.

b
NASA T T F-9294,
APPL I CATI ON OF THE METHODS OF ANALYTI CAL DYNAMI CS
TO THE THEDRY OF OPTI MAL F'LIGHTS
V.S.Novoselov
N65- 19510
ITIIRUI
8
..
-
/
i P TO
(CATIQORYI INASA CR OR TMX OR AD NUMBER)
~ - ~~
Translation of nPrimeneniye metodov anal i ti cheskoy
dinamiki v teori i optimalfnykh poletov".
Vestn. Len. Univ., sere mat. mekhan. i astron.,
NO -3 9 pp -133-146, 1964.
c
NATI ONAL AERONAUTI CS AND SPACE ADMI NI STRATI ON
WASHINGTON MARCH 1965
i
NASA TT F-9294
. -
3%
APPLICATION OF THE MJ 3l"ODS OF ANALYTICAL DYNAMICS
TO THE THEDRY OF OPTIMAL FUGHTS
V . S .Novo s elov
\ qsi o
i
I
I
The author gives an exposition of the theory of optimal pro-
:v !
' I I
' ,
cesses, based on the methods of anal yti cal dynamics.
tegrati ng the variational. foxmula of the acti on functi onal
over the traj ectori es or arcs of a speci al f i el d of extremals,
the maxbum pri nci pl e i s proved, and the suffi ci ent conditions
of the extremumare established. The possi bl e appl i cati on of
the Hamilton-J acobi method to the construction of optimal re-
l3y in-
_ _
' i I gimes i s pointed out; starti ng f r om that method, equations
- ' 1, i
-.. 1 bi ts are obtained.
2 5
defining the optimal impulse transfers between mpl anar or-
I
- ,
,Section 1. The Conditions of Stati onari ty
y .
Given: a control l abl e device whose motion i s described by the system of
.?
>'
, ordinary di fferenti al equations :
(1.1
+( I
;where the 4 (t) are the control s or the control functions, which may be piece-
/ : !
2 !wise continuous; and t i s the time.
I
,'
The controls must satisfy the conditions:
' I - 1
' 4
' ,
u&1 \<uk < U&r, U&1 =Consf, st.
(1.2
I n parti cul ar, f or all or some equations, the equalities 4 1 =- , - 2 = c o m a ;
I
be sati sfi ed.
* Numbers i n the margin i ndi cate pagination i n the-ori gi nal f orei m te&.
- 1
- _
We shal l assume that to any admissible equation there corresponds a uniqu
;
arc system (l-l), starti ng from the poi nt wi th the coordinates
at the ini-
, ~ tial time of motion to . The end-time w i l l be denoted by T, and the coordinate
-
.: 4
. of the end point by 2 .
\
The boundary values of the variables, i n the general case, are correlated
10 ,
by the i mpl i ci t conditions
NASA TT F-9294
4
(1.3
:
- ti ons (1.1) - (1.3).
We shal l cal l a traj ectory allowable or admissible i f it sati sf i es the rela-
/
I
J
Required: to sel ect the controls i n such a manner that the functional
T
(1.4
- - +cal cul ated f or these controls shall be minimal as compared wi th the values
2 , I
- 4 - < !of the functional f or a l l other admissible arcs.
/1
'
~
- . 1-
The functional J on the admissible arcs i s equivalent to the functional
1 - \
%I
. 'where
1
(1-4'
. : )
-0 ~
.- terminate constants.
\ I
d% vanish on the boundaries of the re
* I ]formulas y =I& (Yk ), whose derivatives
Here, A i are the expl i ci t functions of time to be determined and Vp are inde-
From the controls I& we may pass t o the new control s yk by means of the
' I 1
' K
\ - > l
igion (L 2)*
;several papers (Bib1.1.2).
I I nstead of expressing I& in terms of yk , a l arger number of equations can
'be considered and auxi l i ary indetexmhate mul ti pl i ers hn+k , which are function:
I
'of time, can be introduced.
Examples of the introduction of such control s are discussed i n
:' 1
' i
_ _ i
,
To the function L, i n that case, we must add the
7 - 1 -
L
3 sum
- 4
7
r
The total vari ati on of the action functi onal w i l l be defined by the for-
'>
mul a
L .
