VOL. 7, NO. 3, MAY-JUNE 1984 J.

GUIDANCE 315

Target Acceleration Modeling for Tactical Missile Guidance

Paul L. Vergez* and Randall K. Liefert
U.S. Air Force Armament Laboratory, Eglin Air Force Base, Florida

For future short-range air-to-air missile concepts, it has been demonstrated that linear optimal guidance laws
provide significant performance improvements over proportional navigation guidance laws. This paper ad-
dresses the critical problem of estimating the missile-to-target position, velocity, and acceleration (required by
the linear optimal guidance law) when only passive (angle only) seeker information is available onboard a highly
maneuverable bank-to-turn missile concept. More specifically, the problem is how to model the target ac-
celeration to achieve improved missile performance. Four target acceleration models coupled with an extended
Kalman filter are presented and evaluated on a six-degree-of-freedom missile simulation to determine their
estimation effectiveness and their influence on missile guidance.
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

I. Introduction problems can be solved potentially through the use of ad-

PPLICATIONS of optimal control theory to the tactical vanced software techniques, which leads to the second
A missile guidance problem have drawn much attention in
recent years. The most commonly used theory is linear
solution.
Recent advances in estimation theory, coupled with the
quadratic Gaussian (LQG) theory because it is based on a explosion in microprocessor technology, make it feasible to
linear system model and provides a closed-form solution. develop complex yet realizable estimation algorithms. This
Further, it has been demonstrated that for the short-range new capability could solve the information problem in tactical
tactical missile, the LQG guidance law provides significant missile systems without increasing the number and types of
performance improvements over the more commonly used sensors currently in use in tactical weapons. This is done in an
classical proportional navigation guidance laws. Ad- attempt to minimize hardware requirements for mechaniza-
ditionally, in terms of future missile performance tion of advanced high-performance guidance techniques.
requirements and future target characteristics, proportional The most wisely accepted estimation theory is use today is
navigation will be inadequate.l Kalman filtering, which was developed in the early 1960's.
A critical issue that affects performance of the LQG Much work has been done in recent years applying this theory
guidance law is the fact that it is a function of time-to-go (time to the tactical missile guidance problem, in particular, the use
remaining before intercept) and the missile-to-target position, of extended Kalman filters (EKF's).2'5 Still, a critical design
velocity, and acceleration. Time-to-go is typically estimated factor in an EKF is the modeling of the target acceleration. It
from the later information noted in the preceding sentence. is impossible to model the target acceleration for practical
Although not the subject of this paper, much work has been application accurately because one cannot know exactly what
accomplished toward improving time-to-go estimates.2 a target will do, particularly a flying target.
Everyone will agree, however, that the estimates of time-to-go The purpose of this paper is to present four target ac-
and the performance of LQG guidance law are most effective celeration models for an EKF, evaluate their estimation ef-
when the required position, velocity, and acceleration in- fectiveness, and determine their influence on tactical missile
formation is accurate. guidance when coupled with an LQG guidance law. The four
target acceleration models evaluated are 1) first-order Markov
One of the fundamental problems that has limited the process, 2) second-order Markov process, 3) constant target
practicality of the LQG guidance law is the difficulty in acceleration, and 4) zero target acceleration.
obtaining the accurate state information required. Most
present-day missiles can obtain a measure of the missile's
acceleration through onboard accelerometers; however, this
information is not presently used for guidance purposes. In II. Extended Kalman Filter
addition, passive seekers are used to provide a measure of It is important to describe first how an extended Kalman
line-of-sight angle and rate. Since proportional navigation filter works, and then present the differences introduced by
(PN) is a function of line-of-sight rate only, a passive seeker is the four target acceleration models. Figure 1 shows the flow
more than adequate. A low-pass filter is typically used to diagram for an extended Kalman filter. Subsections 11. A
process the noise from the line-of-sight rate for PN. To make through II.E show the derivation of the algorithms of the
use of the LQG guidance law more information is needed. EKF for the missile/target intercept problem.
There are two solutions to obtain the necessary information
for the LQG guidance law: one is to increase the accuracy and
number of onboard sensors. Cost, size, and weight factors A. State and Measurement Models
make this solution unattractive. Further, the target ac-
celeration simply cannot be measured. These hardware State model:

X ( t ) =FX(t) +Bu(t) ~N(0,Q(t) )

