Զѧ
ACTA AUTOMATICA SINICA
1997ꡡ234ڡVol.23No.41997



ڵѧϰĻеֲռ
/λûϿ㷨1
Τ졡ɭ
ժҪڶԳеֲռ/λûϿ㷨ļ򵥻عˣԸ㷨ѵķһֻڵѧϰĻеֲռ/λûϿ㷨ƻеͬ߸նȻӴʱе/λûϿƵĶ̬.ѧϰ㷨֤.ʵ㷨пٵԣܴﵽܸߵ/λö̬ƾ.
ؼʡѧϰƣ/λûϿƣеֲռ.
HYBRID FORCE/POSITION CONTROL OF ROBOT
MANIPULATORS BASED ON ITERATIVE LEARNING
WEI QINGCHANG WENSEN ZHANG PENG
(Department of Automatic Control, National University of Defense Technology, Changsha 410073)
AbstractBy analyzing the difficulties of the conventional force/position hybrid control algorithm in applications, this paper presents a new operational space control algorithm of hybrid force/position control based on iterative learning to improve the dynamic force control performance of robot manipulators of contacting with high stiffness environment. The proof of the algorithm convergence is given. Experiments show the validity of the new algorithm. It converges rapidly and can acquire high-precision force/position dynamic control performance of robot manipulators.
Key words Iterative learning control, hybrid force/position control, operational space control.
1
֪˾߱ƹܣԱֳһܻܹʤθΪӵĲ.Ƽо/λûϿ㷨ۿܴ彨ȻƼȴûдﵽʵûĳ̶.ͬһҪûеõкܴĹϵ߸նȻеͬ߸նȻӴʱƵĶ̬ƾȵ.ǲûڵѧϰĻеֲռ/λûϿ㷨.
ѧϰ1984Arimoto1ݵģǶʱϵͳͨõѧϰ㷨. Bondi21988ڻ˿ƣչһֵѧϰ㷨ϵͳλáٶȡٶźĽÿεĿ룬ͨ淴켣ĳһȷ֮ڣȷ类֤ѧϰ㷨.Kuc31992һϵͳһֵѧϰƲԣǵ㷨ûʹø΢Ľǰѧϰͨջϵͳĸ淴֤õ㷨.
2еֲռ/λûϿ㷨
Ļеֲռ/λûϿ㷨Ŀɣ(1)ǰǲеԼӰ.(2)ͨһԷʹеڲռϵĿƽڲռƶλÿ.ռеֵĶѧΪ
M(x)+H(x,)+G(x)=Fc+Fe
(1)
ʽM(x)ΪеֹH(x,)ΪеָG(x)ΪFcΪеֲռFeΪԻеĩִ.ѲռFcΪǰFaͷԷFb
Fc=Fa+Fb
(2)
Fa=(x,)+(x)
(3)

(4)
ʽ(x,)H(x,)Ĺƣ(x)ΪG(x)Ĺƣ(x)ΪM(x)ĹƣSΪеֲռλú̬ѡΪռѡ.λÿձʹõPD㷨(1)(2)(3)(4)ɻеֲռ/λûϿ㷨.
еڹҵеĴӴҵͬ߸նȻӴĳֽļӹҵȻͬ߸նȻӴʱеƵĶ̬ܲ⣬ʹ˻ƼеĹ㷺ӦãҪ¼ԭɵģ(1)еֲռ/λûϿ㷨͹ϴڶ̬ϣҲǻеּλÿʱFe0ʱռĿʵǾ̬ģ䶯̬ƴά⣬M(x)ķǶԽɵ.еֲռ/λûϵĶ̬άӰе/λûϿƶ̬ܵһҪ.(2)еֲռ/λûϿ㷨ĸ.е/λûϿ㷨ĸ.е/λûϿ㷨Ǹӵ㷨ȫʵָ־ϸĲͽҪļ㣬ϵͳڹ.(3)еδģѧ.ף4жԻеδģѧеĶѧԡеѧԡѧԡϵͳӳٵȶԻеȶԵӰ˷õеδģͬеƶ̬ȶأʱļٶϵͳȶԵĸкܴãڻնȺܸʱPD㷨ѱϵͳȶԣҲǸ߸նȻеͬ߸նȻӴʱܵߣͨﵽֻɾ.
3ڵѧϰĻеֲռ/λûϿ㷨
Ӧõѧϰƣеͬ߸նȻӴʱƵĶ̬ܵ.ǵĿƲǲú㶨PD+ǰѧϰƣPDΪ˱֤е/λûϿƵȶԣеֿܵͨǰѧϰõ.деֲռĶѧΪ
M(x)+H(x,)+G(x)=Ub+Uf+Fe
(5)
UbԺ㶨PDUfǰͨѧϰĽеƵĶ̬ƾ.ǵkε

(6)
еֵ
ek(t)=xd(t)-xk(t)
(7)
Ժ㶨淴ƿΪ

(8)
еֿƵǰĵѧϰΪ

(9)
ʽ(6)(7)(8)(9)ɻڵѧϰĻеֲռ/λûϿ㷨.õѧϰ㷨ԣ½ֻ֤еڲռȫλÿƵԣе/λûϿƵ֤ǻеλõĸ淴֤Ҳͬ.
1ڻеֵĿƷ(5)M(x),H(x,)G(x)Lipschitz
M1(x)-M-1(y)mx-y, x,yS
H(x,)-H(y,)h1x-y+h2-x,yS
(10)
G(x)-G(y)g.x-y, x,yS.
SΪеֲռĿܵ״̬m,h1,h2,gΪ.㶨PDΪUbKDc+KP.eڲͬǰ룬ϵͳ״̬

