ntrs.nasa.govsot adequately desa-ihe the aulti-task fli~ht renaaenent s:tustlcn. either. the...
TRANSCRIPT
A O
UF
UE
IhC
YO
DEL
OF
P
ILO
T D
EC
ISIO
W
WA
KI'
IG
IW A
CU
LTI-T
AS
K
FLIG
HT
~N
AC
FY
EN
T SIT
UA
TIO
S*
R.
S.
Va
lde
n
and
W.
9.
Rouse
All
~c
atL
on
of
de
cis
ton
m
a k
in^
res
po
ns
ihil
itv
b
etw
ee
n
pil
ot
and
co
mp
ute
r is
c
on
si4
ere
d
an
d
a
fliv
ht
ma
na
~e
me
nt
tas
k,
de
sip
ne
d
for
the
s
tud
y
of
pil
at-
co
mp
ute
r in
tera
cti
on
, is
dis
cu
ste
d.
L
qu
eu
ein
n
the
ory
w
od
el
of
oil
ot
de
cis
ion
m
akin
p
in t
his
mu
lti-
tas
k,
co
ntr
ol
and
mo
nit
ori
np
s
itu
ati
ln
is
o
res
en
ted
. Ln
ex
pe
rim
en
tal
inv
es
ti a
ati
on
o
f p
ilo
t d
ec
isio
n
%s
kin
u
and
the
re
su
ltin
p
mo
de
l p
ara
ne
ters
a
re
dis
cu
sq
ed
.
IWTR
OD
UC
TIO
W
Au
tom
ati
on
h
as
fo
un
d
inc
rea
sin
p
us
e
in
a
irc
raft
, n
ot
oc
ly w
ith
re
sp
ec
t to
me
ch
an
ica
l s
ys
trm
s,
hu
t w
ith
re
sp
ec
t t
o
joh
s
no
rma
lly
th
ou
gh
t o
f a
s b
elo
nli
na
to
th
e a
irp
lan
e p
ilo
t.
As
air
pla
ne
n
cm
be
rs
inc
rea
se
and
na
vip
ati
on
beco
mes
%ore
co
mp
lex,
the
tr
en
d
wil
l c
ert
ain
ly
co
nti
nu
e.
Wem
pt
111
N
N
su
en
es
ts
tha
t nn
ad
va
nce
d
au
tom
ate
d
na
vie
ati
on
syste
m
wi
ll
beco
rn-
ne
ce
ssa
ry
in
th
e fu
ture
, b
eca
use
th
e r
eo
uir
ed
co
ntr
ol
pe
rfo
rm~
nc
e w
il
l
exce
ed
h
um
w
capability.
Yow
eve
r.
he
p
res
en
ts
two
rt
as
cn
s
why
co
n l
ete
l a
uto
ma
tic
c
on
tro
l o
f
air
cra
ft
is
un
lik
clv
in
th
ed
kd
re
:
a m
ach
ine
re
qu
ire
d
to
rep
lac
e
a
pil
ot
wo
uld
b
e
qu
ite
c
os
tly
; fu
rth
erm
ore
. m
ach
ine
s do
no
t a
da
pt
we
ll t
o u
na
nti
cip
ate
d e
ve
nts
. C
on
tro
l o
f a
irc
ra
ft t
hu
s w
ill
be
th
e t
as
k o
f a
aa
n-m
ach
ine
syste
m.
Wit
h
the
re
ce
nt
rap
id
ex
oa
ns
ion
o
f
diu
ita
l te
ch
no
lo~
y
(es
pe
cia
lly
m
icrn
aro
ce
ss
ors
),
thr
wo
rds
-man-m
ach
ine"
and
wn
an
-co
mp
ute
r~ h
ave
h
tco
re
pra
cti
ca
llv
syn
ory
mo
us.
The
co
mp
ute
r ca
n
ha
ve
s
ev
era
l ro
les
in
aq
a
irb
orn
e m
an-c
om
pute
r s
ys
tm.
Sp
ec
ial-
ou
rpo
se
p
roc
es
so
rs
ca
n b
e d
ed
ica
ted
to
su
ch
th
inp
s
as
au
top
ilo
t and
na
viu
ati
on
a
ids
, d
isp
lav
3,
co
mm
un
ica
tio
n
sy
ste
:s,
or
su
bsyste
m
mo
nit
ori
np
; in
su
ch
a
pp
lic
ati
on
s,
the
c
sm
pu
ters
w
ou
ld
se
rve
m
ere
ly
as
di
~i
ta
l
co
ntr
oli
ers
o
f
the
v
ari
ou
s
de
vic
es
. T
he
pil
ot'
s
wo
rklo
ad
w
ou
ld
thu
s be
red
uce
d
as
the
c
om
pu
trrs
h
an
dle
m
uch
of t
he
lo
w-l
ev
el
info
rma
tio
n p
roce
:fsin
p.
Or,
to
p
o
a
ste
p
furt
he
r,
the
co
ma
ute
r c
ou
ld
ha
ve
a
more
n
en
era
l ro
le.
Th
r c
om
c~
lte
r mip
ht
ha
ve
in
~c
~t
s
fro
m
a v
ari
etv
o
r a
irc
ra
ft
de
vic
es
a
nd
su
bsyste
ms.
and
inc
rea
se
d
co
ntr
ol
ca
pa
hil
itit
s
as
we
ll.
The
co
mp
ute
r w
ou
ld
the
n
br
ex
pe
cte
d
to
init
iate
a
cti
on
l o
f
it
s
own
(i.e
.,
mak
e
Th
is re
se
arc
h w
as
su
pp
ort
ed
by
the
na
tic
na
l A
ero
na
uti
cs
a
nd
S
pace
A
dm
inis
tra
tio
n u
nd
er
NASA - 4
mts
p
ran
t MSG-2119.
de
cis
ion
s),
b
ase
d
on
th
e
ina
uts
it
rrc
eiv
es
. T
he
p
ilo
t's
w
ork
loa
d w
ou
ld
be
re
du
ce
d a
t a
hig
he
r le
ve
l,
by
th
e c
om
pu
ter
sh
ari
ng
in
de
cis
ion
-ma
kin
g
tas
ks
. I
t i
s t
5is
la
tte
r c
on
ce
pt
of
the
ro
le o
f t
hc
co
mp
ute
r in
air
bo
rne
sv
ste
ns
wh
ich
i
s t
he
c
on
ce
rn o
f th
e re
se
arc
h d
erc
rib
ed
he
rein
.
A c
om
pu
trr
wit
h
sig
nif
ica
nt
decis
ion-a
akln
u
res
po
ns
ibil
itie
s
in
the
fl
ivh
t-a
an
a~
ea
en
t ta
sk
ca
n
cre
ate
p
ote
~tia
l p
rob
lem
s,
ho
we
ve
r.
One
o
f th
e
mo
st
cru
cia
l o
rob
lem
s
is
th
at
of
ail
oc
ati
on
o
f
de
cis
ion
-ma
kin
e
res
po
ns
ibil
ity
b
etv
ren
p
ilo
t and
co
mp
ute
r [2
1.
hll
oc
ati
on
ca
n
eit
he
r be
sta
tic
(s
itu
~a
tio
n-i
nd
ep
en
de
nt)
, o
r d
yn
am
ic
(sit
ua
tio
n-d
ep
en
de
nt)
.
Co
ns
ide
r fir
st
a
systw
m
in
w
hic
h
all
oc
ati
on
o
f
res
po
ns
ibil
itie
s
is
s
tati
c.
Th
ere
it
a
pre
-de
fin
ed
, n
on
-ov
erl
ap
pin
a
se
t o
f re
sp
on
sib
i?lt
ies
fo
r
ea
ch
d
ec
isio
n-m
ak
er
(mar
and
co
mp
ute
r).
Such
a
s
vs
tem
h
as
on
e
de
sir
ab
le fe
atu
re
in
th
at
no
c
on
fus
ion
i
s p
os
sib
le o
ve
r w
ho
sh
ou
ld
pe
rfo
rm
a
giv
en
ta
sk
. H
ow
eve
r.
the
re
are
s
ev
era
l s
eri
ou
s
dra
wb
acks
to
th
e
sta
tic
a
llo
ca
tio
n
of
re
sp
on
sib
ilit
v.
Th
ese
in
clu
~e
:
1.
Po
or
uti
liz
ati
on
o
f syste
m
res
ou
rce
s
du
e
to
ta
ch
d
ec
isio
n
make
r b
ein
g
pro
hib
ite
d
rom
p
erfo
rm
tn~
the
o
the
r d
ec
isio
n
ma
ke
r's
ta
sk
s,
eve
n
whe
n o
ne
d
ec
isio
n
make
r i
s o
ve
rlo
ad
ed
a
nd
o
ne
Is i
dle
.
2.
Dif
fic
ult
y
for
the
p
ilo
t w
hen
un
fore
se
en
c
irc
ws
tan
crs
fo
rce
h
im
to
ta
ke
o
ve
r ta
sk
s
for
wh
ich
h
e
no
rma
lly
h
as
no
re
sp
on
sib
ilit
y.
-
3.
Po
ss
ible
p
ilo
t re
se
ntm
en
t a
t n
ot
be
ing
a
llo
we
d
to
p
erf
orm
c
ert
ain
ta
sk
s,
es
pe
cia
lly
w
ken
he
has
no
thin
@
els
e t
o d
o.
Fo
r th
e
ab
ove
re
aso
ns,
a
dyn
am
ic
all
oc
ati
on
o
f
res
po
ns
ibil
ity
m
igh
t be
%o
re
ap
pro
pri
ete
. S
imu
la'i
on
s
tud
ies
o
f
this
a
lte
rna
tiv
e
by
R
ouse
C
31
su
pp
ort
.
Is
c
on
clu
sio
n.
Su
cc
inc
tly
, th
e
ide
a
is
th
at
the
se
rve
r w
ho . ,
mo
st
ab
le,
in
a
Riv
en
s
itu
ati
on
, w
ill
ass
um
e re
sp
on
sib
ility
for a
ta
sk
.
Dyn
am
ic a
llo
ca
tio
n o
f r
es
po
ns
ibil
ity
is
no
t w
ith
ou
t i
ts
d
raw
ba
cks,
ho
we
ve
r.
The
mo
st
Imp
ort
an
t p
rob
lem
to
bt
so
lve
d
is
th
at
of c
om
mu
nic
ati
on
b
etw
ee
n
pil
ot
and
co
mp
ute
r.
