Академический Документы
Профессиональный Документы
Культура Документы
R. Grover Brown
E l e c t r i c a l a n d Computer E n g i n e e r i n g Department
Iowa S t a t e U n i v e r s i t y
50010
Ames, Iowa
ABSTRACT
appropriate
applications.
four
models
state-space
matrix
The
recursive equations.
filtering
"model" i s c o m p l e t e l y s p e c i f i e d by
so-called
parameters
Kalman
for
the
and
initial
conditions
of
the
Once t h e s e a r e d e t e r m i n e d , t h e d i e i s c a s t ,
Thus,
foreverafter.
a model
finding
s i t u a t i o n a t hand i s a l l i m p o r t a n t .
difficult
that
Also,
fits
the
physical
i t i s o f t e n t h e most
a s p e c t of d e s i g n i n g a Kalman f i l t e r .
F o r m u l a t i o n of
descriptions
is discussed.
Finally,
i s pointed
it
out
t o Kalman f i l t e r modeling.
INTRODUCTION
Kalman f i l t e r i n g i s now w e l l known,
nique a r e given i n a
number
of
and t u t o r i a l d i s c u s s i o n s of
standard
references
[1,2,3].
t h e techThe
filter
r e c u r s i v e e q u a t i o n s a r e summarized i n F i g u r e 1 f o r r e f e r e n c e p u r p o s e s h e r e .
+k * % * %* Qk>
It s h o u l d be n o t e d t h a t once t h e i n i t i a l c o n d f r i o n s and t h e
parameters
measurement
are
sequence
s p e c i f i c a t i o n of
effect,
the
specified,
is
die
proceased
t h e s e parameters
the f i l t e r "model",
is
is
cast
and
completely
the
way
determined.
is e s p e c i a l l y important
--
in
which
the
Thus,
the
t h e y are, i n
Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.
1. REPORT DATE
3. DATES COVERED
2. REPORT TYPE
NOV 1984
00-00-1984 to 00-00-1984
6. AUTHOR(S)
8. PERFORMING ORGANIZATION
REPORT NUMBER
Proceedings of the Sixteenth Annual Precise Time and Time Interval (PTTI) Applications and Planning
Meeting, Greenbelt, MD, 27-29 Nov 1984
14. ABSTRACT
see report
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF:
a. REPORT
b. ABSTRACT
c. THIS PAGE
unclassified
unclassified
unclassified
17. LIMITATION OF
ABSTRACT
18. NUMBER
OF PAGES
Same as
Report (SAR)
12
19a. NAME OF
RESPONSIBLE PERSON
1 ENTER
x;,
INITIAL
AND P i
COMPUTE G A I N
r
PROJECT AIIEAD
'ro NEXT STEP
xi+l
+kik
Pi+l
QkPkQk
T
+
UPDATE ESTIMATE
Xk = X i
Qk
Kk(zk - HkXi)
P, = ( I
KkHk)
Pi
F i g u r e 1 Kalman f i l t e r loop
measurement
equations.
The
random
process
under
consideration
is
where
k refers to the
k-th s t e p i n time,
$ k i s t h e transition m a t r i x ,
xk is
a v e c t o r random
and wk i s a G a u s s i a n w h i t e
process,
sequence w i t h a
covariance s t r u c t u r e given by
where
vk
sequence, uncorrelated
with wk '
and
described by t h e covariance
(1)
t h e p r o c e s s i n g o i n g f r o m s L e p k t o k-i-1.
(2)
is
Hk
the
linear
(noiseless)
rnatrix
cotlnectlon
hctween
relationship
the
gives
that
zk
measurement
ideal
the
the
and
process t o be e s t i m a t e d xk.
(3)
Qk d e s c r i b e s the a d d i t i o n a l n o i s e t h a t comes i n t o t h e xk p r o c e s s
i n t h e A t i n t e r v a l hetween .;tc?p Ic a n d k + l .
Rk d e s c r i b e s n d d i t t v e nt:asilrem~?nt
(4)
It
is
important
through
that
note
to
system,
nor
does
(1)
I t i s n o t a n a p p r o x i m a t f a n of
some
to
have
it
Eqs.
discrete
the
( 4 ) s t a n d s i n i t s own r i g h t .
continuous
11oLse.
be
inodel
described
to
related
by
another
contlnuous
Once t h e d i s c r e t e model i s a s s u m e d , t h e
l i n e a r d y n a r n i c a l s y s t e m k n a n y way.