(11)
c,dΪ.
1.еֵѧϰ㷨(6)(7)(8)(9)ĳΪ
I-M1(xk)ݡܦѣ1.
(12)
֤.ǵkk+1οƣ

ʽֱǵek=xd-xk, k=d-k, k=d-k

(9)ʽʽ


ȡǵ1

ʽСr1=KP(1+M-1(x)m)+M-1(x)m(h1+g)+mm(Hm+Gm+KDk+1m+KPek+1m+m), r2=KD(1+M-1(x)m)+M-1(x)mh2,R=r1*c+r2*d, ж塬f(x)m=(f(x)), x(t)Rn[0,)Ƕ˷Ϊx(t)=e-t.x(t)ʽ߳e-t ȡ˷



ֻҪIM-1(xk)ݡܦѣ1ܿҵִˣʹۡIM-1(xk)ݡ1ѧϰ㡬kKDkKPekˡ0ԺѡƲKDKPСek0ԡxd(t)-xk(t)0ѧϰ֤.
˵.ڶ1֤Уǲûϵͳĸ淴֤ϵͳѧϰƵԣûҪϵͳȶһʣûвø淴֤ϵͳѧϰԣڻеϵͳ˵ǸʣǲýϵPDֻҪ֤ϵͳ޶ȵȶԼɣϵͳĿɵѧϰ𽥸ƣزøӶϵͳƵĶ̬ȶ.㷨ҪϵͳٶȱĽϵͳǰ룬һϵͳмٶȱ׵õҪĳֹƷ᲻ɱźŵʱӳ٣õļٶźǲӦϵͳʵʱƵ.㷨ŵƹǰѧϰʵԷḶ́Ҳǰѧϰ߽̿УٶźҲ߽дԶԼٶźŽ˲ҿԽ˲ʱӳ٣Ӷõϵͳٶź.ȡʽѧϰͬPDȣ㷨CPUһЩ洢ʱ䣬˸õ㷨кܺõʵʱ.
4ʵ
еĻڵѧϰĻеֲռ/λûϿʵͨϵͳGKD3ɵģGKD3еϵͳӲṹDELTA-3000ҵƻ̨PUMA-562е̨֣ά.VME߲ʽṹDELTA-3000ҵƻΪϵͳƽṹ壬ʵεĿƣϲ֯Э͵ײŷƼ.GKD3ɫʵ˻еֲռά()/λ(̬)Ļŷ.ÿνеѧϰǰó/λûϿ㷨ƻе˶ѧϰƵĳʼλụ̀Ȼʼõѧϰ̣ÿοƹУǽѧϰҪPD㶨еֲռϵٶȣеֲռĹ(x)ڴУڿƹвÿοƹ̽ȶԻеֲռٶȽ˲ʱͺ󣬶вֵõеֲռϵļٶźţȻٰѧϰ(9)һοƵǰ.Ϊǽеĵѧϰʵ飬(x)ȡΪԽǾϵͳΪ2.4ms.
1)еֲռ䴿λÿʵ
еxλƶΪ0.10myλƶΪ-0.10mzλƶ-0.30m.е˶ʱΪ5s.еֲռλÿƲ
P=500500500500500500ݣ
D=202020202020.
ʵͼ1ͼеһŴѧϰĴ.


ͼ1(a)xλ


ͼ1(b)yλ


ͼ1(c)zλÿ
2)ƽʵ
еxyΪλÿƣxƶ0.20myƶ0.0mzΪƣzָΪ-10.0N.еƶʱΪ10s.еֲռ/λƲΪ
P=500,500,1,500,500,500ݣ
D=20,20,200,20,20,20.
ʵͼ2ͼеһŴѧϰĴ.


ͼ2(a)xλÿ


ͼ2(b)yλÿ


ͼ2(c)z
Ϊ˱ȽϸõѧϰɵĿЧѡȡʵ8εĿЧ볣еֲռ/λϿ㷨ĿЧбȽϣͼ3.㷨PDƲΪ
P=5000,5000,1,5000,5000,5000;
D=2020200202020.


ͼ3(a)xλÿ


ͼ3(b)yλÿ


ͼ3(c)z
ϻеֲռ/λûϿʵ˸㷨пٵԣܴﵽϸߵ/λö̬;̬ƾȣͬĻеֲռ/λûϿ㷨ȣõѧϰеռ/λûϿ㷨Խ.
1)һܻרоĿ.
߼:Τ졡У1969.19917ڸѧӹϵѧѧʿѧλ.19935ڹƼѧԶϵʿѧλоΪеܿƣڹʿѧλڼƻĿһԤĿоĿǰҪоΪܿƺͻе.
ߵλ:ƼѧԶϵɳ410073
ο
1ݡArimoto S, Kawamura, Miyazaki F. Bettering operation of robots by learning. Journal of Robotics System, 1984,1(2):123140.
2ݡBondi P,Casalino G, Gambardella L.On the iter ative learning control theory for robotic manipulators. IEEE Journal on Robotics and Automation, 1988, 4(1):1422.
3ݡKuc T Y, Lee J S, Nam K. An iterative learning control theory for a class of nonlinear dynamic systems. Automatica, 1992, 28(6):12151221.
4ݡΤ죬ɭ.еƶ̬ȶԷ.ˡ199618(3)173178.
ոڡ1995-05-29