Submitted June 18, 1981; revision received Jan. 3, 1983. This paper X(0)~N(X0tP0) (1)
is declared a work of the U.S. Government and therefore is in the
public domain.
*Captain, USAF, Program Director, Guidance and Control Measurement model:
Engineer.
tCaptain, USAF, Guidance and Control Engineer. Zk=g(X(tk))+vk, vk~N(0,Rk) k= 1,2,3,... (2)
 316 P.L. VERGEZ AND R.K. LIEFER J. GUIDANCE

where (10b)
X(t)=EKF states AT(t)=\TAT(t)+wT(t) (lOc)
F= state matrix
Bu (t) = state forcing function In state space format, the system model is:
w (t) = accuracy of states, white Gaussian noise (WGN)
with zero mean 0 J
0
2(0 = standard deviation of WGN of states
Zk = EFK measurements AF(0 0 0 /
vk — accuracy of measurements, WGN with zero mean
Rk = standard deviation of WGN of measurements 0 0 -T1
X0 = initial value of states 9x1 9x9 9x1
P0 = initial error covariance of states

B. State Transition Matrix

0 0 0 0 0
The state transition matrix, along with the forcing function, 0 J
0 AM(t) + H> M (0
is used to propagate the filter states and error covariances.
The state transition matrix, $ (t k > tk_ / ) , is 0 0 0 0 WT(t)
9x9 9x1 s xl
*(tk,tk_,)=£-'(SI-F) (3)
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

( 11)
C. State and Covariance Propagation where
The propagation equations of the state estimates (X) , and = three components of relative position, ft (inertial
error covariances, (P),are: coordinate)
AK(0=three components of relative velocity, ft/s
(inertial coordinate)
(4)
AT(t)= three components of target acceleration, ft/s 2
(inertial coordinate)
P(tk) = 3>(tk,tk_1)P(tk_])$>T(tk,tk_1) AM(t)= three components of missile acceleration, ft/s 2
(inertial coordinate)
\T = target acceleration response time coefficient, s ~ l
$(tk,T)Q(T)3>T(tk,T)dT (5)

For the system model in Eq. (1 1), the state transition matrix
[ derived from Eq. (3)] is:
where X ( t k ) is the propagated state estimate and P(tk) the
propagated error covariance.
1 ! Af/ ! $]3I
D. Kalman Gain Matrix
The Kalman gain matrix, K, is used to minimize the 0 \ I ! **/
diagonal elements of the error covariance matrix. It is ob-

111,7"
0
tained from the following equation. 9x9
(12)

K(tk)=P(tk)HT(X(tk))(H(X(tk))P(tk)HT(X(tk)) where
!
+R(tk)]~ (6) (13)
where
dg(X(tk))
H(X(tk)) =
dX(tk)
(7) \T (14)

E. State and Covariance Updates

IV. System Model with Second-Order Markov Process
The final step is to update the filter states and error
For this model, the three components of relative position
covariances in order to drive the errors between the actual remain the same as in Sec. Ill; however, the target ac-
measurements and the filter's estimate of the measurements to celeration as a second-order Markov process is a function of
zero. These updates are obtained from the following target velocity. The nine filter states are now the three
equations. components of relative position, target velocity, and target
acceleration, all with respect to inertial coordinates. The
\=X(tk)+K(tk)[Z(tk)-g(X(tk))} (8) differential state equations now are:
*) = [/-A-(f t )^(^(/ t ))]P(^) (9)
(15a)
III. System Model with First-Order Markov Process
The EKF is set up as a nine-state filter with the system VT(t)=AT(t) (15b)
model in Cartesian coordinates. The nine states are the three
components of relative position, relative velocity, and target
AT(t) = -u2(t)VT(t)+wT(t) (15c)
acceleration, all with respect to inertial coordinates. The
missile's acceleration is measured onboard and is used as a
forcing function of the states. when V M ( t ) is the missile's velocity with respect to inertial
The differential state equations are: coordinates and is obtained by integrating the measured
missile acceleration, and V T ( t ) is the target's velocity with
(10a) respect to inertial coordinates. The system model for the EKF
 MAY-JUNE 1984 TARGET ACCELERATION MODELING FOR MISSILE GUIDANCE 317

now becomes: V. System Model with Constant Target Acceleration

For this model the nine filter states are the three com-
AS(t) 0 7 0 ' A£(0 ponents of relative position, relative velocity, and target
— acceleration. The differential state equations now are
&V(t) 0 0 7

_ /MO. _ 0 -co 2 U)7 0 /• r ( * ) (21a)

9X1 9X9
'9X1 AY(t)=AT(t)-AM(t)+wM(t) (21b)
-7 0 0 ^(0 ' , wj!f(0 " /ir(0 =H> r (0 (21c)
f 0 0 0 0 + ^ The sytem model for the EKF now becomes:
0 0 0 0 - ™T(t) . Asm i 0 7 0 1 ASOf)
9x9 9x1 9x