'
f or which the following notati on has been adopted:
I '
(1.6
I
If the control s Yk are not varied, then eq.(l.5) coincides wi th the f o r
0
mul a of the total . vari ati on of the action functional, uhich i s extensively use
- 4
- ' +i n anal yti cal mechanics (Bib1.3).
>< I
- - 7
: i>
1'
- 2 4
L et the functional. J have a smooth maximum.
Then, by equating AU to zero
:we arrive at the necessary conditions of the extremum
) .
-
:) j
i t :
gS expressing, i n eq.(1.9), the total vari ati ons of the vari abl es xi in
-
,terms of isochronous vari ati ons by the formulas ibq = 6% + q A t , and equating
i to zero the terms i n A t and 6- at the times to and T, the conditions (1.9)
, I
a -
{will be reduced to the form of
/1
The conditions (1=7), (1.81, and (1.10) are necessary conditions for the
vari ati onal problem of Bolza (Bibl.4).
of the system (1.10) must be omitted and i f T i s prescribed, the second equa-
If to is prescribed, the first equati o
-
- ti on of that system.
1
) +
' d
We now note the important problem when the functional to be minimized nee
' 0 .
, ; 'not have a smooth minimurn. T hi s i s a rapid-action problem. I n this case, Fo
12 '
1 -
T
Hence we obtain the equality
and the conditions of transversal i ty
= HT = -1, as well as eqs.(1.7), (1.8)
The function R, - i n this case, is defined by the fonnula
I
The necessary conditions f or other optimal problems wi th a non-smooth ex-
tremum of the fundamental functional are obtained similarly.
I f a l arger number of controls are being considered, then eqs.(l.8) are
replaced by the rel ati ons
(1.81
Now l et the controls yk not be introduced, and l et us i nvesti gate the pro
l em with only the ori gi nal controls 4 . I n thi s case, eqs.(l.8)must be re-
placed by the conditions
- Note 1. The system of equations (1.1) and (1.7) i s a canonical system
' , with the Hamiltonian H, the pulses h i , and the coordinates
-
- and in that the addi ti onal vari abl es 4 and yk are present.
. T hi s canonical
system differs from the usual canonical system i n that H depends l i nearl y on A:
L
\
i
Since the system of equations (1.1) and (1.7) i s canonical, whi l e the
1
7
s 0
1 :
I :
1
boundary control s
function H w i l l be equal to the parti al time deri vati ve of this function with
the group of vari abl es 9 , 1 1 , 4 , t.
j ectory with continuous controls, we shall have:
1 and 1+2 are constant, the total time deri vati ve of the,&
I -
,
Hence, on any phase of the optimal tra-
i
(1.l l
,
-.
2 .where h = const and i s the i ni ti al time of that phase.
If the functions Fo and Fi do not expl i ci tl y depend on t, then H w i l l be
constant on any such portion.
i
i
2 s
-- - ' >
1-
- \
' , Note 2. A point at which the controls undergo a di sconti nui ty i s cal l ed
I .a corner. Let us vary the posi ti on and time of some corner poi nt by the quan-
Then, from eq.(1.5), the vari ati on of the fundamental func
.>
ti ti es hxi and AT.
2 4
-<#
dtional U, taken on the optimal arc before and af ter the comer, w i l l be
.I
1
4 1
This vari ati on of the functional must vanish f or the optimal arc.
we obtain the well-known Weierstrass-Erdmann conditions of conti nui ty of hi an
H along the enti re optimal arc.
Hence
1 7 4
I
! i '
1 ,
+
'
' I
For this reason, when the functions F,, and F1 are stati onary, the Hamil-
\ I 1
I toni an H w i l l be one and the same constant f or the enti re optimal arc. For ex
~
c
1
ample, f or the rapid-action problem, as follows from the above discussion, it
1
, w i l l be equal to minus unity.
- 4
,
(, ,Section 2. The Maxi mum Pri nci pl e
-
4
\ -
1
!O 1
, ,over the tubes of the traj ectori es i s widely used i n anal yti cal mechanics (see
1 'Etibl.4).
