Algorithms for Adaptive Control of Two-Arm

Flexible Manipulators Under Uncertainty

J.M. Skowronski
University of Queensland

rc)
Lp705'i
0 9

<A

--

St. Lucia, Australia 4067

1. Abstract

'

. /

t/

---

("'"

i
~

-2
non 'near extension of model reference adaptive
control (MRAC) techKque t a guide a double arm nonlinearirdble robot
manipulator with flcxible links, driven by actuators collocated with
joints subject t o uncertain payload and inertia. The objective is t o
track a given simple linear and rigid but compatible dynmical modcl
in rcal, possibly stipulated time and within stipulated degree of
accuracy of convergence while avoiding collision of the arms. The
objective is attained by a specified signal adaptive feedback controller
and ty adaptive laws, bcth given in closed form. A casc oi 4 W F
nunipirlator illustrates th-.technique.
2. Introduction

Thc MILK technique becomes popular proposition for guidance of recenvrobot manipulators, with demand f o r
precision pointing in difficult Conditions, under the action of full scafe dynamic forces, and subject t o
uncertainty in paramcters. Such manipulators, particularly these used on spacecraft are highly nonlinear and
non1incari:able structures (geometric nonlinearity of elastic links, large angle articulation, nonlinear coupling
of DOF's. nonignorable gyro and Coriolis forces. several equilibria). while c1assi;al MRAC is linear and
applicable t o rigid bodies only. Thus the extension is needed,for handling nonlinearity.see [ l ] , and flexible
links, see 121. On the other hand many robotic objcctivcs, &ain particularly these in difficult space conditions
require at least two arm systems. Thus the tracking has to,'bc a double MRAC (mutual reference adaptive control)
which secures tracking thc same model by two arms while qvoiding mutual collision - cf. [SI, [ 4 ] . If adaptive
(self-organi-ing) control is intended. the tracking relates not t o a given path but t o a given dynamic targetinodcl with prescribed target-parameters. We take the modcl simple thus rigid and linear, but locally compatible
with the nonlinear arms regarding equilibria. Each arm is represented as an open chain with n W F , nonlinear
characteristics m d coupling, elastic links, drivcn by actuators collocated with joints, under uncertain inertia
parameters and uncertain payload. The tracking is done in real possibly stipulated time by a dcsigncd signal
adaptivc feedback controllcr and integrable adaptivc laws in thc s t a t e space, while avoiding collision between
arms of a11 the joints (and elastic nodcs) in Cartesian configdration space.
3 . Motion Equations

I.ayrange motion equations g i v e the rigid dynamics of the arms in the general format
AJ(qJ,sJ)qJ

rJ(qJ,sJ,,\j) + fiJ(qJ,AJ,sJ) = BJ(qJ,~j)uJ , j

1,2

.-

where q J ( t ) .lq Rn , t : to = 0 , is the configuration vcctor of the joint variables q: ,...,qi of the
j-th arm varying in thc known bounded work region ' of the configuration space Rn ; i ( t ) is the correspond3
ing vcctor of joint velocities in the specified bounded subset A* of the space tangent t o Rn ;
q
t S 2n ,
uJ(t)
U +. R n are the control vectors in given compact s e t of constraints IJ ; h J ( t ) E A c R L
are the vectors of adjustable system parametcrs in bounded bands of values A , and s ' ( t ) F S c Rk' is an
uncertainty psr:imctcr within the hnown hand 5
?lorcovcr AJ(qJ,sJ) :ire the L n c r t i a n x n matrices obtained in

the known way from the quadratic form of kinetic energy. The vectors IIJ = (Ilj, ....!I:)'represent pctential
forces (gravity, spring) whilc J'I = (7;.
,i'J)T represent the internal nonpotential acting forces (Coriolis,
gyro, ccntrifugal, damping structural o r viscozs, etc.) and BJ is the actuator transmission (gear) nonsingular
n m matrix. The control vectors u J ( t ) arc selected for the objectives of tracking and avoidance by odaptive
1
2
feedback control programs uJ ( t ) = pJ(ol(t) ,q2(t) ,:'(t) ,i'(t) , 4 ( t ), I ( t ) ) on corresponding products of
Aq x A * x ,\
For convcnicncc the superscripts "j" will be dropped until they are needed t o avoid ambiguity.

