Ada 015815
Ada 015815
Ada 015815
AD-A015 815
Prepared for,:
5 September 1915
H DISTRIBUTED BY:
1<.0
R~pro~yIX4
AA rn0n
NATIONAL TECHNICAL
INFORMATION SERVICE
UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE (When fiata Entered)
INSC'.UCTIONS
RIREAD
REPORT DOCUMENTATION PAGE -'FORACoMPNIRTINGSF'OI!1
I. REPORT NUMBER
_ _/__
ESD-TR-75-255 _ _ _ _
2.
_
GOVT ACCESSION NO.
_ _--,4
.TALOG
CPETSCTLGNME o,...e715
NUMBER
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT, PROJECT, TASK
AREA & WORK UNIT NUMBERS
Lincoln Laboratory, M.I.T.
P.O. Box 73 Project No. 8X363304 I)215
Lexingtop, MA 02173 Program Element No. 6:0304A
11. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE
Baliistic Missile Defense Program Office 5 September 1975
Department of the Army 5_September_1975
1320 Wilb.nVA
Boulevard 13. NUMBER
Arlington, 22209 • OF•• PAGES
I
14. MONITORING AGENCY NAME & ADDRESS (if different from Controlling O)ffice) 15. SECURITY CLASS. (,ftha.. re'l,til
ElectronicAFB
Hanscom Systems Division Unclassified
Bedfsord MA 01715u. DECLASSIFICATION DOWNGRADING
edford, MA 01731 SC EDULE
17. DISTRIBUTION STATEMENT (of the abstract entered in Block 20, if different from Report)
None i
19. K EY WORDS (Continu, on reverse side if necessary and identify by block number)
79. ARSTRACT (Contiaue on reverse side if necessary and identify by block number)
This report presents the relations necessary to define the motion of a target in the gravitational
field of the earth. In order to express this target motion in the frame of a radar, an appropriate
set of coordinate systems (and transformations) is introduced. Target tracking in the form of a
Maximum Likelihood Estimator is discussed. The problem of interceptor miss distance is treated
from the standpoint of the uncertainty volumes associated with estimated target state vectors.
t ~~FO
FDDRM' 1473 EDITION OF I NOV 65 IS OBSOLETE
I JAN 73 / UNCLASSIFIED
"SECURITY CLASSIFICATION OF THIS PAGE (1hen Pain Ent.-,i.
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
LINCOLN LABORATORY
B. L. DIAMOND
Group 37
M. E. AUSTIN
KR EMS
5 SEPTEMBER 1975
LEXINGTON IMASSACHUSETTS
7. =71 71l11177
ABSTRACT
, :1
Ai
CONTENTS
Abstract iii
Section 1-- Introduction 1
1.1 Program Purpose 1
1.2 Program Features 2
Iv
•, V , •i
CONrTENTS (Cont'd)
vi
.......
...................
II
C(ONTENTS (Cont'd)
.
8.3 landover Error Ellipsoid
References 90
•.2
m-r
"vii
Li
SECTION 1--INTRODUCTION
I
areas. Originally ARIES was used in the strategic BMD area to estimate the
metric state vector (position and velocity) of tracked targets, and then to
extrapolate ahead in time to determine an intercept point for an ICBM. The
1q
More recently, a modified endoatmospheric version of ARIES, known as the
HWLPTR (Hostile Weapons Location Projectile Tracking Radar) Program, has been
used in the tactical area for the evaluation of a radar's performance in
"backtracking" an incoming artillery or mortar shell to determine its point
of launch. The resulting point estimation error CEP values also assist in
the evaluation of the drag model error effects and in the study of the overall
.. 1
iJ
equations) to obtain realistic dynamic RCS simulations. Constant
also available.
b. Noise and propagation effects --- radar measurements are corrupted
by receive; noise (S/N dependent), range-independent noise effects
algorithms.
S. Target Discrimination--- (Not presently implemented.) Conceptually,
discrimination algorithms would be implemented to det-rmine whether
a particular simulated target constituted a threat to the defended
area.
~ ~ ~• ~~w
-~ •,•.jg~ ,•4.,_z
-'•~,, ....
ilk-•_ . .. .,.. .
conduct a command guided intercept. Currently, the program extra-
polates the estimated and true target state vectors to various time
(or altitude) points after termination of track to obtain miss
distances. Miss distance statistics are computed from the accumiu-
lated miss distances observed on a series of Monte Carlo tests.
simulation "scenario" For output, an 8½1" x 11" ARIES Test Report (see
Appendix II of Reference 1) is generated which provides the engineer with
descriptions of his input model parameters and scenarios, along with the
resultant shiulation data and statistics. The outputs are all organized into
Flogical sections which are indexed -for ready reference. Outputs from ARIES
also include trajectory plots superimposed on a world map, plots of true and
measured target cross-section, and a radar measurements tape containing metric
and RCS data fo-,c processing by other programs.
4
1.3 Program Documentation
subroutines. All of the options available to a user are discussed and the
The second and third volumes contain all of the relevant mathematics and
distance calculations, etc.) are in the second report. The third report is
cross-section.
1.4 Organization of This Report
In Section 2, some aspects of the WGS66 earth model are presented. This
likelihood estimate of the trajectory which best fits the accumulated radar
measurements. From the state vectors generated by trajectory fitting on
several Monte Carlo runs and the known (true) target position, one can develop
an error covariance matrix representing the uncertainty in predicted target
position at an assumed intercept point. This covariance matrix and the related
6
SECTION 2--EARTH MIDEL
The model for the earth used in the ARIES Program is taken directly from
the 1966 World Geodetic Survey (WGS-66). In this model the earth's shape is
ellipsoid is defined by the Equatorial and North polar radii (semi-major and
semi-minor axes). These values are:
Re - 6378.145 Km (2-1)
Rn = 6356.760 Km (2-2)
Two other parameters of the ellipse, the flattening factor f and the eccentric-
"Ell1
\-.