The method of i ntegrati ng the vari ati onal formula of the acti on functio
i
i .?
Let us apply it to fi nd the necessary and suffi ci ent conditions of
1 ,
I
. the extremum.
Consider any admissible arc AB of the system (1.1) and (1.7) wi th the con-, 1
!
This arc i s made to pass through any two poi nts A and B satisfyin4
1
!' ry
trol s {h 1 .
eqs.(l.3) at certai n times To and T.
N I
- ( 1
2 I n parti cul ar, the points A and B, as
2 ) 1 1
- 1 - 1 well as the times % and?, may coincide with the i ni ti al and f i nal posi ti ons 1
2 )
_ -
^ i
-
1
c :E* 3 of the optimal arc- 1
;and times of the optimal arc, and the arc AB can differ from the optimal only
iover a certai n i nterval f or which the controls
1
_ I ) ~-
di f f er from the controls ~
I c,
I
Let us now pass through the point A and the i ni ti al point of the optimal
.- 1
- I
., arc i n the ( 2n + 1)-dimensional space of configurations of the coordinates a, 1
impulses hi and time t, a certai n curve (.$: I
~
1
+) iWe note that f or the time being eq~~( 2. 1) may be arbi trari l y assigned, subject 1
- 0
? A X ~ - X ~ ( Q ~ ) , ' - i = ~ ~ ( ~ p ) , t o=t ( ap, go), ap=ap(pJ. ( p a -* -1. (2.1 >I
. ',
at
3V
,: :to the requirements that-> 0, and that the contour of ($ passes through the
'(! i
I
I
I -
r ;below.
!above two points. The addi ti onal restri cti ons on the choice of C$ w i l l be gi v&
I
I
We shal l apply the term extremal to any arc of the system (1.1) and
.(1.7).
;curve t$ included between the initial point of the optimal curve and the point
Let us now pass through points of the curve AB and that part of the
I
- i
1
7 ,
, A , the extremals corresponding t o controls equal to the control s {&I on the
.optimal arc.
itremals crossing the curve AB shall i ntersect it.
Let us continue the contour C$ beyond the poi nt A so that the ex
I
Let the rel ati ons (2.1) be such that for v = v the curve 6 shall be
,
.transformed by these extremals i nto the curve Gr passi ng through the end poi nt
,of the optimal arc and the poi nt B.
,
Lo .
1 1 The contour of G w i l l be defined by the
(2.2
1
Through the poi nts of this segment of the curve I $ l et us draw extremals
wi th control s ctst) such that these extremals, f or v = v, shall completely fix
that segnent of the curve G between the end poi nt of the optimal arc and the
point B, i f the following conditions are sati sfi ed: ~ -
a; ( t ( Qp (PI, uo))=u&
u; (%p(l.4* 4) = Uk (t (ap (PI. d).
(PI, *),
A t any t h e t the boundary values f or the control s {d] w i l l be and
- Gj.
* I
1 % +
\ - I
1- final condition:
Let us now subject the choice of the curve and the control s {d] to thc
I ,
- i
Let us write out the condition (2.3) i n the expanded form:
(2.3
(2.4,
I
7
The following notation has been used i n eq.(2.4):
I
A l l the above conditions can be sati sfi ed, because of the fact that theA
functional dependence of x : and 1; on aP is arbi trarg, the functions up of the
parameter P and $ of the time t are arbi trary, and the function t(Q',, v) i s
1
- likewise arbi trary.
- t 1 %
- <
_ _
We shall now i ntegrate the expressions (1.5) i n the above-mentioned f i el d
_ -
1 of extremals with controls coinciding with the optimal controls {& I . We ob-
~~ ~ ~
- 1
(2.6
- \ c =t
I O
u
Hen
qi =F, ( xj . uk. f ) .