1 follows directly.
r e c u r s i v e e s t i m a t i o n p r o c e s s g i v e n i n Fig.
IMPORTANCE OF THE GAUSSIAN ASSUMPTION
We w i l l d i g r e s s f o r a moment a n d look a t
Eqs.
(4).
(1) t h r o u g h
a n d zk will
If
wk a n d
\tk
he G a u s s i a n p r o c e s s e s .
o f t e n omitted
discussions
111
of
filtering,
n o apology.
really
not
does
make
very
good
for
sense
Gauss-Markov
noise.
The
second
process
which
process
is
has
b e t w e e n +1 a n d -1 at: random p o i n t s
processes
power
are adjusted
spectral
processes!
The l e a s t - s q u a r e s
b o t h cases.
most
likely
wave
they
can h e made
Yet,
they
are
To
Tke f i r s t i s a
typical
uf
which
switches
the two
t o have
identical
radically
different
p r e d i c t i o n f a r o u t i n t o t h e f u t u r e i s z e r o Ear
value.
On
the
o t h e r hand,
z e r o i n t h e random t e l e g r a p h wave c a s e .
waveform i s n e v e r z e r o .
2.
I f t h e p a r a m e t e r s of
t h e mean a n d
predict
functions.
proccsses.
appearance
telegraph
I n time.
appropriately,
density
general
random
the
w e make h e r e w i t h
non-Gausslan
the
4s
t h e mean s q u a r e e r r o r
i l l u s t r a t e t h i s , c o n s i d e r the t w o p r o c e s s e s shown i n F i g .
scalar
t l l e r ~ xk
Even t h o u g h t h e G a u s s i a n a s s u m p t i o n
least-squares
i n
W e would
he b e t t e r o f f
case because z e r o 1s
it
is ridiculous t o
W e know a p r i o r L
that t h i s
t o p r e d i c t e i t h e r +1 o r -1
Thus,
the
Gaussia~l assumption
e s t i m t i o n theory,
is
redsonable
and t o s t r a y f r o m i t l e a d s
11s
one
in
the
least
i n t o dangerous
squares
territory.
SPECTRUM
GAUSS-MARKOV SIGNAL
SPECTRUM
Figure 2
I n Kalman f i l t e r a p p l i c a t i o n s , we f r e q u e n t l y b e g i n w i t h a s p e c t r a l d e s c r i p t i o n of
the various
convert
this i n f o r m a t i o n
through
(4).
random
processes
t o a model
procedure
general
The
involved.
of
the
for
The problem t h e n i s
form s p e c i f i e d
making
the
by
to
Eqs.
(1)
to
the
transition
d i s c r e t e model i s a s f o l l o w s :
(1)
Look
f o r a c o n t i n u o u s dynamical system t h a t
process
when
driven
by
white
noise.
(The
yields
white
the desired
noise
input
a s s u r e s t h a t wk w i l l be a w h i t e s e q u e n c e . )
(2)
(3)
S o l v e t h e s t a t e e q u a t i o n s f o r s t e p s i z e A t and o b t a i n
X
(4)
k-tl - 4kXk
+ w
To i l l u s t r a t e t h e p r o c e d u r e f u r t h e r ,
s u p p o s e t h e y p r o c e s s power s p e c t r a l
d e n s i t y f u n c t i o n S ( s ) c a n be w r i t t e n a s a r a t i o of polynomials i n s2 ( o r
Y
2
2
w , where w = -s ).
The s p e c t r a l f u n c t i o n can then always be f a c t o r e d i n t o
two symmetric p a r t s , one w i t h i t s p o l e s and z e r o s i n t h e l e f t - h a l f
t h e o t h e r with mirror-image
p o l e s and z e r o s i n t h e right-half
s plane,
plane.
i s c a l l e d s p e c t r a l f a c t o r i z a t i o n and i s r e p r e s e n t e d m a t h e m a t i c a l l y as
This
where
4-
4-
and S-
are
t h e l c f t - and right-half
plane
parts
respectively.