(16) 0 0 7 AK(/)
vhere _ AT(t) 0 0 0 _ ^7-(0
H>M<0 9x1 9x9 9x1
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

0 ~N(Q,Q'(t))
0 0 0 0 o :
- w r (0 _ + 0 7 0 -AM(t) -t- ^(0
vith
A^QM | 0 0 0 0 0 0 [ H- r (0 ]
1
yxy yx i 9x1
Q'(t) = 0 ! 0
1
0
( 17 >

For the system model in Eq. (22) the state transition matrix
[derived from Eq. (3)] is:
The vector w(0 is the three components of target turning rate
and is obtained by / AJ27
A/7
2
(18) *(/*.'*-/) = (23)
0 7 Atl

VT(t) and AT(t) are the estimated states of the filter. The 0 0 I
state transition matrix for the system model, Eq. (16), is

VI. System Model with Zero Target Acceleration

For this model the EKF is reduced to six states; the three
sinco(^)A^ 7-cosco(/ A: )A^ components of relative position and relative velocity. The
differential state equations are:
«(^)
lS(/)=AK(0
Z (24a)
sinco(/^ ) A^
0 cosco(^)A/7 (19) AKU) = -/MO+H^(0 (24b)
w(^*).
0 -«(^)sin«(^)A^/ cos«(/*)A// The system model for the EKF becomes:

It is possible that any of the three components of

would approach zero at any given instant in time. To com-
9x9

--^--- =
A5(0 1 [ 0 71
-— — —
L 0 0 J6X6
1 A5(0 1
-----
L AK(0 J 6 x l
putationally handle this it is necessary to take the limit of
$ ( t k , t k _ j ) as co^0; therefore,
2
A/ /
[..«..«..]
L o / J 6X6
I ° - --1
L -AM(t) J 6 X | ++ [- -°--- 1
L w M ( t ) __ 6x1
7 Atl (25)
2
------- For the system model in Eq. (25), the state transition matrix
\im(3>(tk,tK_1) = 0 7 A/7 (20) is:
w—0

0 0 7 A/7
9x9 *('*>'*-;) = (26)
6x6

Note that in this model, the state transition matrix is time

varying as opposed to the sampling frequency-dependent state VII. Measurement Model
transition matrix of the first filter model. For the purposes of The same measurement model was used for all four filter
this study, co, (0 <0.01 is assumed to be zero, i=x,y,z. system models. The filter has two measurements: azimuth
 318 P.L. VERGEZ AND R.K. LIEFER J. GUIDANCE

Table 1 Filter parameters

Parameters Value

0.3
, ft 2 /s 4 150
40,000
/, s 0.01

Table 2 Initial conditions

State Mean Variance

SR True value 10ft 2

V* True value 10ft 2 /s 2
AT Qft/s 2 20,000 ft 2 /s 4
Fig. 1 Angular measurements related to filter states.
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

(0) and elevation (a) angle in inertia! coordinates. The With a variance of 10 the filter is informed that the initial
relationships between these angles and the filter's states is guess of relative position and velocity is within ±3.16 ft (or
illustrated in Fig. 1. ft/s) of its true value. With a variance of 20,000 the filter is
The equations relating the angles to the filter states are informed that the initial guess of target acceleration is
within ± 4Ag of its true value.
(27)
IX. Guidance Law
(28)
The LQG guidance law used in the missile simulation is
where Sx, SY, and Sz are the three components of relative expressed in the following equations:
position in inertial coordinates. In state space format the
measurement model is VRx/tgo +KTATx (34a)

(34b)

(34c)
2x1 2x1
(29) The quantities appearing in the guidance law equations are
described below.
VIII. Noise Statistics and Filter Parameters SRx,STy,SRz = three components of relative position
The same noise statistics and filter parameters were used for vector SR referenced to the missile body,
all four filter system models. ft
VRX>VRY> VRZ = three components of relative velocity
A. Noise Statistics vector VR referenced to the missile body,
ft/s
0 0 ! 0 - A TX ,A Ty ,A TZ = three components of target acceleration
vector A T referenced to the missile body,
Q(t) = 0 QM! !1 0 (30)
ft/s 2
0 0 1 Qr/ _ AMX,AMY,AMZ =three components of missile acceleration
9x9 command vector AM referenced to the
missile body, ft/s 2
0.25/R2 + 5.625 xlO" 7 KT is the target acceleration gain, where
(31)
2 7
0.25/.R +5.625 x l O '
KT= (35)
B. Parameters
QM is the process noise covariance for missile acceleration. where T is the target acceleration response time coefficient,
With an rms error of 0.4g for an accelerometer, and /go is the time to go, s (Ref. 2).
= [0.4(32.174 ft/s 2 )]' (32)
=2SR/-V *x (36)
QT is the process noise covariance for target acceleration.
With an estimated variance on target acceleration of 20,000, where Axx is the difference between missile acceleration
command and KT times target acceleration in the axial
Qr=2P (33) direction, ft/s 2
Filter parameters are summarized in Table 1.
An additional input to the EKF is the initial conditions of Axx=AMx-K'TATx (37)
the filter states and their variances. These values are listed in
Table2(Ref.3). and K'T is KT evaluated at the previous time interval.
 MAY-JUNE 1984 TARGET ACCELERATION MODELING FQR MISSILE GUIDANCE 319