It s i de is calculated on the opthal arc, the
first i ntegral of the ri ght si de over the curve
the optimal arc to the point A) the second i ntegral , over the curve Q f r om t h
end point of the optimal arc to the point B; and the thi rd i ntegral , over the
traj ectory AB.
from the i ni ti al point of
The quanti ti es Q are defined by eqs.(l.l) i n the form of
The function H and the factors i n the thi rd i ntegral of the ri ght si de are
~ ..- -
I
8
-
cal cul ated on the extrds of the fi el d under consideration and the poi nts of
thei r i ntersecti on with the curve AB. The control s entering expl i ci tl y i nto H
1 therefore coincide with the optimal controls [ Q 3 .
I
.
'
I
, I
-
1 On i ntegrati ng eq.(2.3) i n the fi el d of extremals with control s Ed], we
' !
E3y defi ni ti on of the function H [eq.(1.6)], we fi nd
. I
R
Starti ng out f rom eq~~( l . 41) ~ (2.6) - (2.8), and remembering that the op-
, 7
_ .
timdl arc and the arc AB both sati sf y the conditions (1.3)s we have:
- ' i
i J
'?n I
-
(2.91
"i ;
Hence we arri ve at the necessary condition f or the fundamental functi onal
-,
I 'to be minimum. T hi s condition i s that H shal l be maximum on the optimal arc:
-
f f j x i , I i k . f ) 2 H ( K i . )i, I l k . 4 ) . (2.10;
'i 1
7 , I The condition (2.10) i s the fundamental condition of L.S.Pontrgagh1s
,maximum pri nci pl e (Rib1.S) and i s equivalent to the necessary Weierstrass con-
:di ti on (Bibl.4).
follow from the rel ati on (2.10).
1 : 1
I
- 1
3
T ~ , More parti cul arl y, the conditions (1.8), (1.81) and (1.8tI)
Let condition (2.10) be sati sfi ed for any control s {<1, some controls { Q
: I 1
- 4
I
' 1
I
:and f or a and A i which are a solution of eqs.(l.l), (1.7) f or any q, 0 A: and
' I
Ifor these controls{& 3 .
the conditions (1.1) - (1.3), (1.7), and (1.9) are sati sfi ed, then, as shown bj
:the above discussion, that arc w i l l be optimal.
If, in thi s case, on some arc with the control s {& 1,
-
I
I
\ I
We have reached the conclusion that the max5malit.y of H f or certai n
9
,
. .
. __
trols Is) wi th xi and 11 that are a solution of the system (1.1) and (lo),
for arbi trary initial values, and the sati sfacti on of the conditions (1.1) -
(l.3), (l o), and (1.9) f or a traj ectory wi th these controls, are suffi ci ent
conditions f or the optimality of these controls.
..
-
.
) ; Note. The necessi ty of condition (2.10) f or an optimal arc can be shown
I O
I
,by me a ns of the following simple construction:
Consider the f i el d of extremals corresponding to controls equal to the op
timal controls
certai n i nternal i nterval which can be as small as may be desired.
i nterval include the poi nt (a, 11, t) f or which condition (2.10) i s to be
proved.
and to the arc AB, di fferi ng from the optimal onl y over a
Let this
Let us draw the arc AB outside the optimal arc such that it can be
ei ther tangent to certai n extremals or i ntersect them, but necessari l y at two
2 1 ,
2 (7 ~ ~
- 5 .points.
- 1
-
On i ntegrati ng eq.(1.5) over the extremals i ntersecti ng the arc AB, we ob
tab, i nstead of q.(2.6),
i
(2.61
By means of eqs.(2.61) and (2.8) we arri ve at eq.(2.9), and consequently also
at the necessary condition (2.10).
>

Section 3. Application of the Hamilton-J acobi Method
The optimal traj ectory {a] and the hdetemi nate factors { h i ] satisfy th
canonical equation with the Hamiltonian function (1.6) wi th controls {s ) whic
are to be considered as expl i ci t functions of time. This canonical system may
be solved by the Hamilton-J acobi method.
For this purpose, l et us set up the parti al di fferenti al equation
I
10 ~ - __
On the basis of eq.(l.6), eq.(3.l) w i l l be of the form
.
)
'Equation (3.19) shows that the Hamilton-J acobi equation f or the problem of the
sel ecti on of optimum controls i s a fi rst-order linear parti al di fferenti al equ
ti on.