...

q
CansiJcring the links c l a s t i c he intrducc the Jrfonatiori coordinates f o r the i-th link as shown in Fig. I ,

while using thc Rit:-Kantorovitch

scries :xpansion

-1

ri(yi,t)
and f o r

v (y

i'

t)

, w i ( yl. , t )

v- 1

ri(ui)ry(t)

ri(yi)ri(t)
m *

analogously, with t h e exact s o l u t i o n expected f o r

-.

(2)
We take m

large

-lengthFlexible link

Figure 1.

enough so t h a t t h e Kantorovitch l i n e a r i z a t i o n is physically j u s t i f i e d . The technical way about it i s to

stepwise subdividing t h e l i n k s between g r i d a s long a s the d i t f e r e n c e of r e s u l t s f o r successive a's b e c a e s smrll.
following

[j]

Q.

Dc(q,4,0,n)

[ol(t),...,on(t))T

, where

oi(t)

(ri(t),vi(t),wi(t)IT

and

w r i t e t h e hybrid system as

, ro(o,ti) , I1ri [o,s)

A (o,s)

where

o;t)

we form t h e vector

ttaving s p e c i f i e d ( 2 )

, Pc(q,ri)

a r e the c l a s t i c correspondents of

and t h e i n t e r n a l damplng

A,

r,

I1

while

Ac(q,Q)

a s well a s t h e hybrid r e s t o r i n g c o e f f i c i e n t s
These matrices a r e formed by i n t e g r a l s over

D(*l,q,rl,r))

P(q,n) a r e matrices coupling thc e l a s t i c and j o i n t coordinates.

tha shape functions, see IS]. L e t t i n g

A(q.n,s)
t o be t h e hybrid i n e r t i a matrix which

positiv,. dcfinite,we i n e r t i a l l y decouple ( 5 ) :

i s nonsingular

[q,6)'r

D(q,G,ri,;i,\,sJ

P(q,rt,,\,s) = B(q,{.s)u

(4)

A-l(Pcr:+ll.Pn+il )T are successively v e c t o r s of nonpotcntial and p o t e n t i a l

D = A-'(Dc;l+I',D;+I*
) T and P
rl
ri
forces and the meaning o f th.? matrix 8 is obvious. The vectors q , (1, '1, r i form the s t a t 5 vector
T :. [ q ( t ) , n ( t ) , G ( t ) , ; , ( t ) ) T .\ 8
A Ax .IA :' R N , N = Jn , f o r each a m . For
x ( t ) = (xl(t)
x,(t))
q
ri
q
ri
convenience ( 4 ) may be then w r i t t e n in t h e genera! s t a t e form

where

,...,

i =

with

(fl,

...,fN)

(5)

E(X.U.i,S)

of t h e shape specified by ( 4 ) i n an obvious way.

Formally ( 5 ) may be written in t h e

contingent form:

i
which for s u i t a b l e

f ( . ) , p ( - ) , ,A[*)

through each xo = x(0) in

o f s(t)
in (5) a t each t

if(x,u,\,sJ

has s o l u t i o n s

x(tJ = k(xo,t)

i5)'

S'

t :0

, absolutely continuous c u n e i

We s h a l l consider the c l a s s of such s o l u t i m s

K(xo)

by exhausting a l l values

4. The Reference Model

We l e t the given Cartesian "world" coordinates representation of the rcference sodel i n general t e r m s
'rn =

with

2n

WF,

[(t)

c R

3.31

, and

F(Im)

F [ArnJ:

, b)

s u i t a b l e matrix, be o f f - l i n e recalculated t o t h e j o i n t c o o r d i n a t e

format of the r i g i d l i n e a r system

;im
with t h e Zn-vectors

qm,

Grn

1-1

Dm(im)yn P m ( f r n N m= '1
+

of joirit coordinates and v e l o c i t l c s , b t a t e

246

'cm(t) = ( q , ( t ) , G n ( t ) ) T

'\

- R N , and

Om, P,

Am

s u i t a b l e matrices, while

(Aml

,...,Ami)

,e

const c A c Re

Pm(Am)(qe#ne) = 0

Then

E.(%.\)
and s u b s t i t u t i n g (7),

Moreover
(8)

with (qa,rle) denoting t h e q u i l i b r i a of (3) on t h e surface i = 0

w i l l be denoted by Em(t&,fm]
i n t h e world coordinates and E m ( \ , i )
t o one another.

= n

fi

The t o t a l energy of t h e model

i n t h e j o i n t coordinates, obviously q u a l

iiT<

P,,,(Am) da

<

(9)

i,~%.i,,~
= -vp,,,i(i,,,)
2 .

(10)

The model is s e l e c t e d such as t o allow achieving o f a s t i p u l a t e d t a r g e t behavior i n t h e s t a t e space. To focus

a t t e n t i o n on something s p e c i f i c and y e t general enough, l e t it be s t a b i l i t y o f t h e o r i g i n , guaranteed by the
nonaccunnnulation of t h e t o t a l energy i.e. non-negative damping

tim($,4J

5 0

v% rl

(11)

while
> 0

VE,,,($,$J
in-the-large i.e. on same CAL = A

where

AL

(12)

is the set i n

AL

RN

enclosing a l l t h e e q u i l i b r i a .

5. Objectives
Now we consider both arms
Fig. 2 .

= 1.2

Figure 2 .

Define two product 2N-vectors

which vary i n

A*K',

The block scheme of t h e system i s shown in

and t h e model together.

Block scheme of the systcm

X J ( t ) = ( ~ ~ ( t ) , x , ( t ) ) ?I
~

4 A'

(XJ(XJO,t), ,xJ(,JO,t)),

generating t h e product t r a j e c t o r i e s

( x J ( t ) = AJ(t)
im ,
t 3 0 , X J O = XJ(0) ,

and two
.
. 2-vectors

(aJo = ~ ~ ( 0 )Then we define t h c "diagonal" s e t s

MJ = ( ( X J , x J )
and given s t i p u l a t e d

uJ

> 0

1"

S'X!

xJ =

Xm,

= 0)

, t h e l r nelghbourhoods

M; =
Moreover we l e t

t ( ~ * , i - j,

be a dezired subset of

A:*.

1xI-x

rn'

+J,

,,Ji

4'

, ) = 2

where we want t h e tracking t o occur, and l e t

247

t C be the

s t i p u l a t e d time a f t e r which t h e tracking is a t t a i n e d with accuracy

The manipulator ams ( l j a r e n r t r u i l y p-tracking the t a r g e t (7) on .A

P i n t Objoctivw

i f t h e r e Is a pair o f

of (4) i n K(xJo) , the set

c do
2 x A and given tC, f o r each

,t

1.2 such t h a t f o r each solution k'(xJo,t)

c o n t r o l l e r s p 1( e ) , J
A 2 x A is p o s i t i v e l y i n v a r i a n t : (XJO,aJO)e :A x A -(X'(t),ca'(t))
k (.) c K(xj0)

pJ

2 0

t h e product t r a j e c t o r i e s s a t i s f y

The convergence is i l l u s t r a t e d i n Fig. 3.

Figure 3.

Convergence of product t r a j e c t o r i e s

Suppose t h e transformation f r a a j o i n t t o world coordinates (forward kinematics) is g i v e n by

Ci =
and denote

4 (x1(t),x2(t))

Z(t)

l ~ ~ - CL ~d.l v o , v

( 2 ..Aa

be t h e c o l l i s i o n set between arms t o be avoided.

AA
i > 0

be t h e "slow down" s a f e t y zone. with

any p a i r

ItJ(*) e

c = 1,...,3*2n

Then we l e t the set

Second O b j e c t i v e :

::(qJ,nJ)

We define

4 [Z

<.