Th
is
is
ne
ce
ss
ary
to
a
vo
id
co
nfl
icts
w
hic
h
oc
cu
r w
hen
the
tw
o
de
cis
ion
m
ake
rs
ac
t fn
de
pe
nd
en
tly
. re
su
ltin
a
In
d
eu
rad
ed
p
erf
orm
an
ce
. (R
ouse
[3
3 p
res
en
ts
sim
ula
tio
n
res
ult
s
tha
t il
lus
tra
te
this
e
ffe
ct)
. A
po
ss
ible
s
olu
tio
n
to
th
is
pro
ble
m i
s t
o *
~s
e
the
co
mp
ute
r 8
s a
h
acku
p
for th
e p
ilo
t.
to
b
e
turn
ed
o
n
when
the
p
ilo
t's
w
ork
loa
d
be
co
ne
s
too
E
rea
t a
nd
/or
his
pe
rfo
rma
nc
e
de
gra
de
s.
https://ntrs.nasa.gov/search.jsp?R=19790009326 2020-01-27T19:35:43+00:00Z
The
men
-,o.n
put.e
r syste
m
is
th
us
-.
een
as
one
in
wh
ich
b
oth
th
e
pil
ct
and
cm
pu
ter
mo
nit
or
ea
ch
o
thrr
--th
e
pil
ot
su
pe
rvis
es
and
Tu
nit
ors
th
e
co
mp
ute
r's
a
cti
vit
ies
, fo
r th
e
pu
rpo
se
or
ins
urt
nff
t.
ha
t th
e
co
mp
ute
r i
s
pe
rfo
rain
a
ad
eq
ua
tely
, and
tha
t n
o
ma
lftt
nc
tio
n
ha
s
oc
cu
rre
d;
the
co
mp
ute
r m
on
ito
rs
the
v
ila
t t
o
de
term
ine
w
hen
he
i
s
ov
erl
oa
de
d
ano
ne
ed
s
as
si-
tan
ce
. S
uch
a
syste
m
wo
uld
h
op
efu
!ly
m
inia
ize
c
on
f:ic
t and
av
oid
th
e
dif
Tic
ult
ies
d
isc
us
re.
above
.
In
ord
er
for
the
c
oq
pu
ter
to
be
a
ble
to
te
ll w
hen
the
p
ila
t n
ee
ds
as
sis
tan
ce
, it
m
sc
fir
st
ha
ve
a
m
eth
od
fo
r d
ete
rnia
inp
th
e p
ilc
t's
w
ork
loa
d,
ic t
erm
s
of t
as
k
dem
ands.
T
hen,
it n
ee
ds
a c
rit
erio
n f
~r
Ie
cid
inp
if i
t s
ho
uld
re
ou
es
t re
sp
on
sib
Lli
ty
for
any
task
:.
Chu
and
Rouse
I
41
h
ave
p
rop
os
ed
a
i~
rn
ula
tio
n o
f th
is
pro
ble
m.
TO
a
pp
ly
this
fo
rru
lati
on
to
any
sp
ec
ific
s
itu
at1
.n.
the
fir
st
ste
p
is
d
ev
elo
on
cn
t o
f
a
mo
de
l o
f th
e
un
aid
ed
p
erf
orm
an
ce
o
f th
e
hu
na
n w
ith
re
sp
ec
t to
ta
sk
s
typ
ica
l o
f th
os
e
en
co
un
tere
d
in
th
e s
itu
ati
on
o
f in
tcrrs
t.
The
pu
rpo
se
of t
his
re
po
rt i~
to
pre
se
nt
the
re
su
lts
of
pu
rsu
ing
th
is
firs
t s
tep
in
th
e
co
nte
xt
of
flia
ht
mana#em
ent.
An
ex
pe
rim
en
tal
sit
ua
tio
n
rhic
h
was
d
rve
lop
ed
to
stu
dy
p
ilo
t p
err
orn
an
ce
in
sin
ula
tefl
fli
eh
t c
on
dit
ion
s i
s
N
N
de
sc
rib
ed
. A
lte
rna
tiv
e
mo
de
l3
thn
t d
es
cri
be
p
ilo
t d
cc
isic
n
Y
aa
kln
fi
in
th
is
flip
ht
na
np
pe
ne
nt
sit
ua
tim
a
re
pre
se
nte
d.
Fin
all
y,
an
ex
pe
rim
en
t i
s
dis
tus
se
d
whose
p
urp
os
e
was
in
ve
sti
(ta
ti0
n o
C
the
s
uit
ab
ilit
y o
f th
e
alt
ern
ati
ve
mo
de
ls.
THE
TASK
The
pilo
t is
pre
se
nte
d
wit
h
a
sim
ula
ted
a
irp
lan
e
ins
tru
me
nt
pa
ne
l,
dra
wn
o
n
a
CRT
(se
e
Fi~
ur
e I)
. T
he
dis
pla
y
inc
lud
es
s
tan
da
rd
air
cra
ft
ins
tru
me
nts
, su
ch
a
s
ar
tif
icia
l h
ori
zo
n.
alt
ime
ter.
h
?3
dim
c
an
d
air
sp
ee
d
ind
ica
tors
. 41so
d
isp
lay
ed
is
a
m
ap
wh
ich
in
dic
ate
s
the
c
ou
rse
th
e
air
pla
ne
i
s
to
foll
ow
. fin
a
irp
lan
e-s
ha
>e
d
sym
bo
l.
su
pe
rim
po
se
d
on
th
e
na
p.
ind
ica
tes
th
e
air
ola
ne
's
ac
tua
l p
os
itio
n.
L s
ma
ll
cir
cle
m
ow
3
alo
n~
the
m
apped
co
urs
e
at
a
co
ns
tan
t sp
ee
d
of
600
fps
(1
83
n
lsl
an
d
ind
ica
tes
th
e p
os
itio
n t
he
a
irc
rs
ft s
ho
uld
h
ave
fo
r it t
o h
e
on
c
ou
rse
and
sc
he
du
le.
Ne
ar
the
lo
we
r e
da
~' o
f t
he
dis
pla
y,
se
ve
ral
dia
ls a
re s
how
n.
wk
ich
re
pre
se
nt
(ab
str
ac
tly
) R
auR
rs
for
su
ch
th
ing
s
as
fue
l,
ele
ctr
ica
l,
or
hy
dra
uli
c
su
bsyste
ms.
The
ga
ua
r p
oin
ters
m
cve
wit
h
ao
pa
ren
tlv
ra
nd
om
n
oti
on
. *e
ne
ratc
d
by
pa
s9
inp
C
au
ss
irn
w
hit
e
no
ise
th
rou
an
a
se
co
nd
-ord
er
sa
mp
lsd
-da
ta
filt
er
.
Th
r p
ila
t c
on
tro
ls
Bo
ein
~ 70
7
air
cra
ft
dyn
am
ics
as
dis
cc
as
ed
b
y
bln
ke
loc
k f5).
He
co
ntr
ol3
th
e a
irc
ra
ft p
itc
h
ac
d
ro
ll
dyn
am
ics
wit
h
a
joy
sti
ck
. a
no
the
r c
on
tro
l s
tlc
k
co
ntr
ols
th
c
alr
sp
ee
d.
The
oil
otN
s t
as
k
is
to
f
ly
the
a
irp
lan
e
alo
n~
the
sa
pp
ed
ro
ute
, m
ain
tain
ing
a
ril
ed
alt
itu
de
a
nd
s
tab
le
pit
ch
an
d
ro
ll a
ttit
ud
e.
The
pil
ot
is
su
pp
ose
d
to k
ee
p
as
clo
se
to
th
e
"on
-sch
ed
ule
w
do
t.
an
d
as
clo
se
t
o t
he
m
apped
co
urs
e.
as
po
ss
ible
. w
hil
e h
old
in*
the
a
irc
ra
ft a
t 40,0
00
ft.
(12,1
92
a.)
a
ltit
ud
e.
In
ad
dit
ion
to
fly
ing
th
+ a
irp
lan
e,
the
p
ilo
t n
on
ito
rs
the
su
bsyste
m in
dic
ato
rs f
or
po
ss
ible
ev
en
ts.
(An
e
ve
nt
ha
s
oc
cu
rre
d w
hen
the
p
oin
ter
mo
tio
n
slo
ws
a
nd
s
top
s.
po
inti
na
dow
nw
ard
as
sh
wn
fo
r s
ub
sy
ste
% 9
in
Fia
ure
1.
) W
hen
the
p
ilo
t th
ink
s
an
e
ve
nt
ha
s
oc
cu
rre
d,
he
e
nte
rs
the
c
orr
es
po
nd
ing
num
ber
on
th
e
ke
yb
oa
rd.
(I
f
mo
re
tha
n
uric
ev
en
t h
as
oc
cu
rre
d,
the
p
ilo
t ch
oo
se
s
the
h
igh
er-
pri
ori
ty,
or
low
er
su
bsyste
m
num
ber,
e
ve
nt.
) T
he
d
isp
l8y
s
hw
n
in
F
iuu
re 2
th
e6
a
pp
ea
rs.
On
this
ne
w
dis
pla
v,
the
s
ub
sy
-te
e
dia
ls
ha
ve
b
ee
n
rep
lac
ed
w
ith
tw
o
row
s
of n
um
be
rs,
one
lab
ell
ed
'B
RA
ICH
D,
the
o
the
r lr
5e
lle
d
'ST
AT
E'.
Th
is
co
rre
sp
on
ds
to
th
r
fir
st
lev
el
of
a c
he
ck
lis
t a
ss
oc
iate
d w
ith
th
e s
ub
sy
ste
a w
hic
h u
as
se
lec
ted
. T
he
pil
ot l
oo
ks
to
r a
b
ran
ch
w
ith
a
s
tate
of
'0'
an
d
en
ters
th
e
bra
nc
h
num
ber
on
th
e
ke
yb
oa
rd.
If
the
c
he
ck
lis
t fo
r th
at
bra
nc
h
co
nti
nu
es
. th
e
ne
xt
checklist
lev
el
is
d
isp
lay
ed
. Y
hen
the
e
nd
o
f
the
c
he
ck
lis
t i
s
rea
ch
ed
, th
e
su
bsyste
m
ind
ica
tors
re
ap
pe
ar,
w
ith
th
e
su
bsyste
m
ind
ica
tor
mo
tio
n re
sto
red
. T
hese
a
cti
on
s a
imla
te
the
c
he
ck
lis
t p
roc
ed
ure
s
a
pil
ot
mig
ht
KO
th
rou
Rh
w
he
n
pe
rfo
rmin
a ta
sk
s i
n a
re
al
air
cra
ft.
The
ap
pa
ratu
s
use
d
for
the
e
xp
eri
me
nts
, th
e
sim
ula
tio
n
so
ftw
are
~
st
d
to
pro
du
ce
th
e
ab
ove
ta
sk
. a
d
es
cri
pti
on
5
f t
ht
air
cra
ft d
yn
am
ics,
an
d
an
ex
pla
na
tio
n
of
the
s
uh
sy
ste
7s
p
ort
ion
o
f th
e
sia
ula
tio
n
are
d
isc
us
se
d
in
ma
re
de
tail
in
W
ald
en
's
the
sis
161
.
THE
MO
DEL
The
ro
le o
f th
e h
um
an
op
era
tcr
in
co
mp
lex
man-m
ach
ine
syste
ms
is
ch
an
uin
a
fro
*
tha
t o
f in
-lo
op
c
on
tro
lle
r t
o t
ha
t o
f
su
pe
rvis
or
or
ra
nit
or
of
au
tom
ati
c,
po
ss
ibly
c
c~
pu
ter
op
era
ted
, c
on
tro
ls.
In
su
ch
a
role
. it
is
th
e
hum
an's
m
o~
!ito
rin
g p
trfo
rna
nc
e
ah
ich
i
s o
f
ma
jor
inte
res
t.