Let, f or certai n Q (t), the following complete i ntegral of eq.(3.1) be /1
found:
(3.2
. where ai are n independent constants and a,, i s an addi ti ve constant. Then, t h
- .
.sol uti on of eqs.(l.l) and (1.7) f or these controls 4 (t) i s given by
(3.3
(3.4
.
- ~ where are new independent constants.
t
It i s generally known that the complete i ntegral of e~~( 3. l ) i s determine
'
with an accuracy to an addi ti ve constant, by the acti on i ntegral . We, there-
I '
1 ,
-' ,fore, have I
. ',
. -
W = ) Ldt+rq.
The function R of eq.(1.4*) depends on the i ni ti al and f i nal values of the var
i abl es and w i l l be a certai n addi ti ve constant.
minimized will be equal, with accuracy to within an addi ti ve constant, to the
complete i ntegral of eq.(3.1).
1"
Hence, the functi onal to be
On the basi s of eq~~( 3. 1) and (3.4), for any expl i ci t representation of a
functional on the traj ectori es of the system (1.1) and (lo?), we have the re-
i l ati ons
11
(3.5
Equations (3.5) may be used to work out programs of numerical solution of
,,
vari ati onal problems, f or instance, by the divergence method (Bibl.6).
then the complete integral of the Hamilton-J acobi ~- If ai = A : andh =a,
0
h ) ~ +
, equations i s written i n the form of (Bibl.7)
Section 4. Optimal Impulse Transfers between Coplanar Orbi ts
As an example of the application of the Hamilton-J acobi method, l et us
fi nd the conditions of optimal bpul si ve coplanar transfers between el l i pti cal
-
orbi ts i n the gravi tati onal f i el d of a spherosymmetric central body.
- _
I
- 1 .
L et us consider the motion of a rocket ship in a pol ar coordinate system
- 5 ,
>-
-
(r, v) with i ts ori gi n at the center point of the central. body.
of motion wi l l be of the form
The equatzons
- -
- )
(4.1
-. .
where k i s the gravi tati onal constant; vr and vg are the proj ecti ons of the
vel oci ty of the ship on the radius vector and transversal , respectively; B =
i = -I+ I n m; CL, i s the constant effecti ve exhaust vel oci ty; m i s the mass; y i s
i
I the angle between the thrust and the radius vector, measured i n the direc- /1
ti on from the radius vector to the velocity of motion of the ship.
% 1
- 4 I
-
;.>
..
il
1 1
4
For any possible i nterpl anetary fl i ghts, the acceleration developed by
modern rocket engines considerably exceeds the other terms on the ri ght si des
of eqs.(L.l).
impulsive cri teri on of velocity:
1
Therefore, f or the boost phase, l et us take the equations of t h
I
~-
~ ~ Z , I - = ~ ~ C O S + . vF==ijsint$, i = ~ , Cif=o. (4.1'
Starti ng out from the capabi l i ti es of rocket engineering and the allowabl
accelerations, a restri cti on on the control rel ati ve to the value of B must be
adopted :
- I
'
-
'
0 (4.2
' I ' i
' I
The equations of motion on the passive phases are obtained fromthe syste
I _
, ' (4.1) by setti ng b = 0.
Understanding the term optimal i n the sense of minimizing the consumption
of mass, we shal l minimize the functional
T
(4.3
_ I
. where i s the i ni ti al mass and mrc tne final mass of the ship, while and T
- 1
~~
- ,
'are the respective times of start and fl ni sh.
I
The functi onal (4.3) must be minimiized by means of the constrai nts (4.1)
2 ) . 1
:,>
-
.
~
'or (4,l v),restri cti on of the control of i n the form of eq.(4.2), and the
- I
. boundary conditions (Bibl.8):
Here the subscript H i ndi cates the characteri sti cs of the i ni ti al orbi t
and the subscript K those of the terminal orbi t.
used i n eqs.(h.l+) and (4.5): e = eccentri ci ty; p =parameter; w = angular di s-
tance of the peri center f r om the polar ax i s ; f = true anomaly.
