RECURSIVE

NEWTON-EULER

FORMULATION DYNAMICS
M. G. Nasser

OF

MANIPULATOR

Lockheed

Engineering & Sciences P. O. Box 58561 Houston, Texas 77258

Company

1.

INTRODUCTION

This paper presents a new recursive Newton-Euler procedure for the formulation and solution of manipulator dynamical equations. The procedure incluaes rotational and translational joints and a topological tree. This model was verified analytically using a planar two-link manipulator. Also, the model was tested numerically against the Walker-Orin (ref. 1) model using the Shuttle Remote Manipulator System data. The hinge accelerations obtained from both models were identical. The computational requirements of the model vary linearly with the number of joints. The computational efficiency of this method exceeds that of Walker-Orin methods. This procedure six-by-six formulation the model. may be viewed as a considerable generalization of Armstrong's is adopted which enhances beth the computational efficiency method (ref. 2). A and simplicity of

In section 2.1, we begin with assuming an open chain, rotational joints, and prescribed base motion. In section 2.2, the procedure is extended to translational joints. Section 2.3 extends the formulation to a topological tree. Section 3 includes the algorithm summary and computational efficiency. The appendix contains descriptions of coordinate frames and notations and a summary of the standard kinematlc relationsused in the algorithm. 2. DYNAMICS FORMULATION

Let'sbegin with a quick look at the procedure. The first step isto set up the equations of motion for a generic link i (rotational) in the i - I frame in a 6 X 6 formi.namely, SzU 4 -- Fi*. U_ is a 6 x I

vector consistingof the reaction loads from link i - I on link i and 0 i, the hinge acceleration of link i.S i is a coefficient matrix, and Fi* consistsof the mass and inertiaof link i (inertial parameters) acting on the inertialmotion of the i - I frame, nonlinear terms, body forces and torques, control torques, and reaction loads between link iand link i + 1. The procedure consistsessentiallyof two phases, the inbound and the outbound. In the inbound phase, one begins at the free end, i = N. Since there isno outbound link,the reaction loads from link N on link N + I are zero. Therefore, F_ is given by F_' N = A N N- I qN- I N- 1 BN, N- I where AN, N- 1 involves only link N inertial paramet_'rs. ' ' Now and URN_I,N [equation Now (2.1.7.1)] we are may ready be solved for in terms to link first. eN-l; of SN-1,AN, I and N_I,qN_I,N_I, URN_I,N results in .

BN, N_ 1 but URN_I,N

not 0"N' must N)N_2

to proceed

N -

substitute

However,

be transformed into three terms:

to the N - 2 frame the first involving

This

transformation qN-2,

decomposing third,

(URN_I,

the second,

N -2' and the terms

a collection

of nonlinear

and forcing

terms.

This decomposition equation I

enables is

one to group these

with their counterparts

from link N - I. The resulting LN_ 1 UN_ 1 "- FN_

of motion

309

Notethat in thisequation
Repeating equation the procedure containing the hinge pass,

of motion for

for link uRi_I,

N-

I, 0"N does not i = N 2, N

appear, 3 .....

only

ON-I'

ON-2' obtain

etc. the

by solving

i for

we finally

acceleration beginning at

of the base the

link only. link 2, we compute all hinge 0"2, _2)2 , and accelerations. (_2)2, then

For the outbound proceed to link

base link,

3 to compute with

03, (3)3 , and (_3)3 , and so on to obtain description of the model. JOINTS

Now we proceed 2.1 2.1.1

a detailed WITH

MANIPULATOR INBOUND PASS

ROTATIONAL

The translationalequation of motion for the center of mass of link i in the i - 1 frame is (see figure2-I and the appendix)

(@,)

(2.1.1)

. - " " " LINK N

____

END LOADS EFFECTOR

LINK

LINK

_i

(LINK 2) BASE LINK

OPEN

KINEMATIC

CHAIN

LINK

i FREE

BODY

DIAGRAM

Definitions:

(Fi tNi (f_c)i

l)i I)i

- 1 - 1 - l 1

= = = = = 1 : : = = =

the the the the the the the the the

inertia inertia control control reaction reaction reaction reaction position

forces torques forces forces force

developed developed applied applied exerted

in link in link at the at the on link on link on frame

in the _ in the

t i -

1 frame I frame i in the in the i i I l frame

proximal distal joint

joint

of link i

{f=i+lCli_ {[i+l,ili_l (ni+l,j)i_ (fi-l, (ni_l,i)i_l (Pl*;_ _ 1 _ )i

of link

frame i i I I frame I frame

i by link link

i + 1 expressed i + I i I -

in the

moment force

exerted exerted on

i by Link i

expressed

in the i

i by link link

expressed ! i i

in the

frame

moment vector

exerted of the i

i by link

expressed ! frame inthe i

in the and

i = J frame in the i I frame

relative

to the

expressed I frame

{nl + IC)i - I In_ici= - I 1 l

thecontroltorquesappliedatthedistaljo_ntoflink the the the control external external torques forces torques applied applied applied at the at the at the proximal center center joint of

link link of link

i in the i in the ( in the

t i i -

! I

frame frame 1 frame

(ff, t E_ i (NiEii_

= =

of mass of of mass

Figure

2-1.

