Զѧ
ACTA AUTOMATICA SINICA
1997ꡡ236ڡVol.23No.61997




гȶṹ³Ӧ
Ρ󡡲
ժҪԽṹͲδ֪ķϵͳһ־ĳȶ³Ӧ. һģͣPopovȶƣ佨ģBP߱ʶʶǰмԲЧʵ³Ӧ. л˷. 
ؼʡ磬 ȶԣ³ԣ Ӧ. 
A ROBUST ADAPTIVE CONTROLLER WITH
HYPERSTABLE CONSTRUCTION
MA ZIFAN JIANCHAI TIANYOU
(Research Center of Automation, Northeastern University, Shenyang 110006)
AbstractA robust adaptive controller with hyperstable construction is presented for nonlinear systems with unknown parameters and structure. The Popov hyperstable therom is used to design the controller. The modeling error is on-line identified by a BP network. The identified results are taken as a compensatory signal so that the uncertainties of the system are effectively overcome and the robust adaptive control results are obtained. Simulation results are also given.
Key wordsNeural network, hyperstability, robustness, adaptive control.
1
ûLyapunovۺPopovȶ۵Ŀ͸һЩмֵӦñ֮У12. 㷨ûпǽģĴڼӰ죬㷨Ĺ㷺Ӧ. ԴRohrs3ӦƵ³һЩ³Ӧ46ݣЩҪһЩʵļ. ⣬㷨ȶ³ߵͬʱָһ㶼в̶ͬȵ½. Ľ˹㷨ƣʶϵͳڵĽģǰв ԴﵽϵͳȶԶϵͳĲӰϵͳ³ԵĿ. 
2
ضģ
y(t)=fs[y(t-1),,y(t-ny);u(t-1),,u(t-nu)](1)
ʽfs:Rny+nuRnΪӵķӳ.fsʱı仯Сڲʶ㷨ʣ(1)ʽɵȼģ
y(t)=ay(t-1)+bu(t-1)+N(t)(2)
N(t)ΪȼģģͼĽģ.һײӦ(2)ʽΪ
y(t)=ay(t-1)+bu(t)-bu(t)+N(t).(3)
bu(t)=b䦤u(t-1)2ݣʽ(3)ʽݻΪ
y(t)=ay(t-1)+bu(t)-b䦤u(t-1)+N(t).(4)
r(t)ΪϵͳĲο룬 ͿֱΪ
e(t)=r(t)-y(t)u(t)=Kp(t)e(t)+uf(t).(5)(6)
Сuf(t)=K1(t)r(t)+K2(t)r(t-1)+K3(t)u(t-1)+K4(t)(t).(7)
(6)ʽKp(t)Ϊ棻(7)ʽ(t)ΪʶĽģ. (4)(7)(6)ʽ(5)ʽ 
(1+bKp(t))e(t)=ae(t-1)-w(t),(8)
ʽСw(t)=-(1-bK1(t))r(t)+(a+bK2(t))r(t-1)
-(b-bK3(t))u(t-1)+(1+bK4(t))(t). (9)
PopovȶоݣʹϵͳȶӦͬʱ
(1) ǰͨ崫ݺϸʵ
(2) ͨPopovֲʽ
(10)
(8)ʽɵǰͨĴݺΪ
(11)
Kp(t)(a-1)/b, (12)
ɱ֤ǰͨ崫ݺϸʵ. ѡǰӦΪ
(13)
ʽС[1(i),2(i),3(i),4(i)]=[(i),(i-1),u(i-1),(i)];1,2,3,4Ϊʵ÷Żȷ߶̬. Lʽѡȡ
(14)
(t)=k(t)+c,k(t)ΪǰӦɵcΪ(0) ¹ϵ1
(15)
(9)ʽ(13)ʽ(11)ʽʽɵ
(16)
ʽС[1,2,3,4]=[-1/b,a/b,-b/b,1/b]; 
ɴ˿ɼ (12)ʽѡȡKp(t),(13)ʽȷӦɣԱ֤ϵͳȶ. 
BPΪģ繲㣺һڵ㣻mSigmoidԪһSigmoidԪ. ź
I=[y(t),y(t-1),;u(t-1),u(t-2),;(t),(t-1),],(17)
С.(18)
ʦźǽģy(t)-(t). 
(19)
Sigmouiãɻ
(20)
ֱΪ
(21)
(22)
Wjp,jWij,jֱȨֵֵ. ǷֱʽУ
Wjp(t+1)=Wjp(t)+jpiOp(t),(23)
Wij(t+1)=Wij(t)+ijjIi(t), (24)
j(t+1)=j(t)+jj,i(t+1)=i(t)+ii.(25)(26)
Ԫѵʽ
(27)(28)
3Ӧ㷨
ȼϵͳģ(2)ʽʾ
y(t)=x(t-1)T+N(t),(29)
x(t-1)T=[-y(t-1)u(t-1)], =[ab]T.(30)
㷨²ִ.Ϊ˱֤BPԣ֮ǰԤѵһ̶ʱͶ. 
1) ϵͳy(t)ȼϵͳʶ. 
2) ±ʶ㷨ʶģͲ
(31)
Сv(t)=y(t)-x(t-1)T(t-1). (32)
r(t-1)maxP(t-1)K, 0k,(33)
Сr(t-1)=r(t-2)[1+x(t-1)TP(t-2)x(t-1)], r(t-1)0, (34)
򡡡P(t-1)=P(t-1). (35)
С
P(t-1)0;(36)
򡡡(37)
УmaxP(t-1)P(t-1)ֵ. 
3) (t+1),ÿѵs. 
4) ݲοϵͳ. 
5) ݲֵȷKp(t)ȷL. 
6) (13)ʽɵK1(t),K2(t),K3(t)K4(t). 
7) (6)ʽu(t)ʩܿض. 
4
ʵ·ģ

УL=8,m=9,=0.025,S=2. ϵͳĸͼ1ʾ. ϵͳнǿ³ԣȶĿƽĸ. ͼ2˽ģN(t)ͱʶ(t). ͼ3ͼ4ֱǲa,bֵߺͿ. 

ͼ1οw(t)y(t)ͼ2ģN(t)ʶֵ(t)

ͼ3ϵͳߡͼ4
5
㷨⣺(1) һิӵķϵͳ(2) ϵͳδģ̬¾нǿ³ԣ(3) 㷨򵥣ҿ˷˳У㷨мʱɲ̬ϵͳʧ. 
1) йʿѧ͹Ȼѧ. 
ߵλѧԶоġ110006
ο
[1]Tomizuka M et al.Synchronization of two motion control axes under adaptive feedforward control.Trans.of the ASME,1992,114(6):196-203.
[2]骣.һ΢Ӧת̨ŷϵͳ. Զѧ198915(3)193-199.
[3]Rohrs C E et al.Robustness of adaptive control algorithms in the presence of unmodelled dynamics.In:Proc.21st IEEE CDC, 1982,3-11.
[4]Preterson B B, Nanendra K S. Bounded adaptive control.IEEE Trans. Autom. Control, 1982,AC-27(6):1161-1168.
[5]Ioannou P A, Kokotoive P V. Adaptive system with reduced models. New York:Spring-Verlag, 1983.
[6]Parlag L. Robustness of model reference adaptive control. In:Proc 3rd Yale Workshop on Application Adaptive System Theory, 1983.
ո 1995-05-19