I n the case under consideration, "I,, = 0 and the boundary conditions are
The following notation is
expressed by the expl i ci t rel ati ons (4.4) and (4.5).
not introduce the indeterminate constants Vp, and shall use the conditions of
transversal i ty i n the general form:
For this reason we shal l
~~ ~
0.
I
13 I
~~ ~
Here the total vari ati ons of the variables are calculated along the limit
-
' except A t are expressed respectively i n terms of Af" and bfK . I n formulating
i ng orbi ts. I n vi ew of eqs.(4.4) and (4.5), the vari ati ons of all vari abl es
-
the problem we do not consider the speci fi c motions along the l i mi ti ng orbi ts,
so that the vari ati ons of AfH , AfK , At and A t w i l l be independent.
obtain &, HK = 0 and, since eqs.(l+.l) are stationary, we obtain H = 0 along
the enti re optimal traj ectory.
the l i mi ti ng orbi ts are proportional to the ri ght si des of eqs.(4.1), calcu-
Hence we
The variations of vel oci ti es and coordinates o
.
l ated on the l i mi ti ng orbi ts f or B =0.
We arri ve at the following conditions of transversal i ty:
-i j
2' :bits, respecti vdy.
- I
- t
Starti ng out from eqs.(L.l') l et us set up
- 1 1
ti ve phases:
H=b(---l +A,cos*j,+).,s
the function H
for the ac- A
We now wri te the Hamilton-J acobi equation:
* U.
-+I(-] dW +2;F; dW
at
The complete i ntegral of this equation is
The i ntegral s of eq~~( 3. 3) and (3.4) yi el d
f
(4.7
We recal l the conditions of nxdmmH on the optimal traj ectory.
Writing
out the fi rst and second parti al derivatives of the function (4.7) with respec
~
0
to 5 and $, we come to the conclusion that the acti ve phases can include the
tlE3
traj ectori es: (a) of the programmed thrust at -
da
or
(b) phases of mardmMl reacti ve acceleration G at
.
(4.5
Since H = C
and H is a l i near homogeneous function of B, the sign of i nequal i ty for the
deri vati ve -% i s dropped and the indetermhate mul ti pl i ers here are also de-
%
telmined from eqs.(4.9).
1,
Equations (4.8) and (4.9) show that the i ncl i nati on angle f of the thrust
, remains constant f or any powered section of the traj ectory. For each such
-,, phase or section, eqs.(4.l') and (4.8) w i l l yi el d
(4.1
- 1
-.
It follows more parti cdarl y from eqs.(i +.l Of t f i at tim consumption--of mass
during the acti ve phase i s determined by the required increment of vel oci ty an
7- 1
-
does riot depend on the parti cul ar law of burning.
On the passive phases, the controls over B and $ are shut down. These
phases admit of the maximumprinciple. L et us fi nd an expression f or the inde
terminate factors on the passive phases by the Hamilton-J acobi method.
The Hamilton-J acobi equation i s of the form
dW
dt
-
(4.u.
The complete i ntegral i s
W=a,t+ccpp+@(v,, &H r . %
where @ i s a certai n function of these arguments.
On the basi s of eq.(3.4), this yields
'For this reason, ha = const f or both passive and acti ve phases.
vi rtue of
the Weierstrass-Erdmann conditions, 1, is the same constant over the enti re op
timal traj ectory.
-
.
For the passive phases, we have
/complete i ntegral i s written i n the form of
= 0, hence L = 0, and by 9.(3.6) the
.
I
i o I
2 % Starti ng from eq.(3.4), we obtain expressions for the indeterminate multi-
( pl i ers:
' < L et us take the sol uti on of eqs.(4.l) on the passive phases i n the follow-
1
L _
? ' i ng form:
\ $
,
'where e, p, and W are the eccentricity, parameter, and angular distance of the
peri center of the transi ti onal el l i pti cal orbit from the pol ar axis, and f i s
. the true anomaly on this orbi t.
.
'-
' I
,
We note that the true anomaly i s expressed i n terms of the time t by meanr
' 1 .
,- ,of the rel ati ons (Bibl.8):
..