f•(2-3)--- E--=
e
The oblique view of the earth given in Figure 2.1 and the cross-sectional
view given in Figure 2.2 aid in the definition of positions on the earth
surface. Longitude is defined as the angle between the plane containing the
Greenwich meridian and a plane containing any other meridian; longitude is
There are two latitude angles defined for the earth ellipsoid, as indicated
in Figure 2.2. Both angles are referenced to the equatorial plane. The
geodetic latitude € is defined as the angle between the local normal to the
earth ellipsoid and the equatorial plane. Mathematically, it is defined by
tan€ 1
1 z; ir/2 o 7i/2 (2-7)
(i-f) 2 ?n
8
I I•
NORTH [S31SS
GREENWICH
MERIDIAN
EARTH X- LONGITUDE
118i-3-1790911
2r L
2 2 2
9
Note that the extension of the local normal to the equatorial plane does not
in general pass through the center of the earth (exceptions occur when 4-0 and
*=T/2). The geodetic latitude is the usual survey latitude. It is also the
latitude determined from astronomic observations.
tanoc = z (2-8)
c n
expressed as
R = co2c
nn1 (2-10)
41-e2CO2sZc
R Rl-cesin24i
R (2-11)
I10
101
.............................. *.7
ellipse, it is desired to find the minimum of
= 4(,o-n)
,h 2 + (zo-z) 2 (2-12)
S2
2 + z R2(2-13)
(1-f) 2 e
Shh (2-14)
where0 0 l 2 osA
r0 = /r 2 +2 (2-15)
and
A detailed evaluation of the error in this simple relation for h° has not been
made; however, sample computations indicated worst-case errors of the order of
ii _
10 meters for altitudes around 2000 Km and about 4 meters for altitudes around
700 Km.
For the purposes of the ARIES Program, the above model for ho is probably
adequate. However, a refinement trý the relation can be made by taking first
order perturbations of n and z around ho. Such a procedure yields
h h 2cos2A)
(l-c (2-17)
V (r) (2-18)
12
• " ' "." -: .' '". , . .
and r is the radius from the center of the earth. Additional contributions to
the gravitational potential result from the mass distribution of the earth.
These perturbations of the point mass potential are expanded in spherical
harmonics. One series of harmonics depends only on latitude; these are called
the zonal harmonics. All remaining terms of the harmonic series have a
dependence on both latitude and longitude; these are called tesseral harmonics.
The ARIES Program only considers the zonal harmonics in the gravitational
potential model. That is, the potential is given by
9 (Ren
V(r) =
r n=2 \~ i~(sinA]
n (2-20)
where Cn is the coefficient of the nth zonal harmonic and Pn(, is the
Legendre polynomial of the first kind of order n. The angle A is the angle
between the equatorial plane and the vector from the center of the earth to
I
the target position. The first order harmonic is zero as a result of the
zonal harmonic coef/iints, th.)ugh the ninth, are available in the ARIES
Program; they are easily implemented in Subroutine GRAVTY.)
1i
13z
Values for the harmonic coefficients in the gravitational potential model
are
C2 1.08271 x 10 -
C3 =-2.630 x 10-6
Gb = +0.53 x 1
C9 =0.24A x 100 (2-21)
14
• i• i•• ' .!!•f!•!)i¢ •",' -••! j•..w.-............. .. -. ... • . • .. . i .. . , . .. .. h'• "i
Several different coordinate systems are used in the ARIES Program. The
various operations performed i, the Program (e.g., trajectory integration,
target tunbling, radar measurement generation) all have preferred coordinate
frwaes for their evaluation. In the sections below the various coordinate
systems employed by the ARIES Program are defined. The transformations
between the various coordinate systems are presented in Section 5.
Centered Inertial (ECI) frame. The origin of this frame is at the center of
the earth, the z-axis passes through the North Pole, end the equatorial plane
of the earth lies in the x-y plane. At the reference time (typically the
launch time of the target), the x-axis of the ECI frame passes through the
Greenwich meridian. The x, y, and z axes shown in Fig. 2.1 form an ECI frame
if the axes are held fixed in space rather than rotating with the earth.
The equations of motion for ballistic targets have their simplest form in
the ECI frame. Consequently, it is the preferred frame for trajectory genera-
This frame is essentially the same as the ECI frame, except that it
rotates with the earth. Usually the x-axis is defined to point through the
S~15
......... . . ...
equator at the Greenwich meridian and the y-axis is defined to point through
There are four different coordinate systems associated with the radars.
Z"',Two of these are peculiar to a phased array radar (XRF, Y , Z~ and R, U, V).
One coordinate system (R, A, E) is associated with a conventional, mechanically
steered radar. The fourth coordinate system is an earth surface fixed frame
or radar XYZ frame. Radar measurements are typically made in either RAE or
RUV coordinates. The cartesian frames are useful for many computations and
provide the necessary cartesian frames for definition of RAE and RUV coordin-
ates.
The Radar Cartesian Frame is defined such that the X-Y plane is tangent
to the earth ellipsoid and the Z-axis is pointed outward along the local
vertical. The X-axis points East and the Y-axis points North. This set of
coordinates is illustrated in Fig. 3.1. Usually the origin of this coordinate
frame will coincide with the radar location or with the target launch point.
If the radar is not located on the earth ellipsoid, but is actually at height
H above the ellipsoid, then the radar X-Y planeL will be parallrl to the tangent
plane to the earth ellipsoid.
16
61-L .9 ii Ib.L'ý-!-mw&ýý
3.3.2 Radar RAE Coordinates
The RAE coordinates are defined relative to the radar cartesian coordin-
ates as indicated in Fig. 3.2. R is the distance (range) from the origin (the
radar) to the target. The azinuth, A, is the angle from the Y-axis (North) to
the projection of the range vector, R, into the X-Y plane. Azimuth is measured
clockwise from North toward East. The elevation angle, E, is measured (positive
up) from the X-Y plane to the range vector, R.
The cartesian coordinates (XR, YRF ZR) are defined with respect to the
phased array radar face as indicated in Fig. 3.3. The radar face lies in the
XF-YRF plane, with the XRF-axis horizontal and the YRF-axis pointing upward.
The ZRF-axis points outward along the normal to th,,• array face. i-
The orientation of the phased array face is defined by the azimuth and
the elevation of the phased array boresight (i.e., the phased array Z-axis).
Azimuth and elevation, in this case, a~r measured in the same sense as defined
The RUV and Ra8 coordinate systems are def 1 relative to the Phased
Array Radar Face Cartesian Frame. R is the di:, .ice (range) from the origin
(the radar) 1:,t-ie target. Target direction in angle is measured by the
direction cisines with respect to the XRF and YRF axes. These direction
17
. . . . . .. . . . .