S ( s ) t h e n becomes t h e shaping f i l t e r t h a t w i l l s h a p e u n l t y w h i t e n o i s e i n t o
Y
a process y ( t ) w i t h a s p e c t r a l f u n c t i o n S ( s ) .
( S e e Ref. [ l ] f o r f u r t h e r
Y
details. )
Now s u p p o s e t h a t t h e s h a p i n g f i l t e r i s of t h e f o r m shown i n F i g . 3.
a s t a t e - s p a c e model f o r t h a t d y n a m i c a l system.
w(t>
( U n i t y White Noise)
>
m
m- 1
bms + b m - p
t
n
n- 1
s + an-ls
Figure 3
*-
*-
b0
ao
Shaping f i l t e r
4.
The s t a t e - s p a c e e q u a t i o n s a r e then
Define s t a t e v a r i a b l e s as
**
r, t , r , * * *where r i s a n
intermediate variable.
Figure 4
We seek
S h a p i n g f i l t e r redrawn
C o n t r o l system e n g i n e e r s r e f e r t o t h t s a s t h e c o n t r o l l a b l e c a n o n i c a l form,
and i t c a n always be achieved f o r t h e dynamical system as shown in F-lg.
I y i s t h e p r o c e s s t h a t i s a c t u a l l y measured, t h e n t h e
row matrix of b ' s g i v e n i n Eq.
3.
H matrlx is j u s t the
(9).
EXAMPLE
a
Suppose we have
s c a l a r Gauss-Markov
process
~ ( t whose
)
power
spectral
density function i s
We f i r s t f a c t o r S
The
shaping f i l t e r
as follows:
equation
This
is a
simple f i r s t
state variable.
order d i f f e r e n t i a l equation,
Call I t xl.
s o we o n l y have one
Our s t a t e e q u a t i o n Is t h e n
The s o l u t l o n of t h i s e q u a t i o n f o r a s t e p s i z e A t i s
and e
be
We tnight pose a q u e s t i o n a t t h i s p o i n t :
Are Kalman f i l t e r models u n i q u e ?
l e g i th a t e
performing
Optimal
linear
transformation
state
vector.
estimation
the
estimates
can
be
on
The
is
process
transformed
state
the
vector
of
choice
purely
to
leads
coordinate
matter
of
another
for
frame
convenience.
freely
a n o t h e r ( t h r o u g h a l i n e a r t r a n s f o r m a t i o n ) and s t i l l r e m a i n o p t i m a l e s t i m a t e s
i n t h e new frame o f r e f e r e n c e .
ARMA MODEL
Sometimes t h e r.*r~dom p r o c e s s ~nudeL cornes t o
equation
rather
consider
the
than
continuous
auto-regressive
moving
US
i n the f o r m of
differential
average
For
equation.
(AKMA)
model
a difference
that
example,
relates
d i s c r e t e process y ( k ) t o a n i n p u t w h i t e s e q u e n c e u ( k ) .
T h e r e i s a close a n a l o g y between d i f f e r e n c e a n d d i f f e r e n t i a l e q u a t i o n s , a n d
It
works
out
that
this
nth-order
d i f f e r e n c e e q u a t i o n can h e
converted
define a n i n t e r m e d i a t e v a r i a b l e y ' ( k )
a s t h e s o l u t i o n t n Eq.
to
I f we
( 1 6 ) with j u s t
u(k) a s t h e d r i v i n g E u n c t i o n , and t h e n d e f i n e o u r s t a t e v a r i a b l e s as
xl(k)
y O ( k ) , x2(lc)
yO(k+l),
etc.
t h e n t h e system of E q , ( 1 6 ) t r a n s l a t e s i n t o s t a t e - s p a c e form as
(17)
PJote t h a t o u r c h o i c e of s t a t e v a r i a b l e s l e a d s t o t h e c o n t r o l l a b l e c a n o n i c a l
form,
just
as i n t h e c o n t i n u o u s dynamical c a s e .
Of
course,
we
c o u l d have
Eqs.
(18) and ( 1 9 ) .