The
as
su
mp
tio
n
is
th
at
es
se
nti
al
co
ntr
ol
tas
ks
wil
l be
pe
rfo
rme
d
ad
eq
ua
tely
, p
erh
ap
s
at
the
e
xp
en
se
o
f m
en
ito
rin
((
pe
rfo
rma
nce
.
Th
:s,
th
e F
oa
l o
f th
e m
od
eli
ng
e
ffo
rt
rep
ort
ed
he
re
is
to
pre
dic
t th
e
su
bsyste
m t
as
k
pe
rfo
rma
nce
, n
ea
su
red
in
te
rw
o
f m
ean
and
va
ria
nc
e
of
su
bsyste
m
ev
en
t w
aft
inu
ti
ne
, a
s a
fun
cti
on
o
f ta
sk
p
ara
me
ters
1
1
ev
en
t a
rriv
al
rate
, and
2)
co
ntr
ol
tas
k d
iffi
cu
lty
.
Ex
isti
ng
m
od
els
p
rov
e
ina
de
qu
ate
i
n
de
sc
rib
in~
the
a
ult
i-ta
sk
f
li
~h
t m
sna#em
ent
sit
ua
tio
n
dis
cu
ss
ed
h
ere
. C
on
tro
l m
od
els
, fo
r e
xa
np
le t
he
Kle
inn
an
, B
aro
n.
an
d Levison
oo
tirn
l m
3tr
o1
n
od
el
!71
(an
d
it
s m
any
c.x
ten
sio
ns
),
are
p
on
e-r
ne
d
on
lv u
ith
nre
dfc
tiv
co
oir
ol
prr
fov
na
nc
e:
a m
od
el
us
efu
l in
n
red
icti
nt:
su
bsyste
m
s?
:.vic
e
oe
rfo
rna
nc
e
Ls
sc
ue
ht.
In
str
um
en
t-m
on
itc
rin
g
mo
de
ls
do
so
t a
de
qu
ate
ly
de
sa
-ih
e t
he
a
ult
i-ta
sk
f
li
~h
t
ren
aa
en
en
t s
:tu
stl
cn
. e
ith
er.
T
he
ins
tru
me
nt-
mo
nit
ori
na
m
de
l o
f S
en
de
rs
($1
fo
c*~
se
s on
de
tail
s
of
ins
trt~
me
-* s
ca
nn
l np.
su
ch
as
ins
tru
me
n+
f i
ra
tio
n
fre
ou
en
cte
s
and
se
.u
en
ce
s.
T!.at
of
Iarh
on
ell
[9
) is
c
xc
ern
cd
w
ith
s
an
pll
nff
~
glic
y b
ase
d
on
in
st,
um
en
t p
ric
rlt
irs
. W
eft
he
r ty
pe
o
f ro
de
l is
ro
n=
ern
ed
v
ith
;u
h%
rou
en
t a
cti
on
s rc
su
ltin
p t
row
in
Ttr
un
en
t .~
os
rrv
ati
on
s.
A o
ue
ue
inp
th
er?
-v a
od
el
of
the
m
ult
i-ta
w
sit
ua
tio
n
is
p
rop
ose
a.
The
mo
ni?
ori
np
ta
w<
is
mo
de
led
as
a s
ina
le-s
erv
er
qu
cu
ein
a
sy
ste
?,
vit
h
su
bsyste
m
ev
en
ts
as
cu
3to
me
?s.
The
wn
ito
rin
p ?
,a;!,
is.
ir fa
ct,
a
ou
eu
~.i
nff
sv
ste
n,
and it s
ee
ns
na
tura
l to
u
se
a
,,u
~u
ein
R
the
ory
d
es
cri
pti
on
fo
r it
.
The
pro
po
se
d
m3
de
l is
a
sln
ula
tio
n
mo
de
l;
an
ap
pro
pri
ate
a
r~
alv
tic
al s
olu
tio
n
is
no
t a
va
ila
ble
. (S
ee
Chu
[I0
1
for
erp
lan
htl
on
o
f th
e
b2
sic
o
pe
rati
on
o
f th
e
qu
eu
ein
* a
iau
lnti
on
.)
In
th
e
fir
st m
od
eli
ng
att
em
pt
(he
rea
fte
r re
'err
ed
to
as
Mo
de
l 1
).
the
m
on
ito
ri-8
ta
sk
is
d
rsc
rib
ed
as
a
MIE~/I:NPRP/N/N
qu
eu
ein
p
svste
iv.
!See
W
hit
e.
S:h
mid
t.
and
Re
nn
rtt
11
11
fo
r d
efi
nit
ion
s
of
the
s
tan
da
ri
qu
eu
ein
u
term
i~o
lorv
em
plo
ye
d
he
re.)
T
he
~o
nt
ro
l a
s i
s
no
t e
xp
lic
itiy
m
od
ele
d.
Ins
tea
d,
the
c
on
tro
l ta
sk
is
in
clu
de
d
irn
gli
ctt
ly
in
the
s
erv
ice
t.im
e
dis
trib
uti
on
o
ara
ne
ters
(i
.e.,
fo
r th
c
Erl
an
r d
istr
ibu
tio
n.
the
n
ecn
and
sh
ap
e
pa
ram
ete
r k
).
wh
ich
c
ha
n~
e w
ith
in
cre
a-e
d
map
ro
mp
lex
lty
(d~t to
th
e
mn
tro
l ta
sk
).
Co
np
ute
r s
imu
lati
on
o
f th
is
qu
eu
cln
x
sy
str
m
pro
vid
ed
p
reli
ain
arv
d
ata
u
se
ful
In m
akin
g
:ste
t im
prc
ue
me
~ts
. h
ut
did
nc
t n
etc
h
the
e
xo
erin
rn
t~l d
ata
w
ell
u
r.?
er
low
tr
aff
ic c
on
dit
ion
s.
(See
RE
SU
LTS
)
The
se
ccn
d
refi
ne
me
nt
of
the
m
od
el
(re
fcrr
ca
to
a
s
Mo
de
l 2)
att
em
pts
to
a
cc
ou
nt
for
the
in
pa
ct
of
the
c
on
tro
l ta
sk
3n
ta
t
su
bsyste
m
tas
k
pe
rfo
rmm
ce
. T
his
i
s done
by
us
in6
a
hl~
he
r eff
ec
tiv
e a
rriv
al
ra
te f
or
su
bs
ys
ten
ev
en
ts:
Ae
rre
cti
ve
=
Aa
tu
al
+ A+
wh
ere
AX
i
s t
he
n
crc
as
e C
n
arr
iva
. ra
te
ne
ed
ed
to
m
atc
h
em
pir
ica
l su
ksyste
m e
ve
nt
wa
itin
g ti
me
. T
he
co
ntr
ol
tas
k
is
thu
s in
co
rpo
rate
d
into
th
e m
onitor
in^
po
rtio
n ~
f th
e ta
sk
as
inrr
ea
se
d e
ve
nt
arriv
als
in
th
e
su
bsyste
m q
ue
ue
s.
Us
ing
th
e
sam
e s
imu
lati
on
p
rog
ram
as
b
efo
re,
su
bsyste
m
ev
en
t a
rriv
al
rate
w
as
va
cie
d
un
ti'
the
m
ean
wa
itin
a
tis
es
p
red
icte
d
by
the
s
iau
lati
cn
m
atc
ieo
th
e
ex
pe
rim
en
tall
y
ohse
rver2
w
ait
ire
ti
me
s.
ft:i
s m
od
el
pro
vid
es
a R
ood ~
re
dic
txo
n of
the
wa
itin
g
tin
e
dis
trib
uti
on
s
for
ind
ivid
ua
l su
bsyste
ms
(se
e
RE
SUl
S).
H
ow
eve
r.
in t
erw
c oC
the
p
hy
sic
al
sv
ste
n,
the
AX p
nra
.ete
r is
ad
nit
tcd
ly u
na
oo
ea
lin
a
Yit
h
the
th
ird
re
~t?
tme
nt,
th
e
no
de
l (*
oC
el
31
ap
pro
ach
es
the
re
al
sit
ua
tio
n
mo
st
clo
se
ly.
The
co
ntr
ol
tas
k
is i
nc
orp
ora
ted
as
a s
pe
cta
l queue;
cu
sto
src
rs a
rriv
inm
i
n
the
c
on
tro
l ta
sk
q
ue
ue
ca
n
pre
em
pt
the
s
erv
ice
o
f a
su
bsyste
m e
ve
nt.
5
erv
ice
o
f t9
r
su
bs
yrt
em
e
ve
nt
wil
l th
en
be
resu
ne
d
at
the
p
oin
t 0.-
intn
rru
pti
on
a
fte
r
tbr
c
on
tro
l a
cti
on
is
co
mp
lete
. T
his
q
va
ue
in~
syste
m
is
de
sc
rib
ed
as
a
WIE
k/l
:PR
P/W
.'N
queue.
One
o
f th
e p
roh
lem
s v
ith
!c
is
thir
d m
od
el
is
de
fin
in*
a "c
usto
me
rn
for
the
c
on
tro
l ta
sk
o
ue
ur.
I
n
fec
t,
it
is
d
isp
lay
ed
e
rro
r (m
ore
s
pe
clf
icrl
ly.
iac
.re
ne
nts
o
f d
isp
lay
ed
e
rro
r)
wh
ich
q
ue
ue
s
(ac
cu
mu
late
s)
for
ztt
en
tio
n;
bo
we
ve
r,
the
s
ize
(o
f th
e
inc
rem
rnta
l b c
ust
om
ern
i
s
unkn
oun.
A re
as
on
ab
le
ap
pro
ach
i
s
to
cn
ns
lde
r th
e
asr;
um
pti
on
m
ade
ea
rli
er
tha
t th
e
co
ntr
ol
tas
k
cu
sto
me
r p
ree
mp
ts
su
bs
ys
tt~
s
erv
ice
; i.e
.,
wh
en
eve
r a
co
ntr
ol
tas
k
cu
sto
me
r i
s p
res
en
t,
su
bsyste
m
mo
nit
ori
ng
(o
r e
ve
nt
se
rvic
e)
is
ia
ne
dia
tely
p
ree
mp
ted
, and
the
c
on
tro
l ta
sk
se
rve
a.
A cu
sto
me
r il t
he
c
on
tro
l ta
sk
q
ue
ue
c
ou
ld
the
n
be
d
efi
ne
d
(lo
os
ely
\ as
a
*s
l~n
ific
an
tn or
na
cti
on
-ev
ok
ln~
~' am
ount
of
dis
pla
y
err
or;
w
hen
a
sig
nif
ica
nt
err
or
is
pre
se
nt,
s
ub
sy
stc
n
se
rvic
e
wil
l be
ore
em
pte
d
an
d
co
ntr
ol
ins
ert
ed
tc
n
ull
th
e
err
cr.
A
sim
ple
w
ay
to
in
dir
ec
tly
m
ea
su
re
the
C
req
ue
ncy
uit
h
rhic
h
"sig
r.if
ica
nta
e
rro
rs
arr
ive
(q
ue
ue
) i
s
tc
m
ea
su
re
the
fr
eq
ue
nc
y
wit
h
wh
ich
d
isp
lay
ed
e
rro
rs
are
s
erv
ice
d,
1.e.