•,: L:•... . ...• ....• ... ........ .... L. ...,_ / • • : • • . . . .. .. .. .. . ...• . . . . ... ... .....
... •... ...... •.
constant; the intersection of these cones defines the direction to the target.
(In the usual spherical coordinate system (Ro4) 0 defines a cone about the z-
axis and 0 defines a plane containing the z-axis; the intersection of the
conical surface and the plane defines target direction.)
The angles a and 8 can also be used to measure the direction to a target
with respect to the Phased Array Radar Pace Cartesian Frame. In this case, a
and B are the arcsines of U and V, respectively. Note that (n/2-a) and (-n/2-8)
are the direction cosine angles with respect to the XRF and axes. TheR
phased array broadside direction is characterized by U = V = 0, which corres-
The target tumbling and RCS generation program, utilize several different
1. Principal body axes frame - this frame has its origin at the center
18
3. Local horizontal frame - this frame has its z-.axis in the direction
of the ECI velocity vector and its y-axis normal to the trajectory
plane.
These frames are inter-related by the target Euler angles. The local
horizontal frame is directly defined from the target position and the target
II
ki
.I
19
.....................
L
118- 5-11101
yynrt)z(upr) 1
Fig.~~~~~3..eadraAscodiats
pm AXIS O
ROTAT20
~~RF
21
. .
.. .. .... . . .......
.- I K
VV(r) (4-1)
and V(r) is the grav.tational potential function defined in Section 2.2. Tar-
get position is given in the ECI coordinate frame. The three cartesian
22
GX
x--- 1 (4-3)
r3
-Z91 9 (4-5)
r3 r3
where
=9 (R e snA (4-7)
r =jX_
2 +y_+_Z2' (4-8)
and
siPA z (4-9)
r
The following recursion relation for the derivatives of the Legendre poly-
nomials
23
I'M;
Given the target position and the target velocity in the ECI frame at a
particular reference time, together with the above definitions for target
accelerations, the motion of the target in the ECI frame is completely defined
over the entire trajectory. This statement, of course, ignores any effects
The ARIES Program does not include the ta.-get deceleration effects caused
by atmospheric drag. However, for completeness, a brief discussion of
aD -- A Pv g -2)
where
CD coefficient of drag of the reentry body
A cross-sectional area of the reentry body
P atmospheric density
24
LL
~ ... . . . ~ . . . . . . ..........
v = body velocity (magnitude)
w = body weight
g = acceleration of gravity
The parameters CD, A and w depend on the shape and size of the target and
(4-13)
25
. . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . . ~
. .. . . . . . . . . . . . . . . .... . .
.......
771 W7.~ W-..d..
. . . Tl,
val were known, the acceleration over the interval could be approximated by a
straight line:
At t
x(At) = x(O) + v(O)At + I dt I a(T)d¶ (4-16)
0 0
and
26
i ..-. ." : '• " • i. i:•' . . . . .. ... '
is made:
27
:W4
* * 1, 1~ j,
Al
.I
28i
L.
*W~Ž~.~
~~'.I.RŽA,~ i
SECTION 5--COORDINATE TRANSFORMATIONS
At the reference time (usually t = 0), the ECI and ECF frames are
coincident. Both have their x-axes through the equator at the Greenwich
meridian and their z-axes through the North Pole. At a later time t the ECI
frame is unchanged but the ECF frame has rotated with the earth through an
angle Wet. Let the subscript I indicate a component in the ECI frame and the
subscript F a component in the ECF frame. Now it is easily shown that the
k 29
& . ... .- ... . , .. .•
or, by defining the coefficient matrix as 6(t) and the position vectors as x
and X
F
F ((t)I-2)
If one takes the time derivative of Eqn. (5-i), the following transformation
of the velocity vector is obtained:
=WG0jk [w 0 = G(t)
j{ (5-3)
where the dot over a vector (or, a vector component) indicates a derivative
with respect to time. The second term in Eqn. (5-3) results from the contin-
uous rotation of the ECF frame relative to the inertial frame. For the ARIES
Program we require only position and velocity transformations between ECI and
ECF frames. Target accelerations are used only for trajectory integration
(ECI frame) and therefore do not have to be transformed.
The transformation matrix G(t) represents an orthonormal transformation.
G(t) 1 = G(t)
X G(t)T x (5-4)
*Script characters denote matrices (lower case for cohlmn matrices; upper case
for N x M matrices).
30
- -i ††††††††††††-†,†;††,†††,††..,,††.†††††,†.†††
% .... •.....•,-i• .,"...... :-- | i••' ..... .SaS.&,.A....
-.,-. .........-'
for positions and by
JzweYF
forvlcii sr k ~) k x -G-t
- +Wx (5-S)
for velocities.
Subroutine ECIECF (XI,XF,TIME) is used to transform the state vector XI
in ECI coordinates to the state vector XF in ECF coordinates. TIME is the
interval in seconds since the two frames were coincident. Subroutine ECFECI(XI,
XI,TIME) performs the inverse transformation.
east from the Greenwich meridian) and at height h above the earth ellipsoid.
The process of transforming from the ECF frame to a radar XYZ (East, North,
Up) frame can be visualized as the following sequence of operations:
1. Rotation by angle x about the z-axis.
2. Rotation by angle A about the y-axis.
.'
3i'i
._ ......
_ _-- _ _ .~ .%V .
0 0 1
2. cosiy 0' i
C 1 0s where tan c=(1-e 2 )tan4 (5-7)
Fh ¢R
in
-sin(€-c) 0 coso(4-Pc)J
35. -
-:'h= : (5-10)j
h 0 (5-10)
6. [ C 0
L0
5.-h . -
32
7-7 M!
- - .
relations are:
XR = AxF + b (5-13)
where
[ -sinX cosX 0 1
A= j-siný cosX -sino sinx cosý (5-14)
Icosý sinx si
L cos" cosX sinij
and
0
b = Rh sin(-ýc) (5-15)
-Reh cos(0-%c)-h
R= AkF (5-16)
ECF, is
33
L:'A
xFY1 A xR Ab (5-17)
and
k, A kR (5-18)
latitude and height above the earth ellipsoid. The A and b matrices are
formations.
The Radar Range, Azimuth and Elevation are defined in Section 3.3.2. They
R = jx2+y2+z2
-1
A =tan (x/y), 0 A 2w (5-19)
The proper quadrant must be determined for A so that A ranges from 0 to 2w.