Liz

= I,.

C A ? Ad

I{!-<::

s u i t a b l e constant.

CA

z(zO,L)

, vt-

:"' CA ,

such t h a t f o r m y

The tracking arms (1) avoid c o l l i s i o n i f f there i s A,,

K J ( x J O ) the corresponding product t r a j e c t o r y

and

0.

(IS)

6 . S u f f i c i e n t Conditions
We return now t o the f i r s t o b j e c t i v e and specify by

of t h e region

"XA

DJ n MJ =

such t h a t

Then l e t

NL

[)(Aix,l)

N[ J ( J A x A ) ]

, CMG 2

11

(;iA;,\)

Further we consider four C -functions

a neighborhood of t h e boundary

MJ

Vi(-): 5

and introdi-:e
*

R , VJ(

open

1 : DJ

DJ

;(:aAu,l)

fi:

R ,j =

1.2

with t h e p o s i t i v e constants

The f i r s t r e l a t i o n obviously r qulres forming

forming 3A , 3.4 from l e v e l s of s u i t a b l e V,(*)
be t h e secuye choice.
THEOREH

Vi(-)

i:

Vi(.).

ObjCCtiVe I i s a t t a i n e d i f , given

VG[.)

such t h a t f o r a11

(XJ,tJ)

from s u i t a b l e , ( A x A ) taken .isits l c v c l , or m v e r s e l y ,

.I z n a l l c r t h J n their d e s i r e d w i l l
In the ldtter Ca>e a" :ho,
1,

AZ

X:

u thereare

248

prog-;ins

pJ

and fiinct ions

J J ,aJ 1 6
vS(x

v!

for each ut

V(X J .a J I c

wC , J

pJ(X7,X2) ;
<

vi(xJ(t),aJ(t))

o ,

a l o n g t h e product t r a j e c t o r i e s

o < v$J,aJ)

vJ(xJ,aJ)
P

6 vJ*
P

uj

G?,,
n
,j

MG

1
-cJ , vs

along t h e product t r a j e c t o r i e s
n i e Objective 1 holds a f t e r a stipulated

tC <

G-LL ,

t L 0

(18)

t 2 0

i f Theorem I is s a t i s f i e d with

j =

I,Z

1.2 ;

such t h a t

(XJ(XJO,t),aJ(aJO,t))

vj*
cl

= 1.2 ;

t h e r e is a c o n s t a n t

J l (t1.a 1 ( t ) )
O,,(X
Rcaark 1:

(17)

j = I,Z ;

c ~j

p j (X',X*)

v ~ J
c

(XJ(X1O,t)aj(atO,t))

v(xj,aJ)

v(x',aJ)

5 v;-,

f o r each

1.2

1,2

rralected by

IC

O b j c c t i v c 2 is a t t a i n e d i f Thoorem I holds and given

pJ[*) , for a l l 2 'CA ,

THEOREM I :

t h e r e is a C l - f u n c t i o n

VsJ6

VA(*): 1 5 ~
+

such t h a t for t h e t r a c k i n g p a i r
(vi)

[vii)

VA(r)

VA[Z)

f o r each

UJ 8

V z c JA ;

pJ[Z)

,
,

i A ( Z ( t o . t ) ):0
along product t r a j e c t o r i e s
PROOF.
Suppose some Z(:",r.)
which c o n t r a d i c t s ( v i i ! .

t 2 0

AA

Z0c

AA

Zoc

t(to,t) , t :o

crosses

3A

at

tl

.
0 .

(20)

Then by ( v i )

, V,(t(tl))

< VA(Zo)

7. C o n t r o l l e r s and Adaptive Laws

I.ct us s e t up

vA
whcrc
N,

aJ

in
To

F. - " '
1 v 'J1

[sign

IJ,

...,s i g n

IJ)

,j

= l i t m ~ x ' J - E m ~ ix ~ )

(15)

, and E,(xJ)
is E m [ * ) with x
exchanged f o r xJ
Choosing
s p e c i f i e d a d d i t i o n a l l y by ( 1 2 1 , s a t i s f i e s t i ) , ( i i i ) and ( i v ) .