. th
e fr
eq
ue
nc
v
at
wh
ich
c
on
tro
l a
ctl
on
s
are
in
se
rte
d t
o n
ull
th
em
.
To
es
tim
ate
th
e
fre
qu
en
cy
o
f c
on
tro
l re
sp
on
se
s,
the
num
ber
of
se
pa
rate
c
on
tro
l a
cti
on
s
pe
rfo
rme
d
by
the
p
ilo
t ca
n
be
co
un
ted
, a
nd
fr
om
th
is
nu
rbe
r.
an
arriv
al
ra
te f
or
co
ntr
ol
tas
k
cu
sto
ne
rs
ca
n
be
o
bta
ine
d.
To
o
bta
in
an
e
sti
ma
te
cf
the
a
ve
rap
e
se
rvic
e
ra
te f
or
a c
ur
?o
rr
i
n t
he
c
on
tro
l ta
sk
queue.
the
c
on
try
i L
ask
se
rvic
e ra
te p
arm
ete
r in
th
e s
inu
lati
rn
pro
gra
m c
:n b
e
va
rie
d v
ntil
a g
oo
d fit to
e
xp
eri
ne
nta
l d
ata
is
ob
tain
ei.
To
acco
u,t
for th
e i
nc
rea
se
d w
ait
in8
tim
e in
cu
rre
d f
rom
e
rro
rs
su
ch
as
fals
e
ala
rms
a
nd
in
co
rre
ct
ac
tio
ns
. m
od
if*e
ati
on
s
to
the
b
as
ic
sim
ula
tio
n
mo
de
l w
ere
m
ade.
F
als
e
s'a
rms
are
a
cc
ou
nte
d
for
by
as
in*
an
oth
er
ou?ue.
wit
h
ev
en
t (fr
lse
ala
rms
) a
rriv
al
ra
te e
su
al
to
th
e
arriv
al
ra
te
of
fals
e
ala
rms
in
th
e
ex
ce
rim
en
t.
Fo
r th
e t
as
k
dis
cu
ss
ed
h
ere
. th
e
se
rvic
e
tim
e
use
o
for
a
fals
e
ala
rm
is
on
e-f
ifth
th
e
mea
n e
ve
nt
se
rvic
e
tim
e,
si
~c
e on
ly
the
fir
st o
f l
ive
le
ve
ls o
f t
he
ch
ec
kli
st
is
show
n w
%n
a f
als
e a
larm
oc
cu
rs.
Inc
orr
ec
t a
cti
on
s
ere
a
cc
ou
nte
d
for
in
th
e
sia
ula
tio
n
bv
--ir
st
es
tim
ati
ne
th
e pro
bability
o?
a
n
inc
orr
ec
t a
cti
on
u
hjl
e s
erv
icin
(r
a su
bsyste
m
ev
en
t.
Thi.1
i
s d
on
e
sim
ply
h
y
div
fdin
e
the
num
ber
cf
in
rorr
cc
t a
cti
on
s
bv
the
c
wb
tr
of
e
ve
nts
. B
ase
d
on
th
is
rate
, in
co
rre
ct
ac
tio
ns
a
re
sim
ula
ted
. F
or
the
m
on
ito
rin
g
tss
k
dis
cu
ss
ed
h
ere
, th
e
P-n
alt
y
for
an
in
co
rre
ct
ac
tio
n
?s
an
in
cre
as
e
in
wa
itin
g
tim
e
eq
ua
l t
o o
ne-
hal
f th
e m
ean
se
rvic
e t
ime
fo
r su
bsy
stem
rr
en
ts.
(On
the
a
ve
rag
e,
if
in
co
rre
ct
ac
tio
ns
~c
cu
r
ra-r
l^m
ly,
the
p
ilo
t w
il
l
be
hal
fway
th
rou
gh
a
ch
erd
lis
t p
roce
du
re,
and
thu
s
wa
ste
o
ne-
hal
f o
f th
e
av
era
ge
s
erv
ice
ti
me
.)
Th
e w
aiti
ng
-tim
e p
en
alt
v
cou
ld
be
even
g
rea
ter,
if
a
fte
r a
n
inro
rre
ct
~c
ti
on
(an
d
the
su
bsy
stem
in
dic
ato
rs
ar
e
retu
rne
d
to
th
e d
isp
lay
),
the
pil
ot
de
tec
ts a
nd
s
erv
ice
s a
n
ev
en
t in
a
hig
he
r-p
rio
rity
q
ueu
e.
Th
is
is
als
o
acco
un
ted
fo
r in
th
e s
iqu
lati
on
mo
del
.
AN
EX
PFR
IMEN
T
Usi
ng
th
e e
xp
eri
me
nta
l s
itu
ati
on
d
esc
rib
ed
e
arl
ier,
an
ex
per
imen
t w
as
per
form
ed
to
st
ud
y
sub
jec
t "
pil
otn
p
erfo
rman
ce
in
a m
ult
i-ta
sk
flig
ht
man
agem
ent
sit
ua
tio
n.
Th
e tw
o
~n
de
pe
nd
en
t v
ari
ab
les
in
the
ex
per
imen
t w
ere
the
in
ter-
arr
iva
l ti
mes
of
sab
syst
em e
ve
nts
, an
d
the
dif
fic
ult
y
of
the
fli
gh
t p
ath
t
o b
e fo
llo
wed
by
th
e p
ilo
t.
In t
he
fir
st
pa
rt o
f th
e e
xp
erim
ent,
o
nly
th
e s
ub
svst
em
mo
nit
ori
np
ta
sk
w
as
co
nsi
de
red
. An
"a
uto
pil
otn
k
ep
t th
e
air
pla
ne
on
c
ou
rse
, c
oin
cid
en
t w
ith
th
e
"on
-co
urs
e in
dic
ato
rn m
ark
er.
10
-min
ute
tr
ials
, u
sin
g
30
, 6
0,
and
90
se
con
d
av
era
ge
in
tera
rriv
al
tim
es
(pe
r su
bsy
stem
),
wer
e ru
n
wit
h
ea
ch
o
f s
ix
sub
jec
ts.
(Th
e a
ctu
al
inte
rarr
iva
l ti
me
s w
ere
ex
oo
ne
nti
all
y
dis
trib
ute
d
abo
ut
the
mea
n.)
To
e
qu
all
y
dis
trib
ute
th
e
eff
ec
ts
of
pra
cti
ce
o
bta
ine
d
du
rin
g
a se
qu
ence
o
f tr
ial3
, a
bal
ance
d
de
sig
n
was
em
plo
yed
, su
rh
tha
t th
e
thre
e
tria
ls
wy
e
giv
en
in
a d
iffe
ren
t o
rde
r t
o
ea
ch
su
bje
ct
(se
e T
ab
le
11.
Six
su
bje
cts
p
art
icip
ate
d,
al
l
of
wh
ich
w
ere
mal
e st
ud
en
ts
or
fo
rmer
st
ud
en
ts
in
-ng
ine
eri
ng
.
Du
rin
g t
his
fi
rs
t p
art
of
the
ex
per
imen
t,
two
se
ssio
ns,
th
ree
ten
-miq
ute
tr
ials
~n
ea
ch
, w
ere
run
wit
h e
ach
su
bje
ct.
T
he
fir
st
sea
sio
n
serv
ed
t
o
tra
in
the
su
bje
cts
fo
r th
e
mo
nit
ori
ng
ta
sk;
little
im
pro
vem
ent
in
per
form
ance
w
as
no
ted
d
uri
ng
th
e
seco
nd
se
ssio
n.
(On
ly
the
d
ata
fr
om
th
e
seco
nd
se
ssio
n w
as u
sed
fo
r a
na
lysi
s.)
Dat
a fr
om
th
is p
art
o
f th
e e
xp
erim
ent
co
nsi
ste
d
of
the
fo
llo
win
g:
1.
Tim
e o
f o
cc
urr
en
ce
of
a su
bsy
stem
ev
en
t
2.
Tim
e o
f re
spo
nse
to
th
e e
ve
nt
3.
Tim
e o
f c
om
~le
tic
~n
o
f d
iag
no
stic
ac
tio
n f
or
the
ev
en
t.
Fa
ilu
res
to
co
mp
lete
d
iag
no
stic
a
cti
on
a
fte
r a
n
init
ial
resp
on
se
(in
co
rre
ct
ac
tio
ns)
w
ere
als
o
no
ted
, a
s
we
ll
as
re
spo
nse
s t
o n
on
ex
iste
nt
ev
en
ts (
fals
e a
larm
s).
Th
e se
con
d
pa
rt
of
the
e
xp
eri
me
nt
co
nsi
ste
d
of
bo
th
mo
nit
ori
ng
an
d
co
ntr
ol
task
s.
Aft
er
co
nsi
de
rab
le t
rain
inp
. th
e
su
bje
cts
part
icip
ate
d
in
foc
r fo
rma
l e
xp
eri
me
nta
l se
ssio
ns,
e
ac
h
of
wh
ich
in
vo
lved
tw
o
tria
ls
of
a
bo
ut
15
m
inu
tes
du
rati
on
. F
or
the
f
ir
rt
t
ri
al
c
f
a se
ssio
n.
the
su
bje
ct
was
g
ive
n a
:i
m~
le m
ap
(few
tu
rns
) to
fo
llo
w (
Fig
ure
3
).
In
the
se
con
d
tria
l,
the
c
ou
rse
w
as
mo
re
com
ple
x
(Fig
ure
4
).
Th
e m
ean
ev
en
t in
tera
rriv
al
tim
e
for
the
m
on
ito
rin
g
tas
k
rem
ain
ed
the
sam
e d
uri
ng
e
ac
h
sess
ion
, b
ut
va
rie
d
fro
m s
ess
ion
to
se
ssio
n,
as
in
dic
ate
d i
n T
ab
le 1
.
To
es
tab
lis
h b
ase
lin
e p
erfo
rman
ce f
or
the
co
ntr
ol
task
. th
e f
irs
t se
ssio
n i
nc
lud
ed
on
ly t
he
air
cra
ft c
on
tro
l.
bu
t n
o
mo
nit
ori
ng
. T
he
rem
ain
ing
se
ssio
ns
wer
e ru
n
wit
h
the
th
ree
le
ve
ls
of
mo
nit
ori
ne
w
ork
load
u
sed
e
arl
ier.
T
he
bal
ance
d
de
sig
n
for
va
ryin
g
the
m
on
ito
rin
g
tas
k
wo
rklo
ad
was
a
lso
u
sed
in
th
es
e
sess
ion
s (T
ab
le
1).
P
rio
r to
per
form
inn
th
e
fin
al
tria
ls,
the
s
ub
jec
ts
pra
cti
ce
d
"fly
ing
" th
e
air
pla
ne
si
mu
lato
r,
bo
th
wit
ho
ut
and
wit
5
mo
nit
or
in^
task
s.