W
e w i l l not pursue t h i s f u r t h e r o t h e r t h a n t o s a y t h e
c h o i c e of s t a t e v a r i a b l e s i s ( w i t h i n l i m i t s ) a matter of c o n v e n i e n c e f o r the
s i t u a t i o n a t hand.
PROCESSES DERIVED FROM IKKATIONAL SHAPING FILTEKS
The
random
process
straightforward.
modeling
procedures
They may b e t e d i o u s f o r h i g h e r - o r d e r
thus
far
have
processes,
been
b u t they
do n o t c a l l f o r much i m a g i n a t i o n .
p r o c e s s e s where t h i s i s n o t t h e case.
be
discussed
These a r e t h e p r o c e s s e s t h a t cannot
thought of a s t h e result of p a s s i n g v e c t o r w h i t e n o i s e t h r o u g h a L i n e a r
dynarnical
system
of
engineeri-ng l i t e r a t u r e .
finite
order.
Such
processes
are
commonplace
in
For example, b a n d l i m i t e d G a u s s i a n w h i t e n o i s e i s a
v e r y u s e f u l a b s t r a c t i o n i n communication t h e o r y .
It i s G a u s s i a n n o i s e t h a t
It can be t h o u g h t of
t h r o u g h an i d e a l i z e d lowpass
a s a r a t i o of
a s the r e s u l t of p a s s i n g p u r e w h i t e n o i s e
filter,
p o l y n o m i a l s i n s of
f i n i t e order.
f i l t e r can h e made t o a p p r o x i m a t e t h e i d e a l c a s e ,
(Note t h a t a B u t t e r w o r t h
h u t n o t equal it.:)
The
[ 4 ] g l v e s u s t h e f o l l o w i n g c r i t e r i o n f o r Lhe r e a l i z a t i o n of
Chen
linear
dynamical models.
A linear dynarnical model of t h e Eorrn
i s f a c t o r a b l e i n tlze form
i f and o n l y i f , G(t;t)
G(t,r)
M
?~(t)~(,r)
and N a r e f i n i t e - o r d e r
input,
impulsive response G ( t , r ) ,
single-output),
(21)
matrices,
M(t)
so i f G ( t , r )
is
scalar
e . ,
single-
exist
for a
corresponding
factorizarion
provides
the
Fmpulstve r e s p o n s e
necessary
1.nfortnation f o r
(See Chen [ 4 ] f o r f u r t h e r d e t a i l s . )
model.
the
PTTL
because
community
of
Furtllernlurc,
the
realization of
the
We w i l l u s e f l i c k e r n o i s e t o
F l i c k e r n o i s e i s of s p e c i a l i n t e r e s t
to
func~ion.
its
presence
in
precisLon
frequency
It i s c h a r a c t e r i z e d by a power spectral d e n s i t y f u n c t i o n of t h e
standards.
[5,6].
w h i t e n o i s e is g i v e n in Fig. 5.
White
>
Noise
Figure 5
Clearly,
the
Block diagrams r e l a t i n g f l i c k e r n o i s e t o w h i t e n o i s e
transfer
function
phase n ( t ) i s l/s3l2.
response
to
an
impulse
relating
input
white
noise
to
the
output
The i n v e r s e t r a n s f o r m U P l / s 3 l 2 g i v e s t h e i m p u l s i v e
applied
at
t=O.
This
i s 2&lhA.
phase ( t i m e )
F requeircy
w(t>
Thus,
for
an
The q u e s t i o n i s ,
that
it
This
being
is not,
the
"Is G ( ~ , T ) f a c t o r a b l e
although
case,
this
Chen's
i n the
is d i f f i c u l t
theorem says
t o show i n a r i g o r o u s sense.
that
w i l l . e x i s t t h a t c o r r e s p o n d s t o t h e G ( ~ , T )of Kq.
no
finite-order
state
model
will
exactly
form M ( ~ ) N ( T ) ? " I t a p p e a r s
no l i n e a r dynamical
This i s t o say t h a t
(22).
represent
system
flicker
Of
noise!
c o u r s e , t h e s t a t e model i s e s s e n t i a l f o r Kalman f i l t e r i n g , s o t h i s l e a d s t o
T h i s Is t h e s u h j e c t of a companion paper i n t h e s e P r o c e e d i n g s [ 6 ] ,
s o w e w i l l n o t p u r s u e thFs f u r t h e r h e r e .