34
=(tR-zR) (-0
These transformations from Radar XYZ to Radar RAE are performed by Subroutine
L XYZRAE.
The transformation from RAE to XYZ is performed by Subroutine RAEXYZ with
x =R sinA cosE
y =R cosA cosE
z =R sinE (5-21)
and
Jk S sinA + yA
S cosA -xA
where
353
5.4 Between Radar XYZ and Phased Array XYZ
The definition for the Phased Array XYZ coordinate system was given
previously in Section 3.3.3. Basically, the orientation of the arrayIface is
defined by the azimuth A• and the elevation lp of the normal to the arrav face.
The array face lies in the xp-yp plane with x horizontal and vp up. The z 5
axis is aligned with the array normal. (The subscript p has the same meaning
as the subscript RF used in Sections 3.3.3 and 3.3.4.)
If we start with the arrav face coordinates coincident with the radar x)":
i I i
frame (that is, the array is pointed up), then the following sequence of
transformations gives thie desired Radar .ý'C' to Phased .,rrav X.Y"' trans formation:
x -1 1
0Xp 0
1 0 0(0
sinE -cosE os.\
s in\ -sinA\
cosA 0 (5-24)
or,
A rpR
-1p 5 -2 5
where
4
AiPrLI
[
/-io
c
snl
sPnEr
nAs,5-26,
-cos:.p Sin:.L, cos) p
o
. .,
L.sirL cosE• cos-ýA cosE, -i nLJ
L p'
M11
Note that when Ap= E= 0 (i.e., the phased array is pointed north), Xp=-XR,
= zR and zp YR' as would be expected from the definition of the phased
k2 =A (5-27)
p rp R
array normal.
The Phased Array range and "angle"' coordinates RUV are defined by:
R "f&2+y2+z2
ypp pP"
U = xp/R = sina, -1 s U 5 1
37
L ............ . .
R = p / R +z )/R
p p Pp p
U U
IJ)/R
p
S= ,-VR)/R (5-29)
The transformation from Phased Array XYZ coordinates to Phased Array RUV
coordinates is performed by Subroutine XRFRUV(XP,RUV).
The inverse transformation, from RUV to Phased Array XYZ, has the
following relations:
X = RU
Y = RV
Z - R1-U2-V2 - RW (5-30)
and j
u + RM
,U
kV + RV
R(UJ+VV) (5-31)
Z- - -'
The target tumbling and RCS generation programs utilize several different
coordinate systems which are unique to this particular subject. These coordi-
-• 38
these coordinate systems and the transformations among them are presented in
Section 7 of Reference 2.
to transform a target state vector in the ECI frame to a target state vector
in the Phased Array RUV frame, the following sequence of transformations would
be performed:
For example,
I
1. ECI to ECF (ECIECF)
The m.n.es in parentheses are the names of the ARIES Subroutines which perform
Radar (for use in multiple Radar coverage) and TYPE identifies the radar as
either a dish radar or a 3ed array radar.
•4
39
other frame. More than orn• subroutine may have to be called, as in the above
example, but the key point is that the capability does exist.
then zero. A phased array radar, on the other hand, is assumed to have the
capability of viewing a sector in azimuth and elevation about the boresight
pointing direction. For example, if the boresight azimuth of the phased array
is , then one-third hemisphere coverage is achieved by applying azimuth
bounds of Ap - 7/3 and A + T/3 and elevation bounds of 0 and 1T/2. Both
40
ILI•
7777
azimuth and elevation are checked against the phased array radar coverage
bounds to determine whether the target is within the field of view. Note that
it is easier to define phased ar-ray coverage limits in azimuth and elevation,
rather than the normal phased array U and V coordinates.
4'
",
. S . . .. ~
.... .
.......
...... . .
SECTION 6--TRAJECTORY GENERATION
3. Type of trajectory
c. Minimum energy
42
state vector subject to the launch, impact and type of trajectory constraints
in Section 6.2. Finally, the procedures employed in perturbing the initial
state vector to obtain the desired precise state vector are presented in
Section 6.3.
the geometry of the prcblern as viewed in the plane of the trajectory. (Note
that this plane is only defined in the ECI frame.) Since two of the trajectory
r r Tecoso
a le2) (6-1)
where a is the semi-major axis, e is the eccentricity of the orbit and 0 is the
central angle measured from apogee (see Fig. 6.1). Note that the true anomaly
is measured from perigee and is equal to (ir-o). The radius can also be defined
in terms of the eccentric anomaly Ec:
43
~~~~~-~~~l __ _ __ _ __ _ __1__ -_ 3z -- __
TRAJECTORYI
LANC IMAC
00* 0
44
r a(l-ecos) (6-2)
If these two definitions for r are equated, it can be shown that true anomaly
Mean motion:
7n = • (6-4)
Mean anomaly:
M = n(t-T) (6-5)
or.,
M Ec-esinEc (6-6)
where T is the time of passing perigee and G is the gravitational constant de-
T 2w•/n (6-7)
The total energy and the angular momentum are constants Lor the orbit. That
is,
2 +r 2 6 2 ) (6-8)
rG
-(t-- constant
45
L......................................~
1 ,VFh
and
r2 6 H
- constant (6-9
= R
e (6-10)
2
(x-ae)2 Y a2 (6-11)
2
l-e
Then cowpute the unit vectors tangent to the circle and ellipse at the launch
point (Recos0o,Resin~o)0
The elevation angle E is now determined as the inverse cosine of the dot
product of the unit vectcrs iand 'ce:
1-ecoso
cosE=~.c 2-2cs (6-14)
e =le
0
46
Use of the identity cos2E = 2cos 2 E-l, yields
i-e 2 -2ecoso(l-ecoseo)
cos2E = (6-15)
l+e 2 -2ecoseo
Application of the Law of Cosines to the triangle formed by the vectors from
the foci of the ellipse to the launch point (see Figure 6.1) yields:
a = Re (l-ecosOo)/(l-e 2) (6-18)
is used. (This is Eqn. 6-1 evaluated at the launch point r - Re, 0 = 0.)
Inspection of the expressions for cosc and co52E demonstrates that a = 2E.