ORIGINAL OF 310 POOR

PAGE

_S

QUALITY

E F i is the momentum

total force exerted on the center of mass of link of the center of mass of link i in the i - 1 frame. _, Fi =fi+l,i +f_-l,i "/" Ff (2.1.1) -fi+l yields

i in the

i -

I frame.

Pi is the

linear

+f; the following translational

(2.1.2) equation of

motion

Substituting for any link

equation (2.1.2) into equation i in the i - I frame:

(2.1.3)

a.=m _

i -1

+ _i -1 Xr

:)
i I frame (torque balance

(2.13.1)

(2.1.3.2) The proximal rotational joint of link equations i) are of motion for link i in the about the

( _
or

N i i-I

=--dt

p. [i_i i-I r_ X _

v_.i_ , x p i i-1

(2.1.4)

i i-I

_i

r"

X m 5g + _l

(,,)

_ i_l",l

+ _i X

(,,)

i_l_

co

(2.1.5)

(li)i-1

= R,_I,

il_ R,, ,-I

= Ji

(2.1.5.1) (2.1.5.2)

+ N! + P*. xf

N t=N

f + r* X FE

c - P*. xf_+!

c l + ., nc -ni+

(2.1.5.3)

The rotationalequation of motion forarbitrary link i is: ni_l,i z_z* X ( r i + m.r. * X _i-I 0:) -Ji --_i (z i-1 Oi)--a* + _*i +-'r"* X,6 i_ - N _ i t + -i r* x -,a l _ _ +I,i

(2.1.6)

+,. . _t,_+
a.

"m = Ji _i -1

=Ji
Equations (2.1.3) and (2.1.6)

_i-lX_i-lOi

+_.J._o._,

i-.,
in the following matrix form:

may be combined

and written

S U=F:
= UR

There is no reaction torque in the drive direction. S i=I-ZizT-ai,i_lz,ZT, where ! is a 6 X 6 identity hinge i. matrix, , -z.zTja, , and , inertia associated (2.1.7.2) with

Z i is its last column,

Ji a is the actuator

i =Ai,

i-lqi-l,i-I

+Bi,i-I

4"

i +1

4"

p_X_i,i+l

31t

(2.1.7.4)

=
Bi, i-I A = $i $i+l * * I B _fc+fc ..,Ci _i t _FE 1 (2.1.7.6) [I] is a 3 x 3 identity matrix, and/_*] is a skew symmetric matrix associated with r *. link N (the link at the

Since the formal structure of equation (2.1.7) has been defined, free end) and make use of the followingboundary conditions:

consider

o
_N,N+IJ LN Therefore, equation (2.1.7) applied A__ SN to link N is qN-I Vi=l,2 + BN, N_ ! ..... N FN = AN, N-1 G
|

(2.1.8)

(2.1.9)

(2.1.10) (2.1.11) (2.1.12) 0N cannot be computed (2.1.7) for i --- N - I. relation is used: until

ALT_
I

U N = GNF N Although and _N-l When the are. expression Therefore, (uRi_I, for 0_/was obtained proceed to link N in equation (2.1.12), I and set up equation the following

_N-1

transforming

i) into the i - 2 frame,

recursive

()

qi, i i-I

=pT

qi_l,i_l+oi,

i_l+oi,

")
i_!

Yi=l,2

.....

(2.1.13)

i,i_I = O'i Z.,

(2.1.13.1)

(2,1.13.2)

(2.1.13.3) I is a 3 X 3 identity matrix, and [.P_*] is a skew symmetric matrix associated withPi".

N-2

312

UN
at

I,N 4N-2
at

FN_ I = AN_I,N_2qN_2,N_

2 4- BN_I,N_

2 4-

PN_I _I,N
N-2

at

UN-I'N 0

= [RN_2,N_IYGNAN, N-2

N_IRN_2,

I,N-I

4- RN-2,

N-IYGNBN

N-I

* AN, N_ 2 = RN_2,N_I * AN, N-2 BN, N-2

atT YGNAN,N_IRN_2,N_ T AN, N-2PN-I

(2.1.14) I (2.1.15) (2.1.16)

= PN-I
at

= RN-2,N-I N_2

YGNBN,
at

N-I
at

(2.1.17) 2

BN, N_ 2 ----PN_IBN, The superscript T denotes the transpose

4- AN, N_2ON_I,N_

operator. (2.1.18)

Ri_l, Obviously, upon substituting

i :

Ri_l,i into equation (2.1.18) for i -- N I, we get

for U R N-I.N qN-2,N-2

FN_ I =

AN_I,N-2

4- AN, N-2

4" BN_I,N_
o.