SUMMARY
V a r i o u s a s p e c t s of Kalman f i l t e r i n g modeling have been d i s c u s s e d briefly i n
t h i s paper.
processes
c o n s i d e r a t i o n must be modeled i n v e c t o r s t a t e - s p a c e
T h i s c a n o f t e n be done w i t h e x a c t methods.
h e r e c a n n o t be used,
t h e e x a c t methods d i s c u s s e d
as i n t h e c a s e of f l i c k e r n o i s e ,
approxtmate f i n i t e - o r d e r
filter.
If
v e c t o r models
form.
[2]
P.
S. Maybeck, S t o c h a s t i c Models,
Academic P r e s s , 1979.
[3]
A.
Gelb ( E d . ) ,
E s t i m a t i o n and C o n t r o l ,
(V0.l.
1),
T. Chen, I n t r o d u c t i o n t o L i n e a r System
Theory,
Winston, 1970.
[ 4 ] C.
and Kalrnan
H o l t , R i n e h a r t and
[5]
[6]
QUESTIONS A N D ANSWERS
VICTOR R E I N H A R D T , I-IUGHES A I R C R A F T COMPANY: I t h i n k you a r e r i g h t
a b o u t t h a t n o t b e i n g a b l e t o be f a c t o r e d , and I t h i n k t h a t I have
a r e a s o n f o r t h a t . You c a n s h o w t h a t f l i c k e r n o i s e c a n b e
m a t h e m a t i c a l l y g e n e r a t e d by t h e s u m o f an i n f i n i t e number o f
g a u s s i a n p r o c e s s e s where t h e b e t a term g o e s from zero t o
infinity. Therefore, there are infinite time constants in the
p r o c e s s . S o , you c a n ' t g i v e 3 s t a t e v e c t o r a t a n y o n e t i m e ,
b e c a u s e t h e b e t a term g o e s from z e r o t o i n f i n i t y .
M R . B R O W N : I a g r e e w i t h w h a t y o u s a y . I t h i n k t h a t i t f i t s my
i n t u i t i o n t o t h i n k t h e same t h i n g , and I h a v e r e a d t h a t paper
t h a t you w r o t e on i t . I t h i n k t h a t i t ' s a v e r y n i c e p a p e r , a n d a
n i c e way t o l o o k a t i t .
Other people have a l s o approximated f l i c k e r noise with a
c a s c a d e d s e q u e n c e o f w h a t we, i n c o n t r o l s y s t e m e n g i n e e r i n g , c a l l
l e a d o r l a g networks, which g i v e s kind of a s t a i r c a s e sort o f
frequency response function, which, t o a c e r t a i n degree of
a p p r o x i m a t i o n , d r o p s o f f a t t e n dH p e r d e c a d e r a t h e r k h a n t w e n t y
dB.
I f you t a k e a n y r a t i o n a l t r a n s f ' e r f ' u n c t i o n , o r one t h a t i s
w r i t t e n o u t i n i n t e g e r p o w e r s , a n d l o o k a t t h e Bode p l o t , t h e
s l o p e s go i n m u l t i p l e s o f t w e n t y d B p e r d e c a d e , There a r e no
t h i r t y dB p e r d e c a d e , o r f i f t y dB p e r d e c a d e s l o p e s .
I n t h e c a s e o f f l i c k e r n o i s e , and c o n s i d e r t h e f i l t e r t h a t
shapes w h i t e n o i s e i n t o f l i c k e r n o i s e , it r e q u i r e s an s t o t h e
n e g a t i v e one-half' power i n t h e t r a n s f e r f u n c t i o n . T h a t would g i v e
a Bode p l o t t h a t d r o p s o f f a t t e n dB p e r d e c a d e i n s t e a d o f
t w e n t y . W h a t y o u would d o i s a p p r o x i m a t e t h a t t e n d B p e r d e c a d e
s l o p e w i t h a w h o l e s e q u e n c e of' f i l t e r s w i t h a l t e r n a t i n g z e r o s and
p o l e s . You t h e n e n d u p w i t h a s t a i r c a s e s h a p e r e s p o n s e w h i c h , o n
t h e a v e r a g e , h a s a t e n dB p e r d e c a d e s l o p e .