Whe
n th
ey
w
ere
ab
le t
o f
ly t
he
co
mp
lex
map
, w
ith
th
e
low
est
(30
se
c.)
su
bsy
ste
n
ev
en
t in
tera
rriv
al
tim
e,
and
a
lso
ma
inta
in
co
ntr
ol
of
air
r .a
ft
att
itu
de
an
d
po
siti
on
, th
e f
ina
l tr
ials
w
ere
run
.
Dat
a o
bta
ine
d
fro
m t
his
pa
rt o
f th
e e
xp
erim
ent
(sam
ple
d
ev
ery
0
.25
se
c.)
in
clu
de
d
the
th
ree
mea
sure
s li
ste
d a
bo
ve,
a
s w
ell
as
co
ntr
ol
tas
k i
nfo
rmat
ion
:
4.
Air
cra
ft p
osi
tio
n a
nd
a
ttit
ud
e
5.
Pil
ot
co
ntr
ol
inp
uts
.
Per
form
ance
da
ta f
or
fou
r le
ve
ls o
f m
on
ito
rin
g w
ork
load
an
d
thre
e
lev
els
o
f c
on
tro
l ta
sk
w
ork
load
w
ere
thu
s
ob
tain
ed
fr
om
th
e e
xp
erim
ent
as
a w
ho
le.
RES
ULT
S
Th
e ra
w
da
ta
ob
tain
ed
d
ur
in
~ t
he
ex
per
imen
t w
as
anal
yze
d
to o
bta
in
sub
syst
em e
ve
nt
se
rvic
e t
ime
and
wa
itin
g
tim
e s
tati
sti
cs
, su
mm
ariz
ed
in
Ta
ble
2
. E
rro
rs
(fa
lse
a
larm
s an
d
inc
orr
ec
t a
cti
on
s)
wer
e a
lso
zo
un
ted
, an
d
ar
e
show
n in
Ta
ble
s 3
and
4
. T
ab
le 5
sho
ws
tnre
e
mea
sure
s o
f c
on
tro
l ta
sk
p
erfo
rman
ce
(RM
S p
erp
en
dic
ula
r d
ista
nc
e
fro
m
the
co
urs
e,
RMS
pit
ch
an
d r
oll
a
ng
les)
, fo
r th
e
sub
syst
em
ev
en
t a
rriv
al
rate
s
and
c
ap
s u
sed
in
th
e e
xp
erim
ent
(als
o
pre
sen
ted
g
rap
hic
all
y i
n F
igu
re 5
.)
Su
bsy
stem
e
ve
nt
serv
ice
ti
me
is
m
easu
red
fr
om
in
itia
tio
n
of
dia
gn
ost
ic
ac
tio
n
to
th
e
com
ple
tio
n
of
the
a
cti
on
, p
rov
ided
n
o o
the
r su
bsy
stem
ac
tio
n
inte
rve
ne
s (a
fte
r a
n
inc
orr
ec
t a
cti
on
, a
dif
fere
nt,
h
iph
er-
pri
ori
ty
sub
syst
en
e
ve
nt
cou
ld
be
serv
ice
d).
F
igu
re
6 sh
ow
s th
e
es
pir
ica
l s
erv
ice
ti
me
s
tati
sti
cs
. T
he
av
era
ge
s
erv
ice
tim
e
ap
pe
ars
tn
be
ind
ep
en
de
nt
of
co
Ss
y-t
e~
ev
-t
~r
riv
a?
*ate
. A
s d
i~c
us
se
d In
TY
E
WD
EL
s
ec
tio
n,
tt
~e
1,r
ee
au
tio
n o
f s
ub
sy
ste
r s
erv
ice
b
y
the
c
on
tro
l ta
ck
i
s r
efl
ec
ted
as
a
n
inc
rea
se
i
n
the
su
bs
vs
tec
s
erv
ice
yim
e.
rb
is i
qc
rea
se
i~
a
pp
are
ntl
y
no
t a
fu
nc
tio
n
of
~o
ntr
o! :;sk
?if
f~
cu
lty
, bu
t s
imp
ly o
f w
he
the
r o
r n
ot
the
co
nt-
ol
tas
c
is
pre
se
nt.
Th
r s
tib
sy
ste
m
ev
en
t w
ait
ing
ti
me
is
me
as
llre
d
fro
m
the
ti
me
o
f
oc
cu
rre
nc
e
.?f
en
e
ve
nt
to
th
e
tim
e
at
wh
ich
d
iap
no
~ti
c ac
tio
n fo
r
'ta
t e
ve
nt
is
co
~o
lete
d.
Wa
itin
a t
ime
i
s t
he
sum
o
f t
wo
te
rm:$
: re
sp
on
se
ti
me
, th
e t
ime
fro
n e
ve
nt
oc
cu
rre
nc
e t~
init
iat.
i?r.
caf
d
iaa
no
sti
c
ac
tio
n;
an
d
se
rvic
e
tim
e,
fro
m
init
~q
tio
n to
c
om
ple
tio
n
of
dia
nn
os
tic
a
cti
on
. A
s sh
ow
n
in
Fip
ure
7,
wa
itin
g
tim
e
inc
rea
se
s
wit
h
s~
~tt
sy
ste
m
ev
en
t a
rriv
al
rate
. T
his
i
s d
ue
t
o t
he
q
ue
ue
inp
e
ffe
ct
on
th
e r
es
po
ns
e
tim
e.
As
ex
pe
cte
d,
wa
itin
p
tim
e
a1
30
in
cre
as
es
u
he
n
the
c
on
tro
l ta
sk
is
ad
de
n.
Th
is
is
du
e
in p
art
to
th
e
inc
rea
se
d
se
rvic
e
tim
e
uh
en
th
e
co
ntr
ol
tas
k
is
p
res
en
t.
Yo
we
ver,
la
rae
r s
erv
ice
ti
me
a
lon
e
do
es
n
ot
as
co
un
t fo
r
al
l o
f th
e
inc
rea
se
in
wa
itin
g
tim
e.
Th
e
co
ntr
a1
ta
sk
a
lso
aff
ec
ts th
e
res
po
ns
e
tim
e.
Th
is
is
th
e
q~
~e
ire
in~
e
ffe
ct.
o
f th
e
co
ntr
ol
tas
k
"cu
sto
me
rsw
or
the
su
bs
~s
tzm
se
rvic
e ,
,erf
orm
an
ce.
Th
is
eff
ec
t i
s a
lso
o
nly
7
fu
nc
tid
n
of
the
m
ere
p
res
en
ce
o
f th
e
co
n:r
ol
tas
k,
rath
er
tka
n
the
c
on
tro
l ta
sk
d
iffi
cu
lty
, ac
. u
as
th
e e
ffe
ct
on
s
erv
ice
tim
e.
Fa
lse
a
lare
s
(Ta
bla
3)
S
en
d
to
oc
c~
lr le
ss
fr
eq
ue
ntl
y
wit
h
inc
rea
sin
d
co
ntr
ol
tas
k
dif
fic
ult
y
an
d
ev
en
t a
rr
iva
l ra
te.
Whe
n th
e
co
ntr
ol
tas
k
was
a
rrs
en
t,
feu
fa
lse
a
larm
s
ue
re
mzd
e;
ma
ny
of
tho
se
w
hic
h
oc
cu
rre
d
we
re
pro
ba
bly
s
imp
ly
typ
inp
m
ista
ke
s
mad
e i
n r
es
po
ns
e
to
re
al
su
bs
ys
tem
e
ve
nts
(
this
co
uld
n
ot
he
d
ete
cte
d).
F
als
e a
larm
s w
as
ted
o
n
the
ord
er
of
1 s
ec
. o
f t
ime
, a
nd
p
rob
ab
ly
ha
d
ins
ign
ific
an
t e
ffe
ct
rq
s
ub
sy
ste
m t.
ank
pe
rfo
rma
nc
e.
1n
:orr
ec
t a
cti
on
s
(Ta
ble
4)
ten
de
d
to
inc
rea
se
u
ith
s
ub
sy
ste
m
ev
en
t a
rr
iva
l ra
te,
bu
t sh
ow
ed
n
o
co
ns
iste
nt
va
ria
tio
n
uit
h c
on
tro
l ta
sk
d
iffi
zu
ltv
. In
co
rre
ct
ac
tio
ns
a
re
si~
nif
ica
nt
be
ca
us
e,
on
th
e
av
era
ae
, o
ne
-ha
lf
the
a
ve
rag
e
se
rvic
e
tia
e
is
w
as
ted
. F
urt
he
rmo
re,
on
ce
a
n
inc
orr
e-t
a
cti
on
i
s
made,
the
p
ilo
t
ca
n
sw
itc
h
to
a
h
iph
er-
pri
ori
ty
ev
en
t,
wit
ho
ut
co
mp
leti
np
th
e
dia
an
os
tic
a
lre
ad
y
be
gu
n.
Th
is
co
rrtr
ibu
tes
u
ne
ve
nly
t
o
the
w
ait in^
tim
e
ac
ros
s
the
su
bsy
st.e
ms;
s
uc
h
an
o
cc
urr
en
ce
a
uri
na
s
erv
ice
o
f s
ub
sy
ste
?l
6,
for
e
xa
mp
le,
mia
ht
re
su
lt
in
th
at
dia
~n
os
tic
ac
tio
n
be
inp
d
ela
ye
d
co
ns
ide
rab
ly,
wh
ile
s
imil
ar
pre
em
pti
on
du
rin
f!
se
rvic
e o
f s
ub
sy
ste
m 2
wo
uld
n
ot
re
su
lt
in
a
s
?o
ne
a
d
ela
v.
Th
e
imp
ac
t o
f
the
in
co
rre
ct
ac
tio
ns
i
s
thu
s
to
in
cre
as
e
bo
th
the
a
ve
rag
e
an
d
sta
nd
ard
de
via
tio
n o
f
wa
itin
? t
ive
fo
r
low
er-
ort
ori
ty
pro
ce
ss
es
.
Ea
ch
da
ta
ite
m
fro^
the
re
sl~
its
of
th
e s
irn
ula
tio
n t
o be
dis
cu
ss
ed
h
ere
i
s
ba
ae
d
on
1
n.0
00
s
imu
late
d
ev
en
ts.
As
mig
ht
be
e
xp
ec
ted
, fo
r ?ill th
ree
s
ctr
of
sim
ula
tio
n r
es
ult
s,
the
co
mp
ute
r s
imu
lati
on
i
5 m
ore
c
on
sis
telt
[s
ma
lle
r s
tan
da
rd
de
via
tio
n)
tha
n
the
ex
pe
rim
en
tal
su
bje
cts
.
Ta
ble
6
sho
ws
bo
th t
he
em
pir
ica
l s
ub
sy
ste
m w
ait
ing
tim
e
an
d
the
re
su
lts
fr
om
ri
mu
lati
on
wo
d-l
1.
Inp
ut
pa
ram
ete
rs
to
sim
ula
tio
n V
ad
el
1 w
rre
:
1.
Pro
ba
bil
ity
o
f in
co
rre
ct
ac
tio
n,
aiv
en
.P
at
a
s~
~b
sy
ste
m
ev
ert
b
as
o
cc
urr
ed
. T
his
s
as
a
pp
ro-.
ima
ted
fr
om
th
e
ec
pir
ic?