= I,?

CAI.

.ice

t h a t ( V I ) holds, observe t h a t E m ( x J ) = Em[SJ , *:?J) of ( 6 ) and t h a t i n c r e a s i n g t h e d i s t a n c e

f o r a t l c a s t one J from i t s ;A value i n c r e a s e s t h e value of V

t h e c h a r a c t e r of

To Lheck upon c o n d i t i o n s

E (.)

A '

( v ) , ( v i i J we d i f f e r e n t i a t e ( 2 1 )

(11).

tm(xJ)
)

aJiJ

i,C

k,i.xm)

b,(xJ)

(23) with respect t o time

aJIJ , ( X J . i J )
+

allJ ,

, ( x J , i J ) , M;

t ) = [ E R ( x 4 ) -Em[x2)

I * [ E,[X'

where
k,(xJ)

= ; E m [ ~ J ) * f J ~, u-JJ,

[XJ,lJ)

C*M:

C-M; ,

) -bR(x2)

I ,

..

tJ)

= (6u-o-P+pmqJ ( q . )

24 9
I

me b r a c k e t r

o f t h e functions 8, 0 , P

kreovcr C'n'U are s u b s e t s of

dropped f o r c l a r i t y .

C*$:

)'&(,E

C-$:

Em(xl)

dP

d e f i n e d by

2 Cmix>

Em(x.)

With a s u i t a b l e c h o i c e o f i n i t i a l s t a t e s t h e f o l l o w i n g set of c o n d i t i o n s imples ( i i ) , (v) and ( v i i ) :

b 0

for

(c 1

pJ;J

,6 #

Ix )
m

,j

'

aJ

1.2

In t h e above

C'h

C*A:

Em(xo) L Em(x')

C-A:

Em(x4)

# 0 ,

Em(x')

CA

a r e s u b s e t s of

d e f i n e d by:

1.2

0.
Observe t h a t f o r '1 I 0 t h e r e is no need f o r a d a p t a t i o n and t h a t the system ( 4 ) c r o s s e s t h e w r f a c e
:I = 0 time i n s t a n t e n w s l y ( v e r t i c a l l y ) so t h e r e I S no need f o r controf i n v i e r o f t h e smoothness o f t r a j e c t o r i e s .
C o n d i t i o n s [ a ) , (b) are called c o n t r o l c o n d i t i o n s h e l p i n g t o d e s i g n p
, c o n d i t i o n ( c ) is c a l l c d a d a p t i v e .
h e l p i n g t o d e s i g n a d a p t i v e laws. Let 11s check t h a t ( a ) , ( b ) . ( c ) indeed imply ( i i ) , ( v ) . ( v i i ) . Consider f i r s t
( a )

E m ( x J ) # Em(xm)

the case

Substituting

(11)

i n t o [ 2 4 ) w i t h ( l @ . w e s a t i s f y ( v ) i n s t i p u l a t e d time
t m ( x J ) = Em(xm)

ll~e case
(viij.

Again f i r s t l e t

im[x2)

iR[x*j

is t r l v i a l a s then
Em(%') # L m ( x ' )

= negative teras

i n t o ( 2 5 ) i n v i e r o f ( i i ) we o b t a i n
the pwer:
i ( x ' ) S 0 t h u s (11).

(c)

Boundcdness of t h e work space necessitates

it

3Em(xm)

tr

(xJ)

S u b s t i t u t i n g ( a ) , (E) ..and

Note t h a t t h i s holds f o r any i n i t i a l s t a t e s .

F i n a l l y we check
* im(xm)
E
-E

.0 ,

it

and o b s e r v e t h a t ( b ) s u b s t i t u t e d t o ( 2 h ) i m p l i e s

(VII).

The c a s e

is obviously t r i v i a l .

Ohscrve t h a t , w i t h ( 1 1 ) ) .

