You are on page 1of 3

WP11 = 4:30

Proceedingsof 23rd Conference


on Decision and Control
Las Vegas, NV, December 1984

9.A.P. Blom
NationalAerospaceLaboratory
NLR
D.O. Sox 90502, 1036 31.1Anstcrdan
The N e t h e r l a n d s
Abstract
For a l i n e a r d i s c r e t e t i m e s y s t e r . w i t h X a r k o v i a n
c o e f f i c i e n t s a new f i l t e r i n g a l g o r i t h m i s given,which
is called the interacting Multiple
Model (IN.:)a l g o r i t h n .
The m a t h e m a t i c a l s u p p o r t f o r t h i s a l g o r i t h m i s o u t l i n e d
anda q u a l i t a t i v e c o m p a r i s o n w i t h o t h e r
known f i l t e r i n g
a l g o r i t h m s i s made. The mainconclusion i s t h a t t h e
r a t i o betweenperformanceandcomputational
comp1exi:y
i s far b e t t e r f o r t h e M
IM a l g o r i t h m .
1

Introduction

An i m p o r t a n t p r a c t i c a l problem i s f i l t e r i n g f o r a
l i n e a rs y s t e r .w i t h? < a r k o v i a nc o e f f i c l e n t s .
Tkerefore
t k i s aroblen h a s h a d o o n s i d e r d 9 l e a t t e n t i o n i n
t h e pas;,
mostly ir, a discrete t i m e s e t t i n g .
3;. now t h e r e are sone
f a l i l i e s ofdiscre:e
' ; h e f i l t e r i n g a l g o r i t h m to eone
arbitrary c l o s e t o I.CrSE p e r f o r n a n c e (3ef. 1 ) . % s p i t e of
no; s a t i s f a c t o r y .
all%hosealgorithmthesituationis
t:-le a l g o r i t h m s t h a t coxe
The cor.pu;ationdcor.?lexityof
c l o s e t o KMSE 2 e r f o r n a n c e is often t o o h i g h f o r p r a c t 1 c a l apolicatioon. A t t k e sane time f r o n a mathen;atical
?oLrt ofview some importantfzmi;iesof
:Y.Gs~ a l g o r i t h m
become less relevant k-hen s h e t i r r e - l e g Se:ween x e a s u r e net; updetesbeconessrnallerand
s n a i l e r . Tcese two cons i d e r z t i o z sf o r n e dc h er e a s o nf o r
a f u r t h e r stuEy o f t h e
Troblen. D u r i r gt h e f l r s c phaseof
the s t u d yt h e a t t e n zlon hasbeendirected
to t h e c o r . t i n u c l s % i x e
prcbiez.
m e e x p l o i t a t i o n of t h e powerful t o o l s f o r S%GCkaStiC
-1

d i f f e r e n t i a le q u a t i o n s (Refs. 2-7) y i e l d e da d d i t i o n a l
insighsintofiltering
for systems withI4arkoviancoeff i c i e c t s( R e f .
8 ) . D j r i n g z k e s e c o r & ?:laseof
t h e st.G&y
Li.Ls
i n c l & % f o r n e d tie g i i d e t o o b t a i n s. ne%- d i s c r e t e
new a l g o r i t h m i s called
t i x ef i l t e r i n ga l g o r i t ' h r c .' T h i s
t h e I?*'%
( I c t e r a c t i n g M d t i p i e :dodel) a l g o r i t h x an& i s
2 and 3 . S u b s e q u e n t l yi r ,s e c t i o n
d e f i n e di ns e c t i o n s
a qualitativecomparisonis
made betweenthe i M . 5 aigor i t h m and t h e main o t h e r a l g o r i t h m s .
L1"!

2 The p r o b l e n
The problemconsidered i s f i 1 t e r i r . g an i n c r e a s i n g
i n f o r m a t i o n f i e l d Yt of r e a l v a l u e d o b s e r v a t i o n s on t h e
discrete time line t ,
yt = h ( c !x +g(c,)w,

(1)

where c i s a d i s c r e t e t i m e f i r i t e
Markov chair,assunir.g
,nT(} a c c o r d i n g t o a d e t e m i n i s t i c
v a l u e s i n M={z
transitionpro.Aabilitymatrix
A , t: 1 s d i s c r e t e t i m e
s t a n d a r dw h i t eG a u s s i a r .n o i s ea n 3
x i s a realvalueprocess which s a t i s f i e s t h e s t o c h a s t i c d i f f e r e n c e
equaticr.,
xt = a ( c t ) x t - , + - b ( c t ) v t
(2)

,. .