2a-Re
snRe RRee ' (6-19)
sin00 sin(Tr -2E -0
0 0
47
or
R F ___
a [r sT-eu
si G(E+%Jo 0
(6-20)
2ae s m-Ffi• =
2 a-Re.(.-21 (6-21) ",<
The use of Eqn. (6-20) in Fqn. (6-21) yields, after some manipulation, the !:
sine
e = i (6-22)
sin (E+O T
0I
If the total energy at apogee is equated to the total energy at perigee, then
it can be shown that the angular momentum, H, is given by
2) (6-23)
H = VGal(-e
Now, the total energy at launch can be equated to the total energy at apogee
(r= a(l+e) at apogee):
2 G _1 H2 G G
Re 2 a 2 (l+e) a(l+e) -
or,
VL = .2aV (6-25)
48
The use of Eqn. (6-20) for a gives the desired result:
vRer2'G sino
[[sino+sion(2E+Oo) (6-26)
L
SCos' _e(6-27)
ECI 2ir-ECL
(6-28)
From ECL and ECI, the mean anomalies at launch and impact can be computed.
The difference between the two is
Use of this value of AM, then yields an expression for time of flight
This completes the sunmmary of the elliptic orbit parameters for object
field.
motion in a central force
49
6.2 Trajectory Initialization
estimate ef the target state vector. In this Section, the equations will be
Let the launch point have longitude XL and geodetic latitude L and let
the impact point have values xI and ýI. The geocentric latitude is readily
computed from the geodetic latitude. A unit vector from the center of the
During the flight of the target the earth rotates an amount wetf, where we is
the rotation rate and tf is the time of flight (initially unknown). The impact
longitude is rotated by this amount
The earth central angle, 200, between the launch and impact points is given
by
y
50
V.Y
L *...
. .. . .. . . . . .-
. ,.
. . . . . . ....
.
. . . . . . . . . . . .
or
This vector is tangent to the unit sphere and in a direction from launch
Since the trajectories are assumed to always be the short way around the earth,
The velocity vector at launch lies in the plane defined by the unit
vectors iL and in. If the launch speed is vL and the launch elevation is E,
then the launch velocity vector in the ECI frame can be written as
This velocity vector combined with a vector, RLiL, from the center of the 14
earth to the launch point gives a complete ECI state vector at the launch
point (RL is the radius of the earth at the launch point). By means of the
, Sl
u ?I
51i
into any desired coordinate system. The remainir, part of the problem --
S0
ET 0 (6-39)
Unfortunately, this is minimum energy in the ECI frame, rather than an earth
surface fixed frame where the minimum likely will occur under different
5?
~%
This relation indicates that the elevation angle which minimizes v2 will only
coincide with the elevation which minimizes v when the last term is zero.
The last term contains the z-component of the cross product of L and Ln which
is non-zero unless the trajectory plane (ECI) contains the North Pole.
where
Se(yLnx-xLny)
Note that the minimum occurs at E = r/4 - 0/2 for a Polar orbit ($-=0). An
0
initial guess of E = n/4 - 00/2 is made, followed by steps of 0.01 radians
until the zero of dV2 /dE is bracketed. Then a two-point linear approximation
of dv 2 /dB is used in an iterative scheme to locate the value of F. which
minimizes v 2 . This value of E is then used to compute the orbit parameters a,
e, ECL' AIM and At. The new time of flight At is compared to the previous
value and the whole procedure repeated, unless the difference is less than one
second.
Finally, the ECI state vector with the position vector, RL•L, and the
velocity vector vL [given by Eqn. (6-38)] is transformed to an earth surface
fixed XYZ frame (Radar XYZ). The final optimization of the state vector (which,
among other things, accounts for flattening of the earth and for the effects
53
6.2.2 Launch Angle Specified
corresponding launch speed (Eqn. 6-26), and launch velocity vector (Eqn. 6-38).
Convert the ECI launch state vector thus determined to an earth surface XYZ
frame at the launch point. Then compute the launch elevation angle from
Eqn. (6-43) and compare it to the desired value ED. If EF is not close enough
to ED (e.g., 0.001 radians), choose a new value of E and repeat the calcula-
Once the elevation E and launch speed vL are determined, the orbit
parameters a, e, ECL, AM and At are computed. The new time of flight is
54
compared to the previous value and the whole procedure repeated, unless the
difference is less than one second.
The final ECI state vector is then transformed to the launch point
"Radar" XYZ frame. Launch speed, azimuth and elevation are computed from the
Radar XYZ velocity vector to provide the initialization of the final trajectory
optimization discussed below in Section 6.3.
This option is very similar to tht described above for the launch angle
specified case. In fact, the initialization procedure is exactly the same,
except that the desired launch elevation is given by
ED = IERI + 0.017
where ER is the re-entry angle in radians (ER < 0). The additional 0.017
radians (1 degree) is the result of observing that typically re-entry angles
(magnitude) are smaller than the launch angles by 0.01 to 0.02 radians.
perturb the launch velocity vector in such a way that the target will impact
55
at the desired point and, in addition, will satisfy the launch angle, re-entry
r~il
interest--launch angle specified, re-entry angle specified, minimum energy-
are considered as three separate special examples.
-
k:• The impact longitude, impact latitude and re-entry angle are nonlinear
functions of the launch velocity vector due to the oblateness of the earth and
due to the gravitational harmonics of the earth's field. In this section, an
iterative technique for perturbing the launch velocity vector, such that the I
trajectory will ultimately impact at the correct latitude and longitude with
the desired re-entry angle (if specified), will be presented. I
Let v be the launch velocity vector and let y be a vector containing the K
quantities--impact longitude, impact latitude and re-entry angle. Actually, I
it is the vector, q = y-•d, which is to be minimized. In this case, Yd is the
vector of the desired values of y. The relation between q and v can be -
expressed as I
q = f(v) (6-44)
where f(v) is a nonlinear function of the vector V. The objective is to deter- '
mine v such that q = 0. A perturbation AV of the vector v results in
56
.....
..
q + q= f(v+Av) + JAy (6-45)
2 3,
3q, q q
av a a
aq2 q2 aq
S -v-- (6-46)
qo f(Vo) (6-47)
If
or,
57
where H J-.