I n c i d e n t a l l y , I t h i n k t h a t t h i s i s a v e r y g o o d way t o m o d e l
f l i c k e r n o i s e . The d i f f i c u l t y i s t h a t e v e r y t i m e you p u t a new
p o l e i n t h e s y s t e m y o u h a v e a new s t a t e m o d e l . I f y o u w a n t g e t a
r e a s o n a b l y a c c u r a t e a p p r o x i m a t i o n o f f l i c k e r n o i s e t h a t way, i t
d o e s i n v o l v e e s c a l a t i n g t h e o r d e r o f t h e Kalman f i l t e r
c o n s i d e r a b l y . T h e r e i s n o t h i n g wrong w i t h d o i n g i t o f f - l i n e f o r
a n a l y s i s p u r p o s e s . I t h i n k t h a t t h e r e a r e sorne o n - l i n e c a s e s
where it would n o t be a c c e p t e d .
MR.
I t h i n k t h a t some p e o p l e h a v e r e p o r t e d on
a
method where t h e y used a f i n i t e number o f f i l t e r s and i t
very w e l l i n an o p e r a t i o n a l case. I f you t r y t o l i m i t t h a t
though, what happens is t h a t a l l t h e p o l e s run t o g e t h e r ,
end u p w i t h a branch l i n e .
REINHARDT:
similar
worked
process
a n d you
M R . B R O W N : I g u e s s my a n s w e r t o t h a t w o u l d b e t h a t , i n a n y o f
t h e s e processes, i n t h e case o f f l i c k e r n o i s e f o r example, a t
z e r o f r e q u e n c y and o u t a t i n f i n i t y , t h e r e a r e s i n g u l a r c o n d i t i o n s
for e i t h e r case. If it drops off a s one o v e r f ,
t h e a r e a under
t h e c u r v e o u t a t i n f i n i t y i s n o t f i n i t e . You a r e t a l k i n g a b o u t a
p r o c e s s w i t h i n f i n i t e v a r i a n c e , which i s p h y s i c a l l y r i d i c u l o u s .
The s a m e t h i n g h a p p e n s a t t h e o t h e r e n d o f t h e s p e c t r u m , t h e
JIM B A R N E S , A U S T R O N , I N C . :
I have done a f a i r amount o f
simulation o f flicker noise with polynomials, t h e lead-lag
n e t w o r k s you m e n t i o n e d , and h a v e o n e comrnent i n t h e i r d e f e n s e :
T h r e e o r f o u r s t a g e s c a n d o a n a m a z i n g a m o u n t . You c a n c o v e r a s
much a s t h r e e t o f o u r d e c a d e s o f f r e q u e n c y w i t h o n l y t h r e e o r
four stages.
M R . BROWN: O h , i s t h a t r i g h t ? I t i s n ' t a s b a d a s i t m i g h t a p p e a r
a t first glance then. I haven't used it, but
t h a t you would need a f a i r l y l a r g e number.
would have
imagined
M R . REINHARDT: A s a n o t h e r c o m m e n t , e v e n a s i n g l e f i l t e r , w h i c h
g e n e r a t e s a random telegraph, w i l l g e n e r a t e a f l a t Allan variance
of about two o r d e r s o f magnitude i n t a u , r i g h t around t h e peak.
Then you r e a l l y h a v e t o p u t a p o l e e v e r y o r d e r o f m a g n i t u d e or
even every two o r d e r s o f magnitude.
M R . BROWN: A l l o f t h e s e a r e , o f c o u r s e ,
t h e reasons which I j u s t c i t e d .
approximate models f o r
ALLAN: I t h i n k , i n p r a c t i c e , t h e p r o b l e m w i t h f l i c k e r n o i s e
is n o t a s e r i o u s o n e , b e c a u s e it's o n l y a t t h e e x t r e m e s , a s you
p o i n t e d o u t , a t z e r o and a t i n f i n i t y t h a t you have d i f f i c u l t i e s
MR.