2 4- BN, N_ 2 4- A it can be moved

,N_2ON_I,N_2 to the left-hand side to

combine

Since A'N, N-2 N-I, with its counterpart Thus, in general,

N-2 is a function from link N - I. of motion

of ON_ I only,

the equation

for any

link i takes

the following

form: (2.1.19)

L.U.=F.

L i = S_ - Ai+1,i_iZ. F i = A at i,i-lqi-l,i-I A*i,i-I A_+l,i_l=ViR__l, = A i,i-I

"

zT
| |

(2.1.19.1) (2.1.19.2) (2.1.19.3) (2.1.19.4) i (2.1.19.5) (2.1.19.6) i,i-I (2.1.19.7)

+ B_ ,i-i + A_ +l,i-I atT pr

* iYi+IGi+IAi+l,iRi-l,_

B: :B, ,i-I _,i-I

+ A_l,,_loat
at

iYi+lGi+lBi+l,i

a prescribed for 0 2 from

compute

F2

the following

313

O;-Once computational namely, proceed 2.2 02 is obtained, cycle for the (_2)2 base and

zT G2Fz can be computed. on to link This 3 and completes repeat the

(2.1.20) outbound sequence Then we

(_2)2 Next

link.

we can move until

the same

compute F3, 0"3, (5-3) 3, (_3)3, to the integration phase. WITH

0"4, etc.,

all hinge

accelerations

are determined.

MANIPULATOR Some manipulators

TRANSLATIONAL a mixture for rotational

JOINTS and rotational joints. The procedure of the

contain section

of translational joints

developed

in the previous

is still applicable

with slight

modifications

expressions involved (using ai.i-l' i, i-I * ,_i, and fl, rotational counterparts,

the kinematics for translational link). These expressions If we denote these variables by a prime to distinguish

include Ui, Z e them from their

we get

(2.2.2)

U',=

[f,_,.,(1)

.f'-I,(2),

0",

_a,_,,,(1)

_ai_,.,(2)

-n,_1,(3)].

(2.2.3)

B'i i-! '

, +f.i

, Xfl'i -Nt

(2.2.4)

i,

i_

-_

'* = a'i-I

_i-IX(_z-lXPi _) g'.=[O
1

+2ffi-lXZi-lOi (2.2.6)

1 0

O] r joints case. in the following form:

(2.2.7)

The remaining Therefore,

variables

are defined of motion

as in the rotational

the equations

for any link i may be written L.U.=F.

where between

the

formulas

obtained

in the

rotational joints

link

case

still

hold.

Note

that

the

only

distinction Zi

rotational

and translational

is through *'i , i - /,andZ'

the use of either

fl, ft.*, U., o l,I--/ .... __I _I I links.

o * l,l ._ 1, and

for rotational 2.3

linksorfl.',fl.*', _l _ TREE

U' _ ' o'. I,i-l'

i for translational

TOPOLOGICAL The

manner.

case of a manipulator with tree topology does In fact, only the root links must be treateddifferently.

not alter

the

formulation

in a fundamental

case

Consider the system shown until the root link is reached.

in figure 2-2. For any branch b., we can proceed Denote the root link by K; hence,

as in the open

chain

314

|_tmcH

/,

ti

Figure 2-2.

(2.3.1)

Recall that in the open chain case (Ui,i+#R)i was transformed to the i - I frame and expanded in terms of A*i +I, i-I'qi -I,i-I' and B* I,i-1" Therefore, we get (2.3.2)

RI

K + I

Rm

or

"= j=l
_

AK+I,K_

(2.3.3)

SJ

BK,

K_

I =

BK,

K_

! 4"

_ j=l

BK+I,K_I

(2.3.4)

For any j, the definition

ofA*JKI,K_ AND

1 and B

*j

K+I,K-!

is the same EFFICIENCY

as that

of the open chain.

ALGORITHM OPEN

SUMMARY CHAIN

COMPUTATIONAL

3.1

KINEMATIC

Start at the free end, i = N. 3.1.1 INBOUND PASS - I,

Repeat the followingsequence for i = N, N I. Compute 2. A*

and B _il,i-I
i,i-l"

(may be skipped for link N).