J Now let
J, J +A (6-49)
estimate. Then
q, =f(V
0 +AV
0) = qo+J 1 AV
0
qO+JoAVo+AAvo
or
q, AAvo) (6-50)
since JoAVo = -JoHoqo "qo" The problem now is to determine the matrix A
such that Eqn. (6-50) is satisfied. If it is further required that
AAu = 0 (6-51)
U 0,
TAV (6-52)
0
qA•vT
A = (6-53)
0 00
S8
This gives a new value for ,he Jacobian, Ji = Jo+Ao, but what is actually
required is the inverse of J
= Hi + qiv 0~
O
0 0
I Hoq I -1
H1 (6-54)
(1+T)- =TI VT
permits H1 to be expressed as
0H
q16v HO
H1 =H o 01 00 (6-56)
AVooVo+AVToHoq
Thus, when initial guesses of v and Ho, are available (or computed), the
59
• • i!
Joii V I 0, kj (6-57)
[Launch speed 1 1
v= /Launch azimuth (6-58) 1
Launch elevation
is converted to earth surface fixed XYZ coordinates at the launch point. This
is followed by a conversion to an ECI state vector for use by the trajectory
integration routine. ',, trajectory is next integrated ' :ither the re-entry
altitude or the impact point (or both) and the appropriate values for the y
vector are computed, where
rRe-entry angle
y= Impact longitude (6-59)
Impact latitude
The q vector is simply the difference between the computed y vector and the
desired vector
60
6.3.2 Launch Angle Specified
For this case, the launch elevation angle is held fixed at the desired
value. Launch speed and launch azimuth in an earth surface (radar) fixed
frame are perturbed as described above. When the iterative procedure has
converged to the impact point (longitude and latitude), the launch velocity
vector still has the desired elevation angle but the launch speed and launch
azimuth are altered to achieve the proper impact point including the effects
The constraint on re-entry angle does not apply in this case, so the
velocity vector v due to re-entry angle errors. Also, the launch elevation
angle is not included in the perturbations so the third row of the H matrix
H
0 0
22 H23 (6-60)
it is readily shown that tie method of updating the H matrix (Eqn. 6-56)
zeroed.
61
6.3.3 Re-entry Angle Specified
For this case, the re-entry angle enters the canputations as a constraint.
called three times with E = Eopt, E = Eopt+AE and E = Eopt-AE, where Eopt is
62
-A
~~~~~~~~~~~~~~
/-• ;-' • ' • :, , , : , . .--.-
.'.
.::
.•
. : ;. . . '.
I
subject to certain constraints. There is the alternate possibility that a
target state vector for a particular trajectory is available for input to the
program directly. In this case, it is desirable to determine various parameters
(e.g., launch point, impact point, re-entry angle) of the trajectory for
frame as follows.
1. ECI XYZ
2. Radar XYZ
3. Radar RAE
4. Phased array XRF
S. Phased array RUV
Each of these five options requires a state vector time of validity (tag time).
Options 2 through 5 also require the radar longitude, geodetic latitude and
63
kk',63
height above the earth ellipsoid to be specified in the input stream. Options
4 and 5, in addition, require the azimuth and elevation angles of the phased
array boresight. The ECI State Vector (Option 1) is referenced to a particular
longitude at the launch time (TAL - 0); this longitude is a required input.
State Vector is then integrated backward in time to the launch point (Sub-
routine LAUNCH) and forward in time to the altitude of re-entry and to the
impact point (Subroutine IMPACT). The State Vector returned from Subroutine -
Both the LAUNCH and the IMPACT Subroutines require trajectory integration
to a specified altitude (zero for launch and impact points, 300 Kft. for
64
Consider an ECI target state vector, xI, somewhere along a trajectory.
The altitude of the target (above the earth ellipsoid) is given approximately
by (Eqn. 2-14):
R
h =r - n (t-61)
S22cos2A'
where r = 77,y+z+,
cosA = p/r and p = . (As before R is the North
polar radius and E is the eccentricity of the earth ellipsoid.) The angle A
is approximately equal to the geocentric latitude of the point on the surface
of the earth from which altitude is measured (i.e., the normal to the earth
ellipsoid passing through the target position). A first approximation to
the geodetic latitude is therefore
X= - At, 0 X i 2r (6-63)
where the second term accounts for the rotation of the earth from the reference
time of the ECI frame (usually TAL = 0) to the tag time of the state vector.
Uhe parameters X, € and h give the position of the target with respect to
the surface of the earth. It is also desirable to determine the azimuth and
elevation of the velocity vector with respect to an earth surface fixed frame
65
%. - .' ..
at this position. To accomplish this, the following three unit vectors are
defined:
i = - (6-65)
i3 3-sin
= , + P- +coso z
z:- (6-66)
to Zi, parallel to the equatorial plane and points to the east; '3 is normal to j
both -i and i2 and points to the north. That is; iL' z2 and '3 correspond to
the Z, X and Y axes of a radar XYZ frame. The components of the velocity
ýF I_+weYI
"~F I ~eXI :
iF ±I (6-67)
The target speed in the ECF frame (or radar XYZ frame) is
•F 'F F'
=v0/7kX;2;+_:,_2_+,2 (6-68)
66
W.WWW"
The components of the velocity vector in the earth surface fixed frame are
obtained by taking the following dot products
"YIX:F+XIYF
VX k
S-V F
2 _____
P (6-70)
Vy VV 3 F I I+(6-71 +COSý
+-sin ±1 (6-71)
where vF = YxF+iyIF+z.F Note that the effects of the rotation of the earth
cancel in the expressions for vz and vy. The elevation angle of the velocity
vector (which may be the launch angle, impact angle or re-entry angle) is given
by
Ev sin-t( (6-72)
Av tan'(•) (6-73)
formed in Subroutine ANGLES. Subroutines LAUNCII and IMPACT both call Subrou-
tine ANGLES after the ECI state vector has been integrated forward or backward
to the relevant target altitude. Various parameters of launch (X, ,EvAv),
67
- j:
re-entry (x,0 ,E,)and impact ( y are saved in the target information
files.
Section 2 of the ARIES Test Report for a particular run of the Program
(see Reference 1, Appendix II) contains trajectory information for each target.
In particular, the following information is available in the Test Report:
4. Re-entry parameters
S. Impact parameters
i.