where v i s d i s c r e t e ;ir.e s t a n d a r dw h i t eC - a u s s i u lr - o i s e .
R r t k e r w and v are r n u t u a l l y i n d e p e n d e n t , h ,g2 ,a and j 2
a r e f i n i t e rea; f u n c t i o n s on I"? and f o r s i n p l i c i t y t i m e i n v a r i a n t . Also r e s t r i c t i o n t o s c a l a r x and y i s f o r :he
sake o f n o t a t i o n a l s i m p l i c i t y only.
3 e f o r e g o i n g t o t h e IhQ4 a l g o r i t h n t a k e a C l o s e r lo&
at t h i s f i l t e r i n g p r o b l e m . F i r s t o f a l l v e r i f y t h a t
a
sufficientstatisticforthesolutionofthefiltering
problem i s a s e t of c l o s e d f o r m e q u a t i o n s f o r t h e
evolu-

~~2093-318410000-0656
$1.00 0 1984 IEEE

656

**I

Branching

I ,

Interaction

Branching

Filter

t A
\
t

557

Filter
for
C

methods f o r t h e o t h e r t r a c k i n g s u b p r o b l e m s t h e
IRN a l g o rithm h a s b e e n e x p l o i t e d t o r e a l i z e
a real-time%racking
a l g o r i t h nw h i c ho u t p e r f o r m ss t a t e - o f - t h e - a r tt r a c k i n g
a l g o r i t h m s (Ref. 1 5 ) . For completeness we have t o mention that to reach this
we needed an a d d i t i o n a l r e s u l t :
t h ee x t e n s i o no ft h es y s t e m
( 1 ) and ( 2 ) t o i n c i u d e
IM a l g o r i t h mf o rs u c h
an
suddenadditivejumps.
The N
extended system will b e p r e s e n t e d i n
a TZturepujlication.

References

'.