: d
ata
h
y d
iv
id
in
~ th
?
nu
mb
- o
f
inc
orr
ec
t a
cti
on
s
by
th
e n
um
be
r o
f e
ve
nts
.
7.
Fa
lse
ala
rm a
rr
iva
l r
ate
an
d
se
rvic
e
rate
. T
he
s
erv
ice
r
ate
was
a
ssu
me
d
to
be
fiv
e t
ime
s
the
av
era
pe
s
ub
sy
ste
m
ev
en
t s
erv
ice
r
ate
fo
r th
e t
as
k,
be
ca
us
e
on
ly o
ne
o
f t
he
fi
ve
le
ve
ls o
f t
he
c
he
ck
lis
t n
ee
d
be
s
ca
nn
ed
t
o r
ea
liz
e
a
fals
e a
larm
ha
s
be
en
m
ade.
T
he
a
rr
iva
l r
ate
fo
r
fals
e
ala
rms
w
as
ob
tain
ed
b
y
div
idin
pr
the
n
um
be
r o
f
fals
e
ala
rms
b
y
the
to
tal
tin
e d
ur
in
~
wh
ich
fa
lse
a
larm
s c
ou
ld
oc
cu
r.
(Th
e
tota
l ti
me
d
uri
np
w
hic
h
falq
e a
larm
s c
ou
ld
oc
cu
r w
as
eq
ua
l t
o
the
to
t-1
e
lap
se
d
tia
e
min
us
th
e
tota
l ti
me
s
pe
nt
se
rvic
ing
fa
lse
a
larm
s.
sin
ce
fa
lse
a
larm
s c
ou
ld n
ot
oc
cu
r u
hi:e
o
ne
w
as
be
ing
se
rvic
ed
.)
3.
Su
bs
ys
tem
e
ve
nt
ar
riv
al
rate
, o
ne
o
f
the
in
de
ee
nd
en
t v
ari
ab
les
i
n t
he
ex
pe
rim
en
t.
4.
4v
era
ge
s
ub
sy
ste
m
ev
en
t s
erv
ice
ra
te,
ob
tain
ed
fr
om
th
e
em
pir
ica
l s
erv
ice
tim
e d
ata
.
5.
Erl
an
g s
erv
ice
ti
me
d
istr
ibu
tio
n
sh
ap
e
pa
ram
ete
r k,
th
e
sq
ua
re
of
the
ra
tio
of
mea
n t
o s
tan
da
rd
de
via
tio
n.
Th
is
wa
s c
alc
ula
ted
fr
om
th
e
em
pir
ica
l s
erv
ice
t'
me
d
ata
(T
ab
le
2).
Th
is
va
lue
a
ve
raR
ed
19
fo
r
the
m
on
ito
rin
g-o
nly
ta
sk
s,
an
d w
as
3 fo
r a
ll
oth
er
tas
ks
.
A s
ub
sy
ste
m
ev
en
t c
ou
ld
no
t h
e
de
tec
ted
im
me
dia
tely
. R
y a
ve
ra
~in
g th
e
min
imu
m
tim
es
fr
om
e
ve
nt
oc
cu
rre
nc
e
to
init
ial
res
po
ns
e
wh
ich
u
ere
re
co
rde
d
in
th
e
ex
pe
rim
en
tal
tria
ls,
4.5
se
c.
was
e
sti
ma
ted
a
s
the
a
ve
raa
e
tia
e
aft
er
o
cc
urr
en
ce
o
f a
n
ev
en
t b
efo
re it c
ou
ld
be
d
ete
cte
d.
In
th
e
co
mp
ute
r s
imu
lati
on
, e
ve
nts
a
re
de
tec
ted
im
me
dia
tely
; to
a
cc
ou
nt
for
the
de
tec
tio
n t
ime
d
ela
v,
4.5
se
e.
is
rd
drd
i
n
the
s
imu
lati
on
t
o
the
b
as
ic
wa
itin
p
tim
e,
to
g
ive
th
e
ad
jus
ted
wa
itin
g t
ime
.
Su
ma
ari
zin
p
the
re
su
lts
fro
m
Po
de
l i, w
he
n
no
c
on
tro
l ta
sk
i
s p
res
en
t.
the
m
od
el
pre
dic
tio
ns
m
atc
h
the
e
mp
iric
al
da
ta
we
ll.
80
th
mean
an
d
sta
nd
ard
d
ev
iati
on
o
f
wa
itin
g
tim
es
fo
r
ind
ivid
ua
l s
uts
vs
tem
s
are
c
los
e,
Ho
we
ve
r,
uh
en
th
e c
on
tro
l ta
sk
is
pre
se
r~t,
a c
los
e m
atc
h
is
ob
tain
ed
o
nly
fo
r
the
h
i~
h
sub
syst
em
arr
iva
l r
ate
co
n+
itfo
n.
Wit
h lo
wer
a
rriv
al
rate
s,
the
m
odel
p
red
icti
on
s a
re
sma
lle
r th
an
th
e
em
nir
ica
l re
su
lts
. T
his
ca
n
be
e
xp
lain
ed
in
th
e f
oll
pw
ing
w
ay.
Kit
h
hiq
h
arr
iva
l ra
te,
ev
en
ts
oc
cu
r fr
eq
ue
ntl
y
and
m
ost
c
on
tro
l a
cti
on
s
ore
emo
t su
bsy
stem
d
iae
no
sti
cs
in
p
rop
ress
. T
he
co
ntr
ol
tas
k
eff
ec
t is
th
ere
fore
a
lmo
st
en
tire
ly
inc
lud
ed
in
th
e
inc
rea
sed
s
erv
ice
ti
me
fo
r su
bsy
stem
e
ve
nts
. W
ith
;e
wer
a
rriv
al
rate
s,
a s
ipn
ific
an
t o
ort
ion
o
f th
e
co
ntr
ol
sc
tio
ns
ca
n
be
per
form
ed
wit
ho
ut
pre
emp
tin
g
sub
syst
em
dia
gn
ost
ics,
b
ut
may
in
cre
ase
th
e t
ime
t
o
resp
on
d
to
a
Sl~
hS
y~
ten
! ev
en
t.
Th
us,
fo
r th
e
low
er
sub
syst
em
ev
en
t a
rriv
al
rate
s,
the
c
on
tro
l ta
sk
is
n
ot
ad
ea
ua
tely
mo
del
ed
as
sim
ply
an
in
cre
ase
in
se
rvic
e t
ime
fo
r
sub
syst
em e
ve
nts
.
Sim
ula
tio
n
rod
e1
2
, w
hich
u
ses
an
eff
ec
tiv
e
arr
iva
l r
ate
(s
ee
TH
F H
OD
EL
) t
o
ac
co
un
t fo
r tn
e
co
ntr
ol
tas
k,
pro
du
ced
a
good
m
atch
to
em
~ir
ica
l da
ta
(se
e T
ab
le 7
).
Th
e si
mu
lati
on
in
pu
t p
ara
me
ters
a
re
th
e
sam
e a
s
tho
se
abo
ve,
e
xc
ep
t th
at
an
eff
ec
tiv
e
arr
iva
l r
ate
of
sub
syst
em
ev
en
ts
Xeffe
ctiv
e =
ctu
al
+ A
)I
is u
sed
in
ste
ad
of
th
e a
ctu
al
ar
rl
va
~ ra
te.
+
was
o
bta
ine
d
by
vary
in^ A
)L
un
til
a c
los
e
fi
t
to
Ffi
e e
mp
iric
al
wa
itin
g
tim
e
da
ta
was
fo
un
d.
(Sin
ce
Mod
el
1 a
rod
uce
d
an a
de
ou
ate
fi
t t
o e
m~
iric
a?
d
ata
fo
r tn
e
hig
h
arr
iva
l r
ate
c
on
dit
ion
s.
the
e
ffe
cti
ve
a
rriv
al
ra
te w
as
on
ly
com
pute
d fo
r lo
w
and
m
ediu
m
arr
iva
l N
rl
rate
s,
bo
th
sim
ple
an
d co
mp
lex
~
ao
s.)
The
v
ari
ou
s v
alu
es
of
AA
w
ere
aver
aped
(
to g
ive
0.0
09
5).
an
d
this
ev
era
ae
u
sed
t
o c
alc
ula
te t
he
eff
ec
tiv
e a
rriv
al
ra
tes
use
d
to
pro
du
ce t
he
re
su
lts
su
msa
-ize
d
in T
ab
le 7
.
Th
e re
su
lts
of
sim
ula
tio
n W
ode?
3
are
sho
wn
in T
ah
le 8
. T
he
pro
ba
bil
ity
o
f in
?o
rre
ct
ac
tio
n,
fals
e
ala
rm
arr
iva
l r
ate
an
d s
erv
ice
ra
te,
and
su
bsy
stem
e
ve
nt
arr
iva
l r
ate
are
th
e
sam
e o
ara
me
ters
a
s
for
Yo
del
1.
T
he
sub
svst
em
ev
en
t s
erv
ice
r
ate
was
th
e
av
era
ge
o
f th
e
thre
e
rate
s
ob
tain
ed
fr
om
th
e
ba
seli
ne
m
on
ito
rin
p-o
nly
ta
sks.
T
he
Erl
anfz
sh
ap
e
pa
ram
ete
r k
was
th
e a
ve
rag
e o
f th
e t
hre
e v
alu
es
ob
tain
ed
fo
r th
e m
on
ito
rin
g-o
nly
ta
sks.
Th
e a
rriv
pl
ra
te o
f c
ust
om
ers
in
th
e c
on
tro
l ta
sk
qu
eue
A w
as
ca
lcu
late
d
fro
m
the
nu
mbe
r 3
f d
isti
nc
t c
on
tro
l a
ctf
on
s co
un
ted
(N
),
the
ela
pse
d
tim
e
of'
the
co
rre
spo
nd
ing
tr
ials
(T
e),
and
t
6~
a
ve
rag
e d
ura
tio
n o
f th
e c
on
tro
l a
cti
on
s
(Td
) (
all
em
pir
ica
l q
ua
nti
tie
s)
us
in^:
(Ac
tua
lly
, o
cly
a
ile
ron
c
on
tro
?
inp
uts
w
ere
cou
nte
d;
elc
va
tor
inp
uts
w
ere
infr
eq
ue
nt
and
o
f ~
ho
rt
du
rati
on
co
mp
ared
to
afl
ero
q i
nn
uts
. S
ee
F
ipu
re 8
for
en
ex
amp?
e o
f a
ile
ron
cc
ntr
ol
inp
ut
vs.
ti
me
f
or
on
e t
ria
l.)
T
he
se
rvic
e
ra
te f
or
the
co
ntr
ol
tas
k
qu
eue
was
v
ari
ed
u
nti
l c
los
e
fi
ts
t
o e
mp
iric
al
wa
itin
g
tim
e
av
era
ge
s (a
cro
ss
all
su
bsy
ste
ms)
w
erP
o
bta
i~e
d.