68
-~~ w
At present the only "tracking" alaorithm built into ARIES is the Maximum-
k Likelihood Estimator (MLE). The MLE does not provide estimates of a target's
position after each radar measurement, as do-recursive-type tracking schemes
such as the cz-ý tracker or a Kalman filter. Instead, the MLE provides ARIES
only with an endpoint state vector estimate, valid at the end of the track
interval(s) of interest. Duiring the "track" interval itself, the MLE Subrou-
tine \tkXUIK simply stores up the (noisy, biased) radar position measurements,
along with the variances ol each measurement.
Once all such measurements have been accumulated, the MLE attempts to
find that endpoint state vector estimate which minimizes the "residual errors"s
between the physical trajectory (uniquely defined by the endpoint state vector)
and the set of radar measurements. The error used in this minimization is
formed as the sum of the weighted, squared differences between the physical
trajectory and the set of radar measurements. Each error residual is weighted
according to the standard deviation of the corresponding position measurement.
F". Since each error residual is weighted according to the variance of the
corresponding position measurement, the final state vector estimate corresponds
to a minimum variance estimate. This minimum variance estimate is also a
maximum likelihood estimate of the trajectory, when the measurement errors
have a zero-mean, Gaussian distribution. If there are no bias eriors in the
data (such as those due to inadequate calibration or those due to imperfect
correction of tropospheric or ionospheric refraction effects), the trajectory
69
estimation procedure described below corresponds to Maximum Likelihood
dependent on whether the radar is a dish type or a phased array type. The
reference ECI state vector x. is integrated successively to the times of the
radar measurements. At each of these times, the state vector xs(ti) is
transformed to the appropriate radar frame (RAE or RUV) to form the set of
"estimated" position vectors, 4r(ti). It is the weighted, squared differences
between rm(ti) and ts(ti) which are minimized in the trajectory fitting
process.
It has been shoun in Section 5.2 that the transformation from an ECI state
70
'I'-I
where the transformation matrix G(t) accounts for the rotation of the earth in
time t (see Eqn. 5-1). The transformation matrix A(¢,X) and the translation
vector b(o) are those used to transform from an ECF frame to a radar XYZ frame
(see Eqns. 5-14 and 5.15). For a phased array radar, the additional trans-
formation from radar XYZ to phased array XYZ (see Eqn. 5-26) must be included
so that the vectors xsr are defined in the proper coordinate system for the
radars.
Also, we have denoted
s [y] (7-2)
u JECI
*!i[ and
Xsr (7-3)
radar
71
case, we use Eqn. (5-21) in Eqn. (5-20) to determine the relation between
= ¢ 6sr (7-5)
-q.
For a phased array radar (RUV coordinates) the matrix C is
RU RV RW
R -N U(7-7) --
where W =l -U2-V. Note tha. the matrix C varies as the target moves along. . . . -
its trajectory; for simplicity, this time dependence has not been explicitly
included in the above relations.
Combining Equations (7-4) and (7-5) one can then relate perturbations in
vector:
Back to the NLE problem: Given the reference state vector x., the MLE Sub-
routine MAXLIK first uses the equations of motion (Section 4.3) to obtain x
si
72
S..i
..................
and then Equation (7-1) to obtain Xsr at each measurement time. Next the xsr
state vector is transformed into %s (the set of target position estimates
based upon the ECI reference vector xs), using the appropriate transformations
(given in Section 5.3 for an RAE radar, and in Sections 5.4 and 5.5 for an RUV
radar).
In deriving the MLE estimator, we assume that perturbations in the refer-
fts(ti, 6
x 6
X)
U, F(ti)[6xs + (ti-ts)6ks] (7-10)
At this point we have a set of radar measurement data tm(ti), a set of "esti-
mated" target positions ts(tj) based on the reference ECI state vector xs(ts)
73
follows:
N
Q = 1{in~ti) - S(ti) -
6
&L5(ti, 6 X5 ,6ks)T
where W is a weighting matrix taking into account the quality of the radar's
is02(ti)
10 0
i
.
(O(ti) 0 1C(7-13)
oY2 (t.)
0 01
74
6&(i) r(i) (7-14)
in s
and
i"=1
If the components of 6xs are denoted by 6xs, 6ys, 6zs, then to minimize Q one
must set the partial derivatives of Q with respect to 6xs, 6 ys, and 6z, to
zero: .
T
sN]T( 6f
(i',Xs'6(s)
2a=1
.[(i,6x 6)
S
W(i) 0 (7-17)
75
Similarly, if the components of 6kare denoted by As 6ýs 61 and if the
partial derivatives of Qwith respect to 6±.t6s-9and 62 re set to zero, then
the following matrix equation results:
S. sv
N
~[6(i,t6xs6k)]rW(i)F(i)(t -t) (0,0,0) (7-19)
The ML Estimator simply has to solve these linear equations for ýxandU
then the end-of-track state vector estimate can be updated: th
s@
Xse X~~l +~ (7-20)
Cs,new ~s,old + 7-1
N
i=1
The matrix equations for the N Estimator, Eqns. (7-18) and (7-19), can bes
written in a more compact form as follows. First take the transpose of these
equations adsbttethexprsinfr6itsk ro Eq. (71)
Then define the following vectors and matrices
d EF (i)W(i)&'i(i) (7-22)
N
i~76
N
M2 (ti-.S F (i)W(i)F(i) (7-25)
N
M3 ; i (ti-ts)2FT(i)W(i)F(i) (-7-26)
6X + M2 6k (7-27)
dl M~I M2 •s i
- . ----
.. ..... (7 -2 9 )
2 Mj2, M3 6s ,;
A (7-30)
M2 ',
,•
and [!]!
nB
(7-31)
77
,a r . -*;aL'z.:% r-. .-
These are not to be confused with the A and b which were defined above in
Equation (7-1).
Note that the ML Estimator is an iterative procedure: An initial,
reference state vector xs is determined (equal to the true target state vector
in ARIES), then perturbations 6xs and U are determined by the above pro-
cedure. The new value of x from Eqns. (7-20) and (7-21) is then used to
repeat the process. Since ARIES starts with an excellent value for x., the
convergence of this process is very rapid; in fact, only two iterations are
used.
The residual errors between the target-position measurement for the radar I
tn(i) and the target-position prediction kst(i) at each time ti (where rs(i) is.
computed from xs, the final-iteration value of the MLE end-of-track state
N".1AX
Mean square range residuals S [6r 1 (i)] 2 (7-32)
where NPIMAX is the number of measurements made during the Monte Carlo run, '3
and where 6r1 is the first component of the residual error vector 6M defined j
in Equation (7-14). Similar mean square residuals are computed for the other
two components of 6r. These mean square residuals are computed for each Monte
Carlo run, and then averaged over all runs. The resultant mean-square-resid-
uals are printed out in Sect Lon 11 of the ARIES Test Report.