I
9 1
c j iwhere
,
i s the time of transi t through the peri center and a i s the semi-major
,
1
I
I
I
. axi s, which for an el l i pti cal orbi t is equal to p(1 - ea)-'.
Let us denote by a ( k =1, 2, 3, 4) the Keplerian elements e, p,
and
- 1
,Equation (4.12) then takes the form of
1
16 ..
.
~_
If we put t = to i n eqs.(k014) and (4.15), we arri ve at the expressions
To obtai n eqs.(4.16), the Keplerian elements must be represented in t er n of xr
Here, on the basis of eqs.(4.15), we shd
-
. .by the ai d of eqs.(4.14) and (4.15).
i
where Y i s a certai n function of these variables.
8 ' The quanti ti es e, p, w, and f will now be expresse, on the basi s o
' J ,
eqs.(4.14), expl i ci tl y i n tenns of q. We wi l l , therefcre, have the i denti ti e:
On the basis of eqO(4.l8) we have
- _ - _ - _ -
Equation (4.12) and the Hamilton-J acobi equation w i l l f i nal l y yi el d & = -H.&
I n this case, therefore, & = 0, and the expressions f or the indeterminate mul -
ti pl i ers w i l l contain only three b-determinate constants.
a4r
ax,
To determine the derivatives - , l et us di fferenti ate eq.(4.14) wi th
- 5 .respect to a We then have
- 1 - ,
1
i
-1 q2
t i The coeffi ci ents Bl j vanish f or i f j, Bl 1 = Be2 =k p , 833 =-p'-' (1 +
I
- I
j + e cos f)Z, b4 =1.
I
Making use of the obvious equality
i
I
and assuming that e f 0, we obtain, af ter simple calculations,
-
> l '
,
_' 1 '
? ,
1- I
- ?
Equations (4.19> are of great independent significance, si nce they deter-
'mine the vari ati on of the elements e, p, 0) f or a smal l vari ati on of the posi-
' ti on of the point and of the vel oci ty of the satel l i te going i nto orbi t.
-
-
_ I
~ _ _
- .
I ntroducing the following notation f or new constants:
- 1
A = - &le-lpll2A3, B = k- Ie--Ip1/2A,, D = k-Ie-1pV (ea -
we shal l have, from eqs.(~.16) and (4.19):
-l
% >
.%
; i
5 r )
2 -
(4.20
h=Ac os f+Besin f,
h= - A sinf(2 + e cosf) + B ( 1 +e c o s f ) + D
A,= - k f ~ - ~ ' ~ ( l +ecos f)* [Asinf(A +ecosf)-'-
a, = - &ep-'" A.
- B - D(1 +e c os f ) - ' ] ,
. -
The constants A, B and D coincide with the Lawden constants (Bibl.9) f or a
Our discussion has now establ i shed the connection noncircular transfer orbi t.
of the Lawden constants with 1; and the elements of the transfer orbi t.
our study was i n a pol ar coordinate system,
than Lawdents corresponding formulas (ab1.9) f or the indeterminate mul ti pl i er;
Since
eqs.(4.20) turned out simpler
~~
18 . .
.
.
4
on a noncircular orbi t.
L et us obtain expressions f or the indeterminate mul ti pl i ers that w i l l be
.
sui tabl e f or any transfer orbi t, including a ci rcul ar one, f or which e = 0.
the new elements q = e cos w and 1 = e sin w,
1
,
Substi tuti ng for e and
1
1
we now wr i t e eq~~(4.14) i n the form of
I
1
I
v,=kp-l n(qsl n 'p-Icoscp), ~ , = k p - l ' ~ ( l $-qcoscp$-Isl ncp),, 1
r =p( 1 +qcoscp+fsincp)-*. I (4.21)
1 I
Di fferenti ati ng eqs.(4.21) with respect to a, we obtain equations f or
.
determining the parti al deri vati ves :
i
- y p - ' ( q si n 'p-lcos
1
I ( 1 +gcos 'p +i s i n y )
The matrix of coeffi ci ents I ~ G J 11 i s of the form
It-'p'" 0 0
/1 8 k - V n 0
0 -p-'(1+gcos'p+fslncp)'
Hence, we fi nd the following apressi ons f or the derivatives:
Equations (L. 21) determine the vari ati on of elements q and I f or a small
'
vari ati oc i n the vel oci ty and position of the ship.
