Ϣ
INFORMATION AND CONTROL
1999 28 6 Vol.28 No.6 1999



ϵͳԶо
붬 лκ
ժҪԶָеļȨűջϵͳƵڵĽϵΪһµ÷÷Ҫŵǲ⸴ӵľRiccatiҲɺ׵ȷָջҪ״̬󣮱Ļָջѡ񷽷˵ּ÷ЧԺͼԣ
ؼʣϵͳ LQ⣬ Ȩ Riccatị 
ͼţTP13ױʶ룺B
STUDY ON THE LINEAR QUADRATIC EXPECTED POLE ASSIGNMENT
PROBLEM FOR THE CONTINUOUS-TIME SYSTEM
FENG Dong-qing2 XIE Song-he2
(1. Information & Control Institute, Zhengzhou University of Technology, Zhengzhou 450002; 2. Department of Control Engineering, Zhengzhou Institute of Light Industry, Zhengzhou 450002)
AbstractThis paper presents a new method of the expected pole assignment based on the analytical relationship between the weighting matrices in the linear quadratic performance and the closed-loop systems in the frequency domain. The advantage of the method is that the state feedback matrix satisfying the pole assignment can be easily determined without solving the complex matrix algebraic Riccati equation. The choice of prescribed closed-loop poles is discussed and the effectiveness and simplicity of the method is expounded based on an example.
Key wordscontinuous-time system, LQ problem, weighting matrix, Riccati equation, expected pole assignment
1
һԶɿϵͳ
X=AX+BU(1)
ʽABֱnnnmά
ԶָΪ
(2)
ʽQ0R>0ſ֪ʹָ(2)СſΪ
U=-KX=-R-1BTPX(3)
ʽPRiccati
ATP+PA-PBR-1BTP+Q=0(4)
(1)ʽ(3)ʽɵűջϵͳ
(5)
νԶ⣬Ҫȷһ״̬KʹӦıջϵͳһԤָļcii=12nʹĳָ(2)СĳȨQ0R0KӦ[1]ˣоڣڵϵͳѡıջ㣬Ӷ֤֮ӦΨһŵģڶϵͳԣָ֪һǰ£üõɶȣѰһ㼫Ҫķ森ˣʵǴͳļ÷LQƷۺ⣬ͨۺƷõıջϵͳм÷ö̬ԺLQƵ³Ե˫ŵ[3]
ͨķ֣ԼƶΪֵ[35]ʹLQ۵Ľ[68]ǰһַûдӱϽʾȨQRKԼci֮ĶֵӦϵڵҪԾ󷽳̣ǳӣһַAͦci̶ͬȵҪʵûеõԲһԼƶΪ·صǲҪԵľRiccatịҲԺ׵ȷָջҪŷKԼ֮ӦļȨQRʵּ÷ǼЧģ
2ŷKĴ
Ѿ֪LQſϵͳü㣬Ըһջ㣬һӦQ0R0ʹɴƵLQջϵͳָļ㣮ڶҪθıջ̬ѡָıջ㣻γöϵͳõɶȣȷһ״̬Kʹ֮Ϊŷ󣬲ﵽָջõҪ
1 ڸķKĳȨQ0֮Ӧҽ

ʽF(j)=I+K(jI-A)-1B.


1 oi ,ci(i=1,2,,n)ֱϵͳ1ͣ5Ŀջ㣬²ʽ
(6)
(7)
(8)
ʽ
֤ detF(s)=Pc(s)/Po(s)1֤6ʽͣ7ʽ[9][10]ѵõ8ʽ
1ΪʵʹѡʵıջṩҪݣѡָǰӼϿҪciڦoi(i=1,2,,n)ƽԭԼԶһЩ
2[8] ڵϵͳԣָıջci(i=1,2,,n)һĳnʽͬʱ

ʽi=1,2,,nҵj>nʱ aj=bj=0
3õĻԭ
[1]
FT(-S)RF(S)=R+BT(-SIn-AT)Q(SIn-A)-1B
ʽͬʱȡʽ
Pc(S)Pc(-S)=Po(S)Po(-S)det(9)
H(S)=(SIn-A)-1B=N(S)/Po(S),9ʽɸдΪ
Pc(S)Pc(-S)=Po(S)Po(-S)det(10)
2[11] ci٦cj(ij)Ҧci(A),QRKԼci¹ϵ:
K=-[ƪ1,2n][ƪ12n]-1(11)
ʽзi,ii=12nʽȷ
i=0 i=1,2,,n(12)
i=H(ci)i i=1,2,,n(13)
2ӱϷӳ˼ȨQRKԼci֮໥ϵĳַʽܹõciӦQRô2ֱȷQRӦK⸴ӵķԾ̣4
3 ֪ϵͳAģ̬ΪMMѷֽΪ
(14)