78
On the mth Monte Carlo simulation run the ARIES Test Program computes a
"correlation" matrix estimate via
T '2
Cm emn (8-2)
where em is the error defined above in Equation (8-1) on the mth Monte Carlo
run. These correlation matrix estimates are then averaged over all runs:
A A MC
C CA (8-3)
m=l
i.e., if the quadratic form xTPx is non-negative for all real values of the
variables xi, and is zero only if each of the xi variables is zero. For the
P matrix, this statement translates to the condition
MC MC
m=l Te l (xe
m)2 1__m=I em+X em)2 > 0
m ~ m 1 im 2 2M1 3 3MW
80
which is obviously non-negative. If MC=l or MC-2, it can be readily shown
that there are an infinite number of non-zero values of the xi's for which the
above quadratic form is zero. This violates the conditions for P to be posi-
form equal to zero with non-zero values for the xi's. Exceptions are possible
if the error vectors are linearly related; this is unlikely irn the case at
hand, particularly for large values of MC. The reason why the position
correlation matrix is singular for one or two Monte Carlo runs can be simply
explained as follows: a volume in space requires three non-parallel. vectors
for its definition (one error vector defines a point, two error vectors define
a plane). Consequently, for the ARIES Program to generate meaningful error
If the mean error e (obtained by averaging over all Monte Carlo runs) is
A T
Vm = (% -) (em.-,)T (8-4)
n1 (8-5)
m=
81
In practice, what is actually done by ARIES is to first compute C via Equation
(8-3), and next compute
I - 1 M
C
e i(8-6)
e.
V eJ (8-7)
I
As with the correlation matrix C this covariance matrix D is symmetric and
thus only the lower half appears in Section 12 of the ARIES Test Report.
The "position" covariance matrix P (the upper left-hand 3 x 3 sub-matrix
of D) is singular for MC < 4; the process of subtracting the means of the
82
To eliminate the correlations between the error vector components requires
- -- -- (8-8)
Then we want to determine a new coordinate system such that the ECI position
ilk error vector eT = (e, ,e2,e3) is related to the position error vector ( T
(e',e',e3) in the new coordinate system via the transformation matrix Q:
e= Qe (8-9)
where Q is the normalized modal matrix of P with the property that Q _OT.
The new "position" correlation matrix P' is then easily shouwn to be related to
P via
pi 2Tp (8-10)
P-XII 0(8
83
where X is an eigenvalue. For this case there are three such eigenvalues;
they are the roots of the cubic equation
P11" P 12 p13
P2 1 p 2 2 -X P2 3 = -X 3 -pX2-qX-r = 0 (8-12)
"where
P = "-(P +P2 2 +P3 3 ) (8-13)
q2 - p2 (8-14)
2 2 2 .2P P (8-1S)
r=P 1 3 P 2 2+P 2P 3 3+P 2 3P 1 1 12P 2 3 13 1 22P33
p
a = q - p2 /3 (8-16)
p3 (8-17)
b7-
+
Cos-[ 2 a (8-18)
1 = 2 ~Cos P- (8-19)
84
[
= [2 cs0j(8-20)
rcsT 3 (8-21) i
X3 [2
Since the matrix P is positive definite, the eigenvalues (X1,~, 3 are all
real and positive. It then follows that P' has the desired diagonal property:
X
0 0
P2 (8-22)
rendered uncorrelated errors, the eigenvalues A',,X2 and X3 alone are insuffi-
cient to specify the uncertainty volume of the target state vector at hand
over. The orientation of the ellipsoid axes must also be taken into account.
In order to avoid this complexity, it is often more useful to think of an
"uncertainty sphere" = namely a sphere of radiw ,, centered at the
85
A
(exti'apolated) handover point, with R chosen such that the error vector lies
assures us that the error components in the new coordinate frame (see preceding
section) are uncorrelated. If further we asstume these errors to have "normal"
or "Gaussian" distributions,'then the probability of the error vector e'
[defined in Equation (8-9)] is given by
•i:1 x2 y
Tq
2. Z2
TX_ 3
x - (8-25)
86
Y
y (8-26)
z (8-27)
then we obtain
2 '2 '
C 1,7
-
8x -- 2
p(Ie' < R) 8
11.5 dxe dye-Y2 e-z 2 dz (8-28)
C =I~rX
a(8-29)
Cy =R/j-21 (8-30)
S--R/V73(8-31)
Note that the integration is now over an ellipsoid rather than a sphere.
then we may replace the two inner integrals of Equation (8-28) withl
87
- .- ~.,,
f fI ff
do re -r2 drr2(=--d (8-34)
0 0 0
This integration gives the area of an ellipse in the y-z plane. The semi-axes
•/i- -Cysi2
2
c
Use of the double angle formula, 2sin2 o = l-cos2o, in Eqn. (8-35) results in
r2(0) = (1 Y (8-36)
CY 2cos20
or,
r2 o- ? +~ os2-0 (8-37)
12
where
k( 2 2 (8-38)
2(18
88
and
C2
2k2 (8-39)
2 xX2 )C2
_r(O) 2 T]
J
df re-r dr i -f c- dJ (8-40)
0 0
When Equation (8-40) is used to replace the inner two integrals of (8-28),
we obtain
Cx"r
p(Ie'l < R) -- L dx ex T -fexp dk+kcsd (8-41)
0 0
The double integral in Ecin. (8-41) cannot be reduced any further; there-
tation of the probability that the error vector lies within the sphere. flow-
89
References
[11 M. E. Austin, "The ARIES Program: A General Overview and Users' Guide,"
Technical Note 1975-14, Lincoln Laboratory, M.I.T. (to be published).
[2] J. L. Mannos and J. L. Katz, "The ARIES Program: Analysis and Generation
of Simulated Radar Measurements," Technical Note 1975-16, Lincoln
Laboratory, M.I.T. (9 July 1975)
[3] G. B. Morse and W. E. Courtney, "Applications and Results," private
communicat ion.
[41 F. S. Acton, Numerical Methods that Work, (Harper • Row, New York, 1970).
gZIA
9.
90