19
I ntroducing the following notation f or constants:
M =- k-'p'*A,, N=&-'p'nA,, Q = 2 k - l PA
we shal l have, from eqs.(ko16) and (4.2l),
Putti ng q =8 = 0 i n eqs.(4.22), we obtain, f or a ci rcul ar transfer orbi t
(4.23
I
As= - kp-yl (M sin 'p - IV cos 7 - Q],
A,= 0.
The constants M, N, and Q here coincide with the Lawden constants (Bibl.9
f or the case of a ci rcul ar orbi t.
responding Lawden formulas.
f act that our formul as,hstead of the true anomaly f, use the l ati tute argu-
Equations (4.23) are simpler than the COI-
The difference between these al s o consi sts i n the
ment cp which corresponds better to a ci rcul ar transfer orbi t. The present
~ s t u d v has al s o es t abl i s hed t h e connection of t he Lawden const.nnt.s wi t h h? and
>
.-
--
,') 1 The conti nui ty of the indeterminate mul ti pl i ers, thei r expl i ci t expres-
,
The general representation [ eq. (4.22)] obtained above i s parti cul arl y
convenient f or considering quasi ci rcul ar transfer orbi ts.
*.
r o 4
i si ons given by eq~~(4.9) and (4.22), the boundary conditions (L.4) - ( 4. 5) ,
I
, the conditions (4.10) f or vel oci ty jumps f or each acti ve phase, and the condi-
ti ons (4.6) of transversal i ty, yi el d a number of equations suffi ci ent f or de-
termining al l unknown quanti ti es f or a prescribed number of impulses.
An i nvesti gati on of these equations would be a separate problem, beyond
the scope of thi s paper.
20
8 NASA 'M' F-9294
I
i BIBLIOGRAPHY
, 1. Ykiele, A. : General Variational Theory of the Fl i ght Paths of Rocket-
I
i Powered Aircraft. Missiles and Satel l i te Carriers. Astronautica acta,
-
I
1
\ * VoLIV, No.4, 1958.
$) i
12 1 malfnykh trayektoriyakh raket ). Priklad. Matem. i Mekhan., Vol.XXV,
j 2. Leytman, G. [George Leitmann]: On Optimal- Rocket Traj ectori es (Ob opti-
1 :
' .
No.6, 1961.
3. Uitteker, Ye.T. [E.T.Whittakerl: Analytical Dynamics (Analiticheskaya
I * dinamika). ONTI , Moscow - Leningrad, 1937. I
4. Bl i ss, G.A.: Lectures on the Calculus of Variations (Lektsi i PO vari ats-
ionnomu ischisleniyu). I nternational Publishing House (I L), Noscow, 19504
~ 8.
,
Ye.F.: Mathematical Theory of 0pti.ma.l Processes (matemakicheskaya t e w , 1
!
optimalfnykh protsessov). Fizmatgiz, Moscow, 1961.
Pontryagin, L.S., Boltyanskiy, V.G., Gamkrelidze, R.V., and Mishchenko,
i
I
Brekvell [J .V .Breakwell] : Optimization of Trajectories (Optimizatsiya i 1
I
trayektoriy). Vopr. raket. tekh., No.1, 1961. [North American Aviation, '
Inc., Report NO.AL-2706, August, 19571 .
I
Gyunter, N.M.: I ntegrati on of First-Order Parti al Di fferenti al Equations 1
1
I
(integrirovaniye uravneniy pervogo poryadka v chastnykh proizvodnykh)
ONTI , Gostekhizdat, Moscow, 1934. I
I
I
Subbotin, M.F.: Course i n Celestial Mechanics (Kurs nebesnoy mekhaniki). 1
I
Vol.1, OGTZ, Gostekhizdat, Leningrad - Moscow, 1941.
- 1
9. Lawden, D.F.: I nterplanetary Rocket Trajectories. Advances i n Space I
I
Science, Vol.1, 1959. I
,
Received 10 May 1963.