Զѧ
ACTA AUTOMATICA SINICA
1997 23 3 vol.23 No.3 1997



ϵͳĽṹ
ʶ1)
ġɭ־
ժҪϵͳMarkovطµõMarkov. ڴ˻ϣƵɫ¶ϵͳKroneckerṹĵƱʶ㷨. 
ؼʡϵͳ ṹϵͳʶ . 
IDENTIFICATION OF STRUCTURE AND PARAMETER INVARIANTS 
FOR MULTIVARIABLE STOCHASTIC LINEAR SYSTEMS
HU DEWEN WEN XISHENG WANG ZHENGZHI
(Dept. of Automatic Control,National University of Defence Technology, Changsha 410073)
Abstract Firstly, this paper introduces the optimal input vector for identifying Markov parameter matrices of multivariable linear systems with correlation analysis method. Based on all of the estimated Markov parameters, then, a recursive algorithm is derived for identifying the Kronecker invariants and parameter invariants of multivariable linear systems in the presence of colored noises.
Key words Multivariable systems, structure invariants, system identification, parameter estimation.
1
ϵͳĽṹʶһֱǱʶ۽Ĳ⣬ Guidorzi1ֱȷ(MIMO)ϵͳĽṹ. ⷽ棬 ģ26ݲòͬо. Ľоɫ¶ϵͳKroneckerṹıʶ. ȫMarkovֵΪݣ ߽ṹʶͲƾ. ͬʱ Ƶṹ㷨Լټ. 
2Markovıʶ
ɢ(MIMO)ϵͳ
y(t)=G(z)u(t)+v(t)(1)
y(t)Rp,u(t)RqֱΪϵͳ룻{v(t)}Ϊƽȵֵ۲v(t)Rp. 
ģ6ݵļ£(1)ʽϵͳɲģ
(2)
G(),=1,,NsϵͳMarkovNsΪϵͳĵʱ. غ
(3),(4)
ģ6֤˲ط
(5)
ϵͳMarkovĸԪɵﵽһµ߱ʶ. ʱMarkovĹֵʽ
()=Ryu().diag(P-11,,P-1q).(6)
ģ6ݻ{u(t)},ʹ(5)ʽϸ. ͬʱùڵεĽ()ĽԣƫԡԡǿһԵ. Ľṹʶȫ() (=1-Ns)Ϊݽе. 
3ϵͳĽṹʶ
״̬ռһϵͳķʽΨһ. ϵͳܹ۲£Lueuberger淶. GuidorziĹ1ݣʱϵͳȫȼ۵p뵥ϵͳĲַʾڲ(v(t)0,t)ʱ
(7)
yi(i=1-p)uj(j=1-q)ֱΪϵͳyuĵi͵jvijΪϵͳĿɹͽṹvij=vi;ik=v1+v2++vi-1+k viΪϵͳKronecker ǿ̻ԶϵͳṹıʣԪئijkΪϵͳĲ. 
(7)ʽУ1Ns-viʱ֤
(8)
gim() ΪMarkovG()ĵimԪ. (8)ʽֻijkԪأ ɼ򻯽ṹʶ. 
µ
i()[gil(),,gi1(+L), gi2(),,gi2(+L),,giq(),,giq(+L)]T,(9)
L=Ns-(vi),(vi)Ǹ֪ʶƵĸviֵֵҪNs2(vi).
˳i()
1(1),,p(1);1(2),,p(2);1(3),,p(3).(10)
еǰأ򲻼. Ŀǰѡl򹹳ɾl ӦĵiĲijklάli.
ڽеl+1ѡ. ѡm(lm+1)µѡ
l+1=[lm(lm+1)],(11)
Цli(i=1-p)Ϊָ룬ʾڦli(.). 
(8)ʽзʱС˷õlĹֵ
lm=lTlm(lm+1), l=(Tll)1.(12)(13)
ǰlѡУlлأTllǷģl. 