ComputeA*

andB*

315

3. Compute 4.

L tand G_.

i -- i - I and repeat untili = 2. PASS

3.1.2 OUTBOUND
I. Compute 2. Compute

Prescribed base motion: _I, _I, _I, and v! are given. Repeat the followingsteps for i = 2, 3 ....N. eitherF orF 0_. (_i) and (-_i) '(i= I).

3. Compute 4. 3.2 3.2.1

i = i + I and repeat steps I through 3. TREE PASS

TOPOLOGICAL INBOUND

Apply the open kinematic chain procedure to all branches until the base node is reached in this case. 1. CmputeA*JK+l,g_ I andB*JKt,K_ j = I, 2 ..... m where m is the number ComputeA*K,K_ ! and B*K,K_ 1. l orA*Y'K+I,K_ I andB_'J'Kl,K_l of branches at the base node. forall

2.

3. Repeat steps 2,3,and 4 as in the open chain unless another isreached; in such case, repeat steps I and 2. 3.2.2 OUTBOUND No change. 3.3 COMPUTATIONAL EFFICIENCY is equal to 258N 119, and the number ofadds is equal to 19IN - 83, PASS

where 4.

The number of multiplies N is the number of links.

CONCLUSIONS

A general procedure for the formulation and solution of the equations of motion for a rigid manipulator has been presented. This procedure includes a solution for the tree topology. The extension to a closed kinematic chain follows naturally. However, the presentation of this extension is pending formal implementation and verification. 5. ACKNOWLEDGMENTS

! wish to thank Mr. Ken Hopping for the model implementation and for his suggestions and recommendations. [ wish to express my deep appreciation to Mr. Carl Adams forhis support throushout the development and verification process - m particular,in taking on the tedious task of verifying the two-linkplanar manipulator case analytically. APPENDIX LINK COORDINATE

FRAME

AND

NOTATION

We adopt a dynamic reference frame. This frame is used here with the Denavit and Hartenberg convention (ref. 3). The joints are points of articulation between links and are numbered such that jointiconnects link i - I and link i. Consequently, jointsiand i + I are the proximal and distaljoints, respectively, of link i. Each link i is assigned a Cartesian coordinate frame, (xi, y_,zi),which is fixed on the link and therefore moves with it.(See figureA-1.)

316

Zi_ !

Zi

Xi =

1_ i_"_'

#/ = JOINT i + !

LINK

i-I_

J."_IIMI(

/i/

_.__"_""-..._

y, t .i-----JOINT i JOINT

_"",,,,_---T'-. )

/-'i_"
J_

Yl

PROXIMAL

Figure The z i axis the common is the axis normal of the rotation/translation fromz=_t tzi' The,yi vector

A-I joint of link i. The _Lx axis the right-handed parenthesis is directed system. notation is along

of the distal axis equalszi

X x i to complete frame,

In order to associate a particular introduced as follows.

with

the

coordinate

an indexed

(Oi ) i-1
(P*i)i_i

= the link i relative = the position vector

displacement of the i frame

with

respect

to and expressed

in the i in the i -

1 frame

relative

to and expressed

I frame is A-2.)

To relate two neighboring coordinate frames, a transformation from defined as successive rotations of 0 i about the.z i -1 axis followed"b'_ This is denoted as Ri, i-1

the i - I frame to the i frame i about the x i axis. (See figure

=.o, ..(+)Rot,(o)
l --! COS 0 l

sin O= cos _ i cos 0 i - sin _ i cos Oi

- cos _ _ sin 0 i sin i sin 0 i R -l i,i-1 =RT l,i-I

sin _ _ cos dpi

(A.l.l)

(A.I.2) = Ri-l,i al t (A.1.3)

a. sin O.
| t

.cosO. J
si

_i

LINK

2LINK i1J

.,
"

,
_-1

Note:

When

the z i_l

and z i axes are aligned, Figure A-2

it implies

that

Oi = 0.

317

The following is a set of standard kinematic relative to a moving reference frame.

relations

(see figure

A-3) for the motion

of a rigid

body

if link

i is rotational (A.2.1) i is translational

(e'),-1

and

(_s)

,-I

zi-l_

if link

zi x t

zi-I

Figure A-3 T

(o,),_, =[oo o,l,_ '

(A.2.2)

(_,),_,--(_s),_1--I _,I,_,
T

(A.2.3)

(_),_,: (_),_,: [o o _,],_,

(A.24)

(A.2.5)

2.

Solution to the Eauations of Motion of an n-Link Manipulator. Proc. of Machines and _dechanisms, Montreal, July 1979, pp. 1343-1346. Notation for Lower-Pair June 1955, pp. 215-221. Mechanisms Based on

3.

J.; and Hartenberg, R.S.: A Kinematic ASME Journal of Applied Mechanics,

31_ _