
Priy
anth
a 20
00R
ande
ll 20
01Fo
xlin
199
8
Wan
t 199
2
Hal
law
ay 2
003
Wel
ch 1
999
Hed
ges
2001
Chi
ntha
mm
it20
03
Hiro
se 2
003
B
ahl 2
000
Eng
e19
99
Wan
g 20
00
D
jukn
ic20
01S
olim
an20
00
New
man
200
1
Judd
199
7
Ran
dell
2003
Foxl
in 1
996
Elec
trom
agne
tic S
pect
rum
Ligh
t Spe
ctru
m