The
s
erv
ice
ra
tps
fo
r si
mp
le
map
w
ere
the
n
av
era
ge
d,
as
w
ell
a
s t
ho
ye
fo
r th
e c
om
ple
x
mao
. 9
n
Er
la
n~
sha
pe
pa
ram
ete
r K
of
2
was
u
sed
'o
r th
e c
on
tro
?
tas
k
qu
eue
se
rvic
e
tim
e
dis
trib
uti
on
. T
he
se
va
lue
s w
ere
use
d
to
pro
du
ce t
he
re
su
lts
sho
wn
in T
ab
le 8
.
It
is
imp
crt
an
t t
o n
ote
th
at
the
re
su
lts
o
bta
ine
d
fo
r
this
th
ird
m
od
el
wer
e h
ase
d o
n
the
ass
um
pti
on
th
at
co
ntr
91
ta
sk
se
rvic
e r
ate
do
es
no
t v
arv
wit
h s
uh
syst
em e
ve
nt
arr
iva
l ra
te.
Th
is
assu
mp
tio
n
may
h
e
ina
pp
rop
ria
te.
Yow
ever
. i
f
co
ntr
ol
tas
k
se
rvic
e
ra
te
va
rie
s
wit
h
bo
th
co
ntr
ol
tas
k
dif
fic
ult
y
and
su
bsy
stem
e
ve
nt
srr
iva
l ra
te,
on
e
co
uld
a
rbit
rari
ly
mat
ch
alm
ost
an
y
da
ta.
Wha
t is
n
eed
ed
is
a m
etho
d o
f d
ete
rmin
iw t
he
co
ntr
ol
tas
k s
erv
ice
ra
te d
ire
ctl
y
fro
m
the
c
on
tro
l si
gn
als
. T
his
is
th
e
po
st
imp
ort
an
t e
xte
nsi
on
n
eed
ed
by
this
qu
eu
ein
g m
od
el.
SUM
MAR
Y AN
D C
ON
CLU
SIO
NS
Th
e s
uit
ab
ilit
y o
f th
ree
alt
ern
ati
ve
qu
eu
ein
p m
od
els
of
the
f
lia
ht
man
agem
ent
sit
ua
tio
n
was
in
ve
stie
ate
d,
and
th
e
res
ult
s
pre
sen
ted
. A
tte
nti
on
w
as
foc
use
d
on
p
red
icti
ne
p
ilo
t p
erfo
rman
ce
in t
he
su
bsy
ste
n m
on
ito
rin
*
task
. I
n t
he
f
irs
t m
od
el,
the
su
bsy
stem
m
on
ito
rin
g
tas
k
was
e
xp
lic
itly
m
od
eled
a
s a
W
/Ek/
l:KPR
P/N
/W
qu
eu
ein
e
syst
em.
Th
e c
on
tro
l p
ort
ion
o
f th
e t
as
k w
as
inc
lud
ed
in
th
e e
mp
iric
al
sub
syst
em
ev
en
t s
erv
ice
ti
me
, w
hic
h
inc
rea
sed
w
hen
the
c
on
tro
l ta
sk
w
as p
rese
nt.
T
his
mo
del
a
de
qu
ate
ly p
red
icte
d
the
su
bsy
ste
m
ev
en
t w
ait
inp
ti
me
p
erfo
rman
ce
on
ly
for
the
ta
sk
s
wit
h
hig
he
st
arr
iva
l ra
tes
.
Th
e se
con
d
mo
del
a
cc
ou
nte
d
for
the
co
ntr
ol
po
rtio
n
of
th
e
tas
k
(wit
h
low
an
d m
ediu
m
arr
iva
l ra
tes
) b
y
usi
ng
an
e
ffe
cti
ve
arr
iva
l r
ate
fo
r s
ub
syst
em
ev
en
ts.
By
ad
din
g
an
in
cre
me
nt
of
arr
iva
l r
ate
to
th
e a
ctu
al
ev
en
t a
rriv
al
rate
, th
e e
mp
iric
al
wa
itin
g
tim
e
per
form
ance
c
ou
ld
be
p
red
icte
d
sa
tis
fac
tori
ly
by
com
pu
ter
sim
ula
tio
n.
How
ever
. us
ink?
a
n
eff
ec
tiv
e
arr
iva
l r
ate
le
av
es
muc
h t
o b
e
de
sir
ed
fr
om
th
e
po
int
of
vie
w o
f w
hat
is
ph
ysi
ca
lly
hap
pen
ing
in
th
e o
ve
rall
ta
sk. T
he
thir
d
mo
del
in
co
rpo
rate
d
the
c
on
tro
l ta
sk
a
s
a se
pa
rate
, s
pe
cia
l q
ueu
e w
hic
h
cou
'l
pre
emp
t su
bsv
stem
ev
en
t s
erv
ice
w
hen
ever
a
cust
om
er
arr
ive
d.
Th
e ta
sk
w
as
thu
s
mo
del
ed a
s a
K
/Ek/
l:PR
P/N
/W
qu
euei
ng
sy
stem
. An
e
sti
ma
te o
f
the
arr
iva
l r
ate
fo
r c
on
tro
l ta
sk
q
ueu
e c
ust
on
ers
was
m
ade.
T
he
se
rvic
e
ra
te f
or
the
co
ntr
ol
tas
k w
as u
sed
as
a
fit
tin
e
par
amet
er.
9y
v
ary
ing
t
hi
~
pa
ram
ete
r in
th
e
stn
ula
tio
n,
a re
aso
na
ble
fit t
o e
mp
iric
al
da
ta w
as
ob
tain
ed
.
Th
e th
ree
ro
de
ls
pre
sen
ted
re
pre
sen
t su
cc
ess
ive
re
fin
em
en
ts
in
an
a
tte
mp
t t
o i
nc
orp
ora
te
the
c
on
tro
l te
sk
in
to t
he
qu
eu
ein
g m
od
el
of
the
fl
igh
t n
anaR
eaen
t s
itu
ati
on
. T
he
sub
syst
em
mo
nit
orf
np
ta
sk
was
e
as
ily
fo
rmu
late
d
usi
ng
a
qu
euei
ng
d
esc
rip
tio
n.
and
it
s
ba
sic
re
pre
sen
tati
on
w
as
the
sa
me
in
all
th
ree
m
odel
re
fin
em
en
ts.
Th
e th
ird
m
od
el
re
fin
e~
en
t -10s'
clo
se
ly
ao
oro
ac
he
s th
e
rea
l s
itu
ati
on
w
ith
re
sp
ec
t to
re
pre
se
nta
tio
n
of
the
c
on
tro
l an
d
sub
syst
em
3
on
ito
rin
k
tas
ks
; it
is
the
refo
re
intu
itiv
ely
a
pp
ea
lin
g
:a:t
,h0
1~
~5
the
co
ntr
ol
tas
k
pa
ram
ete
rs
are
no
t w
ell
-de
fin
ed
).
The
g
oa
l o
f th
is
rese
arc
9
was
to
d
ete
rmin
e
if
a
qu
eu
ein
a
scde:
woul4
p
rov
ide
an
a
de
qu
ate
d
es
cri
pti
on
o
f p
ilo
t d
ec
isi~
n m
a ki
n^
in
a m
ult
i-ta
sk,
co
ntr
ol
and
mo
nit
ori
ng
s
itu
ati
on
. R
easo
nab
le
pre
dic
tio
ns
of
per
form
ance
of
th
e
suh
syst
efl
ta
sk
w
ere
ob
tain
ed
w
ith
l
it
tl
e o
r no
fin
e
tun
ing
o
f p
ara
me
ters
. T
his
su
ap
ests
th
at
the
m
adel
is
o
rom
isin
a,
and
can
h
e
su
cc
es
sfu
lly
em
plo
yed
n
ot
on
ly
in
the
a
da
~ti
ve
co
mp
ute
r a
idin
g
sch
eme
men
tio
ned
e
arli
er.
bu
t p
erh
ao
s in
a
bro
ad
er
cla
ss
of
sim
ila
r m
ult
i-ta
sk
sit
ua
tis
ns
.
Fc
r e
xa
m~
le, t
he
p
ro~
os
ed
qu
rue
ina
m
odel
c
ou
ld
be
fi
t
to m
1 lt
i-ta
sk
s
itu
ati
on
s a
nd
th
en
th
e
fra
cti
on
s o
f a
tte
nti
on
re
qu
ire
d
by
ea
ch
ta
sk
c
ou
ld
ea
sil
y
be
de
term
ine
d
sin
ce
th
ey
a
re
inh
ere
nt
ou
tpu
ts
of
a
qu
eu
ein
g
form
ula
tio
n.
The
a
dv
an
tag
e
of
this
ap
pro
ach
is
th
at
the
fr
ac
tio
ns
o
f a
tte
nti
on
a
re
no
lon
ge
r fre
e
pa
ram
ete
rs
as
th
ey
a
re
in
co
ntr
ol
the
ory
m
od
els
of
hum
an
de
cis
ion
mak
lng
in
mu
lti-
task
s
itu
ati
on
s
[12
, 1
31
.
REFE
REN
CES
1.
T.
C.
Yem
pe,
"F
lip
ht
Man
Pge
men
t - P
ilo
t P
roc
ed
ure
s an
d
Sy
stem
In
terf
ac
es
fo
r th
e
19
80
-19
90
's."
A
XA
A
pap
er
No.
7
4-1
29
7,
AIA
A
Lif
e
Sc
ien
ce
s an
d
Sy
stem
s C
on
fere
nc
e,
Nov
embe
r 1974.
2.
W.
8.
Ro
use
. "D
esig
n
of
Man
-Com
pute
r In
terf
ac
es
fo
r O
n-L
ine
Inte
rac
tiv
e
Sy
ste
ms,
" P
roce
edin
tqs
of
the
IE
EE
, S
pe
cia
l Is
su
e
on
Inte
rac
tiv
e C
om
pute
r S
yst
em
s,
Vo
l.
63
, N
o.
6,
pp.
84
7-8
57
, Ju
ne
1
97
5
3.
U.
B.
Ro
use
, "H
uman
-Com
pute
r In
tera
cti
on
if
! M
ult
i-T
ask
S
itu
ati
on
s,"
IE
EL
T
ran
sa
cti
on
s
on
Sy
ste
ms,
M
an
and
C
yb
ern
eti
cs,
V
ol.
SM
C-7
, N
o.
5,
pp.
38
4-3
92
, M
ay
19
77
.
4.
Y.
Chu
an
d
W.
R.
Ro
use
, "O
pti
mal
A
dap
tiv
e A
llo
ca
tio
n
of
De
cis
ion
M
akin
g
Re
sp
on
sib
ilit
y
Bet
wee
n
Hum
an
and
C
om
pute
r in
M
ult
i-T
ask
S
itu
ati
on
s,"
to
a
pp
ea
r in
th
e
Pro
ce
ed
ing
s o
f th
e
1977
Inte
rna
tio
na
l C
on
fere
nc
e
on
Cy
be
rne
tic
s an
d
So
cie
ty,
Sep
tem
ber
1
97
7.
5.
J.
ti.
Rla
ke
loc
k,
Au
tom
atic
C
on
tro
l o
f A
irc
raft
an
d
Mis
sil
es,
New
Y
ork
: Jo
hn
W
iley
&
So
ns,
In
c.,
1
96
5.
7.
b.
L.
Kle
inra
n,
3.