ԦlǷά. (11)(13)ʽ
(14)
ʽıֵΪо
(15)
÷ֿʽʽȼ۵
Jm(lm)=Tm(lm+1)[m(lm+1)-llm].(16)
ͬʱ÷ֿ湫ʽ
(17)
ѡm(lm+1)ܷѡ(10)ʽУҪǰѡǷأ ҲҪl+1Ƿ?ȼ۵Ҫl+1Ƿ?det-1l+1=0Ƿ?det-1l0ʿ(15)ʽжJm(lm)=0Ƿ. 
ڱʶ(9)ʽĸԪij(.)棬 ʱ(8)ʽз. (12)ʽõlmС˹ֵlm֪(16)ʽJm(lm)ʵϾǹƲвƽ. ʱжJm(lm)0Сȫ. ʽֵоȫȼ. 
ڿɽб. Jm(lm)0ǰJm(.)ֵJm(lm)СKroneckerʣȷvm=lm mӦĲҲ(12)ʽȷ. ʱ(10)ʽѡУm(.)ѡ(Jm(.)0). lmļ̶. 
Jm(lm)0Բ l+1ѡ ִ(16)(17)ʽ.l1lm1. 
ѡµmֵlmѱ̶mֵ. (12)ʽж(16)ʽǷִ(17)ʽ. ѭ ֱplm̶ֵΪֹ. ʱϵͳȫṹȷ. 
(֤)²ֵ
(18)
k=1-vi, i=1-p, m=1-q, ik=v1++vi-1+k.
4
ϵͳ15
(19)
óΪ1023Ϊ1A-6AST-386CԽз. 
1) ʱɱ㷨õ
J2(0)=0.093582, J2(2)=0.070152, J1(1)=1.335807,
J2(1)=0.091158,J2(3)=0.000000,J1(2)=0.000000.
ɳжϳv1=2,v2=3ֵΪ
111=0.500000, 112=-0.250000,
121=0.020000,122=0.000002,
211=-0.020001,212=0.000001,
221=0.200000,222=0.000000, 223=0.400000.
2) y(t)ϵɫ
v(t)=(t)-0.5(t-1),(20)
Ц(t)-N(0,diag(0.25,0.25)).
ʱ жԵõv1=2,v2=3ֵΪ
111=0.484170, 122=-0.254957,
121=0.061712,122=-0.018487,
211=0.043676,212=0.030954,
221=0.208587,222=0.025213, 223=0.414969.
5ᡡ
Ȳ(6)ʽõϵͳMarkov. ԴȫΪ(9)ʽһ׽ṹṹƹʽ(12)(16)(17)֮׵Ľṹо(16) ǰ5ʽ. ͨڵĴ÷(ɫε)ϼˣҾзǳߵĽṹͲƾ. ֤Kroneckerѡȷ£Ĺֵǿһֵ. 
1)õȻѧͺʡ. 
ߵλѧѧԶϵɳ410073
ο
[1] Guidorzi R. Canonical structure in the identification of multivariable systems. Automatica, 1975, 11:361-374.
[2] Tse E, Weinert HL. Structure determination and parameter estimation for multivariable stochastic linear systems.IEEE Trans. on Automatic Control, 1975, 20:603-613.
[3] Suen L C, Liu R. Determination of the structure of multivariable stochastic linear systems. IEEE Trans. on Automatic Control, 1978, 23:458-463.
[4] El-Sherief H, Shinha N K. Determination of the structure of a canonical model for the identification of linear multivariable systems. Proc. of the 5th IFAC Symp. on Identification and System Parameter Estimation, 569-576.
[5] 壬¬.ϵͳĵƱʶ㷨.Զѧ19817(4):569-576. 
[6] Hu Dewen. Optimal input design for identifying the Markov parameters of MIMO systems. Preprints of the 11th IFAC World Congress,1990, 3:192-197.
ոڡ1993-09-10