IC)

is implied by t h e f u l l o w i n g a d a p t i v e l a d s

1).

i: = - s i g n ~ ~ ( 9 m- i 4 ~ i

... .

f o r L~
0 , I 7 I,
.n
P h y s i c ~ l l yt h c .elutions iJC i J o , t ) r c p r c s c n ? t h e d e l e n e r g y flux which bccme
poaitibe o r n c g a t i v c Jcpending upon where i J o is l o c a t e d Lbelow or abobe t h e w r f a c c I J = 0 ) thus r e g u l a t i n g
bJ
A 0
K O zero iron Jnywlicre u u t % i J r t h e *urfdcr
t h e increment or

8.

Modular Oouble RP-Manipulator

Our t e c h n l q u c
I1g. 1.

IZ

i l l u \ t r ~ t c r lbelow on t h e case sturly of t h c f a t r Ml' m.inipulJtor w i t h two

3houn
J
in

thre As, X4 a r c d u p i n g c o e f f i c i e n t s , A1,AA2

spring c o e f f i c i e n t s , g-gravity a c c e l e r a t i o n , tlle r m i n l - . r
1.2 , a r e ignored f o r t h e time being. Ne takr the
of notations s h a m i n Fig. 4. The s u p e r s c r i p t s "J", j
possible payload on t h e g r i p p e r s as u n k n m but within known bounds which makes m2 s p e c i f i e d b:r

-. s m 2 5 i ,
1
Allowing s i n q1 * q, - i;si , cos q1 = 1 - I q f

where E,
p o s i t i v e constants.
(29) by corresponding i n e r t i a c o e f f i c i e n t s we obtain:

Choose

(XJO,

l J 0,)

Cod,

1.2

.and

4,

ri + ni

Z0

C*A

iui

,i

, and

subdividing t h a q u a t i d n s

1.2

Then the c o n t r o l conditions ( a ) , (b) h o l d i f successively

and

25 1

Thus ne choose

and for such

u;

u;

such that for

, re

choose u f

4;

f 0

satisfying

The procedure for' ui and ui is identical utilizing the second inequalities of (M), (55). Assuming
= nf = mI , m i m i = m2 C [!,a]
, r4 = ra r and substituting the exprcs3ions for
symmetry of ams:
j
=
1.2
,
YE
obtain
the
tracking
controllers
i.J
U J , S! , i ,

mi

suitable const:int,

4' :

suit:iblc constant.

Vq:

= I) ,

which i m p l y the control conditions ( a ) , (b) for our example.

1;

;:
for

= 1.2

simulation of
tn3

-(sign
= -[sign

,AJ)\

tJ)i

n341
nl
d"m2

'The adaptive laus 1 2 4 ) .ire

:c

I .

lc]

'.J

identically, since by Jcsign

1 = 'a1 '
1
nodular case. uith the data m I = ;Okg, 5 = j O h p , ii = JOhh. r = 0.hbm.

The first t w o l:iws vanish

OUT

*z

(1

= 5, a4 = 2 , is 5hown

in

I i y . 5, .ind c o n f i n s the convcrpcnce- voidance required.

252

. '

Zlumerical
d
i m l = -0, l r n l
20.

9. References
(1) J.M. Raronski, Control Dynamics of Robot Manipulators, Academic .-rP

[Z] J.M.

Skowronski, Model Reference Adaptive Control Under Uncertainty of Nonlinear Flexible Manipulators,
A I M Paper #'E6
1976 CP, Proc. A I M Guidance, Navigation and Control Conf., Williamsburg,V), 1986.
[3) G. Leitunn, J.M. Skowronski, On Avoidance Control, Journal of Optimization Theory and Applications, Dec.
1977.
141 G. Leitmnn, J.W. Skowronski, A Note on Avoidance Control, Optimal Control Applications and Me;hods,
Dee. 1983.
[S] A. Truckenbrodt, Effects of Elasticity on the Performance o f Industrial Robots, Proc. 2nd IASF-D Danos
International Symposium on Robotics, Switzerland, 1982, pp. 52-56.

25 3