ʽ
(15)
(16)
M-i(i=1,2,,n1)M+j(j=1,2,,n2)ֱʾӦڦ-oiͦ+oj-oiʾAҪֲֵ+ojʾAĽҪƶֵn1+n2=nȨQ
QM-i=0 i=1,2,,n(17)
ôQR>0ȷ״̬Ϊ
(18)
ʽЦi,i(i=1,2,,n2)ֱɣ12ʽͣ13ʽȷҶӦıջϵͳвֵͱƶԺn2Ŀֵ
֤ 2[10]ѵ֤
3 ֪MΪAģ̬
(19)
ô㣨17ʽҪQ
(20)
ʽ
D=DT0.
֤ ΪM-1M=In


ݶ3ͿµԼƶÿһһ㣩Ϊ÷
ٶԭϵͳʽΪ
Po(S)=Poi(S)Pov(S)
ʽPov(S)=0ĸӦҪƶֵPoi(S)=0ĸӦڲֵ
Pov(S)=S2+a1S+a0  Pov(S)=S+a0
ֻһμƶıջϵͳʽΪ
Pc(S)=Pci(S)Pcv(S)
Pci(S)=Poi(S)Pcv(S)Ŀֵ
Pcv(S)=S2+b1S+b0  Pcv(S)=S+b0
ݶ3D=ddTʽd=[d1 d2]Td=d1ˣ[9]

ʽ루9ʽ
Pcv(S)Pcv(-S)=Pov(S)Pov(-S)+dTW(S)WT(-S)d(21)
ʽW(S)=[[AKM^]yN(S)R-1/2]/Pov(S)Pov(S)Pcv(S)ԼW(S)֪ˣɣ21ʽdӶõQָĳһR>0£.ٸ3ȷһKʵĳһһԼƶμķƶnΣɵõҪK
ҪָÿƶеļһҲϣǣÿƶļʱȻƶܴˣÿεļ̾͸ӶˣΪ21ʽչתΪһԷ飮⣬ĸͬһвƶ
4ƶļ㷽
һ ֪Ŀϵͳ1oi(i=1,2,,n)ȻЩʵ飬ÿֻһ㣮
ڶ ݵһĿö12ѡʵıջ㣬з飬Ա֤21ʽdڣ
 j=1,A(j)=A,H(j)(S)=H(S),N(j)(S)=N(S)P(j)o(S)=Po(S),{(j)oi}={oi},ָһȨR>0һȡR=ImͬʱR(j)=R.
Ĳ еj鼫ƶݣ21ʽd(j)ɣ20ʽQ(j)ٸ3K(j)ʱ׼е(j+1)鼫ƶ
A(j+1)=A(j)c=A(j)-BK(j)
H(j+1)(S)=-1B
R(j+1)=R(j)=R
P(j+1)0(s)=P(j)c(s)
{ai(j+1)}={ci(j)}
岽 ȫƵĿλãôָ֤ջҪŷΪ
(22)
֮ӦļȨΪ
(23)
lʾƶתĲ
5[5]

 (A)={-4,-1,-2}
ļΪ(Ac)={-4,-5,-6}ôԷãȽ-2Ƶ-6ٽ-1Ƶ-5-4ֲ䣮
ڽ-2Ƶ-6ʱǿɵ

ѡR=I2

ݣ21ʽɵ

ˣ[d1(1)]2=128



ͬڽ-1Ƶ-5ʱ


ˣd(2)1]2=24900/116,


(24)
֮ӦļȨֱΪ
(25)

(26)
A,B,QR뷽̣4Pݣ3ʽ֤õŷ루24ʽͬЦ(A-BK)={-4,-5,-6} ͳ˵˱ĵȷԣ
߼
붬(1959-)У˶ʿڣоΪſ,ܿ,̼ƣ
лκ(1965-)УڣоΪԶϵͳۣģƣܿƣ
ߵλ붬 ֣ݹҵѧϢо ֣ 450002
лκ ֣ṤҵѧԺƹϵ ֣ 450002
ο
1 Anderson B D O, Moore J B. Liner Optimal Control. Prentice-Hall, London,1971
2 Kalman R E. When is a Linear Control System Optimal. Trans. ASME(D), 1964, 86(1): 5160
3 Solheim O A. Design of Optimal Control System with Prescribed Eigenvalues. Int. J. Control, 1972,15(1): 143160
4 Eastman W L, Bossi J A. Design of Linear Quadratic Regulator with Assigned Eigenvalue. Int J Control, 1984,39(3):731742
5 Juang J C, Lee T T.On Optimal Pole Assignment in a Specified Region. Int. J. Control, 1984, 40(1):6579
6 ҫ,.иſϵͳ.Ϣ,1989,18(3):4145
7 ҫ.LQſϵͳмȨȷ.Զѧ,1992,18(2):313317
8 лκͣ˺.һµż÷.Ӧ,1993,10(1):113116
9 Ӣգ,ܳ.Զеľ.:ѧ磬1979
10 Kawasaki N, Shimemura E. Determining Quadratic Weighting matricces to Located Poles in a Specified Region. Automatica, 1983, 19(5):557560
11 Kailath T. Linear Systems, Prentice-Hall, London,1980
ո:1998-06-18