T u g n a i t J.K., " D e t e c t i o na n de s t i m a t i o nf o ra b r u p t l y
18 ( 1 9 8 2 ) pp. 607-515.
changingsystems",Automatica,
2 . G i h m a n 1.1., Skorohod A.V., " S t o c h a s t i cd i f f e r e n t i a l e a u a t i o n s " ,S p r i n g e r ,3 e r l i n ,1 9 7 2 .
3 . B r o c k e t t 3.W., B l a n k e n s h i p G . L . , "A r e p r e s e n t a tion theoren for linear differential equations with
M a r k o v i a nc o e f f i c i e n t s " ,P r o c .1 9 7 7 ,A l l e r t o nC o n f .
on C i r c u i t s andSystemsTheory,
Urbana,pp. 6 7 1 ~ 6 7 9 .
4. L i p s t e r R . S . , S'niryayev A . X . , " S t a t i s t i c so f ran.dom
I and 11, Spricger,1977,1978.
p r o c e s s e s " ,P a r t s
G . , " S t o c h a s t i cf i l t e r i n gt h e o r y " ,
5K
. allianpur
S p r i n g e r , 1980.
APPEFDIX P r o o fo ft h et h e o r e m
6. Hazewinkel b!. , Willems J . C . ( e d s .) , " S t o c h a s t i c
s y s t e m s : The m a t h e m a t i c s o f f i l t e r i n g a n d i d e n t i f i Assume at moment t - 1 t h e p r o b a b i l i t y n e a s u r e on
c a t i o n a d a g p l i c a t i o n s " , R e i d e l , 1981
RxM o ft h ep a i r{ x , c }
i s p { x X , C ~ - ~ = ~ } w i t hB an
7. S a r a sJ . S . B
, lankenship
G . L . , h'opkins ( J r . )W.E.,
is a
arbitraryinformationfield.tdifythat{x,c}
"Bistence,uniquenessandasyqtoticbehaviour
of
Markov p r o c e s s w h i c h s a t i s f i e s t h e
Chapmu.-Kolmogorov
s o l u t i o n s t o a classofZakaiequationswith
unboune uation,
PB{ x E ~ , c ~ = n } = n r " i R / p ( ( u , n ) , (B~{, ~m~) -) ~ E d u , =n}
c (15)
d e dc o e f f i c i e n t s " ,
IEE3 T r . on Automatic Control,
t
t-1
28 (19831, pp. 203-214.
w i t h Markov k e r n a l ,
8. Blom H.A.P., "Ikrkov j u n p - d i f f u s i o nm o d e l s
and d e c i s i o n - m a k i n g - f r e ef i l t e r i n g " ,P r o c e e d i n g so ft h e6 t h
p ( ( u , n ) , ( X , m ) ) = PIX EX,^ =r~llx = U , C =n?
(
1
6
)
t
t . t-1
t
i n t . C o n f .A n a l y s i sa n dO p t i n i z a t i o no fS y s t e m s ,
With t h i s t h e f i r s t s t e p i s e v a l u a t i n g t h e k e r n a l , by
J u n e 198L, Eds:
Sensoussan
A., Lions J.L., S p r i n g e r
u s i n g t h a t x s a t i s f i e s ( 2 ) and c i s a Markov c h a i nw i t h
Berlin.
t r a n s i t i o n m a t r i x A,
9 .W i l l s k y
A.S. , Chow E . Y . , Gershvin S.B. , Greene
I;(( u , n ) , ( ~ , m ) ) = P{X-EX~c , = m , ~ ~ - ~ = ~ , c ~ - ~ = n } .
C.S., Houpt P.K.,Kurkjian
A.L. ,"Dynamic modelon
P { C ~ = ~ / X ~ - ~ = U ,= C ~ - ~ = ~ }b a s e d t e c h n i q u e s f o r t h e d e t e c t i o n o f i c c i d e n t s
f r e e w a y s " , IEEE Tr. on A u t o x a t i cC o n t r o l , 25 (1980:,
(17)
= ?(x,-a,m) A F T ,
pp. 347-360.
where F(. ,u,m) i s f o r a l l ( u , ~ ) E R ~a MG a u s s i a n p r o b a b i l 1 C . Smith A.F.:,!. , MaBov J . E . , "Bayesiandetectionand
i t l r measure on R,
e s t i m a t i o n o f jumps i n l i n e a r s y s t e m s " , E d s :
F(X,u,m) = X ~ ( ~ b ( m ) ) - ~ e x p { - ~ ( a ( = ) u - x ) 2 ; - ~( 1~8m
) ~ } . ~J a c o b s O . L . R . , e . a . " h . a l y s i s and o p t i m i z a t i o r . of
s t o c h a s t i cs y s t e m s " , Academic P r e s s , London,1980,
The s u b s e q u e n ts t e p i s t o i n t r o d u c e t h e d e c o m p o s i t i o n ,
pp. 333-345.
Tenney 3 . , " F i n i t e
( 1 9 ) 1 1 . Weiss J . L . , Upadhyay T.!T.,
computa5le f i l t e r s f o r l i n e a r s y s t e r n s s u b j e c t t o
S u b s t i t u t i o n o f ( 1 7 ) a n d( 1 9 )i n
( 1j ) y i e l d s f o r
t ime-varyingmodeluncertainty",Proc.ofthe
P:AEC'3II,
3
1983, pp. 349-355.
P Ict=mbO:
2., V h i t i n g 3.H., Gruber 35. ,"A s u b o p t i m a l
12.
Athans
PBIxt EXlc t =m} = R/F(X,u,m) n&AmP B {ct-,=n}.
estimationalgoritk?withprobajilis'icediting
fsr
falsemeasurementswithapplicationstotargettrack.P3Ixt-lEdulct-l=n}/? 5{c,=ml ( 2 0 )
i n g w i t h wake phenomena", IEEE Tr. on A u t o E a t i c
C o n t r o l , 22 ( 1 9 7 7 ) , pp. 372-384.
Finally by defining ,
13. Chang C . B . , Athans .!?I , " S t a t ee s t i m a t i o nf o r
dis,
c r e t es y s t e m sw i t ks w i t c h i n gp a r m e t e r s " ,
1333 T r .
Y
Y
on Aerospace and 3 : l e c t r o n i c s y s t e m s ,
1L ( 1978),
B(m)=E t - l i x
t-l
, ?+(m)=E t - l { x t c,=m}
L
,
pp. b18-425.
R(m)=EYt-l{( ~ ~ - ~ - 2 ( )r2? )C,-~=TT.} and
1 4 . T u g n a i t J.K. , "Comments on
(Ref.13 )", I E E 3 T r .
on A e r o s p a c ea n dE l e c t r o n i cS y s t e m ,
14 ( 1 9 7 8 ) ,
R+(m)=EYt-l{(xt -2 + ( ~ ) ) ~ l c ~ = n } ,
po. 418-425.
t h e r e l a t i o n s j e t w e e n ( ~ ( I E ,R(m)
)
) and (?+(m) , R + ( n ) ) ,
1 5 . Blorn H.A.P., "A s o p h i s t i c a t e dt r a c k i c ga l g o r i t l x
such as givenby (4) t h r o u g n ( 7 ) , f o l l o we a s i l yf r o m
f o r A i r TrafficControlsurveillanceradardata",
(18) a n d( 2 0 ) .
? r o c .o ft h eI n t .C o n f .
on RaEar, P a r i s , I4ay 1981,
pp.
393-398.
Q.E.D.

.L I

...

458

You might also like