Brr
on
, en
d
W.
Y.
Le
vis
on
, *A
C
on
tro
l T
he
ore
tic
A
pp
roac
h
to
ran
ne
d-V
eh
icle
S
yst
ems
An
aly
sis,
"
IEE
E T
ran
sa
cti
on
s
on
Au
tom
atic
C
on
tro
l.
Vo
l.
AC
-16,
N
o.
6,
pp.
82
4-8
32
, D
ecem
ber
1
97
1.
8.
J.
W.
Se
nd
ers
, "T
he
Hum
an
Op
era
tor
8s
a
Yo
nit
or
and
C
on
tro
lle
r o
f M
irlt
ide
~re
e o
f F
reed
om
S
yst
em
s,"
IEF
E
Tra
nsa
cti
on
s on
H
uman
F
ac
tors
in
E
lec
tro
nic
s,
Vo
l.
HF
E-5
, n
o.
1,
pp.
2-5
, S
epte
mb
er
13
64
.
9.
J.
R.
Ca
rbo
ne
ll,
"A
Qu
eu
ein
n Y
odel
o
f M
anv
-In
ntr
u*
nt
Vis
ua
l S
am
plin
~,~
IE
EE
T
ran
sse
tio
ns
on
Hun
an
Fa
cto
rs
in
Ele
ctr
on
ics
, V
ol.
H
FE
-7,
No.
4
, pp.
15
7-t
64
, D
mem
ber
1
96
6.
10
. Y
. C
hu,
wO
pti
aa
l A
da
pti
ve
A
llo
ca
tio
n
of
De
cis
ion
Wsk
inr?
R
esp
on
sib
ilit
y
Bet
wee
n
Hum
an
and
Com
pute
r in
Mu
lti-
Tas
k
Sit
ua
tio
ns
,"
Ph
.b.
the
sis
in
p
rog
ress,
Un
ive
rsit
y
of
Ill
ino
is a
t U
rban
a-C
ham
pei
pn.
1
J.
A.
Wh
ite.
J
. Y
. S
chm
idt.
an
d
C.
K.
Be
nn
ett
, A
na
lysi
s o
f Q
ue
ue
ing
Sy
stem
s,
Nee
Y
ork
: A
cadem
ic
Pre
ss,
Inc
.,
19
75
.
12
. R
. E.
C
urr
y,
D.
L.
Kle
inm
an,
and
W
. C.
H
off
man
, "A
D
esig
n
Pro
ce
du
re
for
Co
n:r
ol/
Dis
pla
y
Sy
~t
ea
r,
~
Hum
an
Fa
cto
rs,
Vo
l.
19
. W
o. 5
, O
cto
be
r 1
97
7.
13,
3.
Bar
on
an
d
W.
H.
Le
vis
on
, "D
isp
lay
A
na
lysi
s v
ith
th
e O
pti
mal
C
on
tro
l M
odel
o
f th
e
Hun
an
Op
era
tor.
*
Hum
an
Fa
cto
rs,
Vo
l.
19
, lo
. 5
, O
cto
be
r 1977.
6.
R.
S.
Wal
den
, "A
O
ueu
ein
g
Mod
el
of
Pil
ot
De
cis
ion
Y
akin
p
in
a
Mu
lti-
Tas
k
Fli
gh
t M
anag
emen
t S
it
~a
ti
on
,~
MSME
the
sis
, U
niv
ers
ity
o
f I
llin
ois
at
Urb
ana-
Ch
amp
aig
n,
Ju
ly
19
77
.
h b l r 1. hxper imrnta l d e s l m
5
C o n t m l t a s k
S i r p l r Cooplex "rp Pap
90 90
60 60
9 0 90
?O 30
60 60
1 0 30
No c o n t r o l C o n t r o l Task t a s k
S impl r C m p l r x S u b j e c t n a p LP
C o n t r o l Task C o n t r o l Task
S imple C w p l e x Yap Yap
?O ?O
30 30
60 60
50 50
90 9 0
90 9 0
Simple coap1.x W.P
f a b l e e n t r i e s a r e mean s c t s g s t c r : e v e n t i ~ t e r a r r i v l l t i r + s ( s c c .
Table 2. S u b s y s t e s e v e n t s e r v i c e performance ( e h p i r i c a l )
Subsvster! s e n t a r r i v a l r a t e ( e v e n t a / s e c ) Subsystem
NO.
1 l e 2 25 3 30 4 24 5 32 6 48
111' 177
1 34 2 43
18 4 5' 5 19 6 31
A 1 1 219
1 40 2 41 3 35 4 46
:i; t A l l 232
289 4 1 1.07 14.62
S i s p l e Pap 44 7.92 2.87 19.69 53 e.49 5.?2 20.56 54 10.01 7.59 24.91 51 9.74 5.20 2P.21 52 8.33 4.23 Y . 9 7 34 e.55 4.43 55.67
28R 9.71 5.?? 29.52
Conclex lrao 41 ?.?7 <.!I? 22.79 U9 12.03 6.94 2i.U6 5 2 ?.28 6.56 27.03 53 7.90 2.44 29.94 47 ? . 9 b . 39.06 ?5 3.71 5.50 52.76
277 9.49 5.79 31.44
Cumulative s t a t i s t i $ , a c r o s s a l l s i x subw'stcns N - Nusber o f %bsystems e v e n t s r e c o r d e d T , a, - Event s e r v i c e t ime a v e r w e , s t a n d a r d d e v i a t i o n ( s e c . ) T:, a, - Event w a i t i n g t i n e a v e r a g e , s t a n d a r d d e v i a t i o n ( a m . )
Table 5. Contml task performance (enoirical)
Subsyrte? event arrival rate (events/seccr.d)
0 :.L!I111 0.01567 3.033?3
Simple Map statistic
Haan poaition error (It. ) 1259 1.352 3@57 2211 RYS position error (It.) 196 1 2?9 1 6135 2638
t4e.n course wror (ft.) 790 1074 1096 1186 RHS course error ( It. ) 1268 1609 17?2 1730
RMS altitude error (It.) 2700 5616 4395 6692 RMS pltch angle (radians) 0.096 9.118 0.134 0.110 W! roll a~le (radilns) 3.2PI O.?OS 0.?2? 0.319
Hean position error (It.) ?958 3247 3443 5203 MS position error (ft. 6526 4081 4209 673 1
Maan course aror ( ft . ) 1626 1638 1467 1921 RnS courrr arcr ( ft . ) 2p ZWl 203? 2611
)!4S altitude wror (It.) 66911 307 ShOZ 5355 ms pitch angle (radians) 0.158 .?. 117 0.136 0.171 RE! roll anele t ratlianr) 0.527 ?.W 0.415 0.406
h b l e 6. Canparison OF anp i r i ca l data and oueurinR Yolcl 1 simulatlon .-rsults
nb.
1 2 3 4 5 6
~ 1 1 ~
1 2 3 4 5 6
A 1 1
I
2 3 4 5 6
A l l
Subsystem event a r r i v a l r a t e (eventa/sec.) 0.0110 O.Clb57 0.03333
Cumulative s t a t i s t i c , across a l l s i x subsvstems am - Podel t w e n t waitinc t i n e averme, s tantard deviation (sec.)
TUe, ox* - Empirical evert waittne t i n e averant, standard deviation (zcc . )
Table 7. Comparison of er .? ir ical :a:* and oueueinu ?'ode: 2 simulation r e s u l t s
Sutwy-te? w e p t e'frctive a r r iva l r a t* (eventslsec.) ( 1-emect:ve ' b actual 0.0°95)
Subsystem 0.3226 0.0262 NO.
1 2 3 4 5 5
9?11
1 3 7 4 5 6
A l l
Simple ?'aF "we Tm
6 .?a 18.89 C.69 21 .w 5 . k 6 ?? .4@ 12.50 2F. 17 2;.&3 26.97 26.39 53.04
cumulative s t a t i s t i c , scrozs e l l r l r subsv~terrs T,, ow, - Yodel i event waitinp t i - e werape, standard deviation ( sec . ) Twe, a,, - Emniricrll event wnitinc t i -e averare, standard deviation ( sec . )
Aeffective - Fffect ive s;rsv?tes event a r r ivn l r a t e used in s i su la t ion
Aactsal - Pctual sutsvste- event e - r i v a l r a t e used in experiment
Table 8. Cm~ariron of empirical data md oueueinr Kodel 3 aireulation results
Subsystem event arrival rnte (evrnts/sec.) (p, n 0.181, ks = 19 for all tasks.)
Zublystem No.
1 2 3 4 5 6
~ll'
1 2 3 4 5 6
All
Simple Pa@ ( kc s C.15, F,: = 0.20, kc = 2)
"m "ue Tw "m %e "we 'b a, =we .e.54 19.17 6.79 20.35 8.75 19.62 7.26 21.74 8.84 20.85 9.65 18.99 8.69 22.19 10.55 29.56 9.75 24.90 11 .K 23.42
11.32 23.92 8.46 23.99 12.58 24.91 12.35 30.68 16.2c 27.81 13.99 24.87 12.50 27.57 17.06 28.21 18.15 43.05 28.V 32.68 17.04 31.43 24.63 32.34 ?3.5? 3.F? 37.04 70.40 60.58 54.08 22.65 34.1e 26.39 b0.75 35.55 55.67 55.&0 138.07 139.16 139.39
Corplex Vap 9.21 22.21 8.57 21.20
10.02 20.7u 8.19 22.82 11.75 25.1C 14.80 25.05 13.26 24.1C 12.63 29.76 16.88 35.5b 27.72 34.38 21.63 47.86 ?4.61 4?. 3! 14.69 27.06 19.97 2n. 35
Cumulative statistic, &cross all six subsystems ?, , a, - Hodel ? event uaitir.: tise averam, staMard Ctviatlon 'see.) rue, a,, - Empirical event vai:!?g tise averrre, standard cevis;i~ (zec.)
1,. pc, % - Control task cus'.omer arrival rate and servi?e re:. (customars/sec.), and Erlrnt distribution shape parapeter
ps, k, - Subsystem event se?vice rate (events/rec.), and Frlanv Cistr!bution shape parameter
0)
- 0
-. -
c
n
No
Con
trol
Tos
k Y
- W
A
C
ontr
ol T
ask
with
Sim
ple
Map
t
0
Con
trol
Tas
k w
ith
Ccm
plex
Map
-
0
- 0 l
i 1
I 0.0
1
0.02
0.0
3
N
Arr
ival
Rat
e (e
vent
s/se
c)
W
0
No
Con
trol
Tas
k A
C
ontr
ol T
osk
w~
th Sim
ple
Mop
>
8
0
Con
trol
Tas
k w
ith C
ompl
ex M
ap
01
I 1
1 I
0.0
1
0.0
2
0.0
3
Arr
ival
Rat
e (e
vent
s/se
c 1
sP.m
0
No
Con
trol
Tas
k A
Con
trol
Tas
k w
ith S
impl
e M
ap
0
Con
trol
Tosk
wit
h C
ompl
ex M
op
Arr
ival
Rot
e (e
vent
s/se
c)
ru
.m
F ~ Q ' J - ~ A . Ailercn c c n t r o l stick input