K'T=KT \ (t-At)
(t_ (38)

The time-to-go algorithm has the advantage of explicitly

accounting for the missile's axial acceleration, which has been
ignored in the past; thus resulting in more optimal lateral and
normal acceleration commands.

X. Performance Evaluation
An evaluation of the EKF with the four system models was
conducted in two steps first, for a given missile/target
engagement, graph of the percent errors in the filters
estimates of the magnitude of relative position, relative
velocity, and target acceleration vs time were generated for o 6.
the four system models. For the model with the second-order TIME (SEC)
Markov process, a graph of the percent errors in the estimates
of target velocity replaced relative velocity. Graphs of the Fig. 2 Relative position error for first-order Markov filter.
errors in the filters estimate of azmimuth and elevation angles
vs time were also generated for the four system models. This
evaluation provides a measure of how well the filters estimate
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

I.OT
the states; however, a more effective evaluation is to deter-
mine how well the missile performs the EKF coupled with the
LQG guidance law.
The second step of the performance evaluation involves the
determination of miss distance (point of closest contact 1.0 v
2.0 3.0 4:0 5lO 7.0
between the missile and target) for a specified number of
initial launch conditions. For both evaluation steps a detailed
six-degree-of-freedom (6-DOF) simulation of a generic bank-
to-turn short-range air-to-air missile was used. The target
used in the simulation incorporated a "smart" target
,-1.0
A
algorithm incorporating a 9g out-of-plane evasive maneuver.
The simulation contains detailed nonlinear math models of
the major missile subsystems including the seeker, autopilot,
and propulsion systems; realistic noise models of the onboard
sensors and seeker models; detailed aerodynamic models of
missile airframe characteristics; and the models that describe
the missile's equations of motion. This missile/target com-
bination was selected because it represents desired per- -3.0--
formance capabilities for the future guided weapons.

A. Estimation Error Graphs Fig. 3 Relative velocity error for first-order Markov filter.
For each of the EKF system models, the percent errors in
the filter's estimates of the position, velocity, and acceleration
magnitudes vs time were generated for the case of 0 deg off
boresight (the off boresight angle defines the angle between
the initial line-of-sight vector and the initial missile velocity
vector, therefore, 0 deg off boresight means that the missile
was launched pointing directly at the target) and 90 deg aspect
angle (the angle between the initial line-of-sight vector and the
target's velocity vector at launch). The initial launch range
was 13,000 ft. The missile and target were coaltitude at launch
(10,000 ft) and were cospeed at launch (0.9 Mach). The target
performs its evasive out-of-plane maneuver when the range
becomes less than 6000 ft. With 1 s left in the engagement, the
target performs a dive, straight down. This launch condition
was selected because it represents a challenging shot for the
inertially referenced guidance and estimation algorithms.
Plots of the filters' estimation errors vs time provide Fig. 4 Target acceleration error for first-order Markov filter.
graphic evidence of the ability or inability of the filter to
estimate position, velocity, and acceleration during a single
flyout in the presence of process and measurement noise. the components of position, velocity, and acceleration but is
Figures 2-4 are a representative set of such plots for filter successful at estimating the overall magnitudes. Sample plots
employing a first-order Markov process. of azimuth and elevation errors again for the first-order
It should be noted that only the errors in magnitude coupled Markov filter are presented in Figs. 5 and 6.
with errors in angle provide an effective evaluation of the
filters' estimation capabilities. Large magnitude errors
coupled with small angle errors indicate the filter maintains B. Missile Performance
the proper ratios of the three components of position, To evaluate the missile's performance with the EKF in-
velocity, and target acceleration, but does a poor job of corporating each of the four system models and the LQG
estimating the overall magnitudes. On the other hand, small guidance law defined in Sec. IV, numerous missile vs target
magnitude errors occurring simultaneously with large angle flyouts were generated for a set of intiial launch conditions
errors indicate the filter does not maintain the proper ratios of listed in Table 3.
 320 P.L. VERGEZ AND R.K. LIEFER J. GUIDANCE

Table 3 Initial launch conditions The selection of these initial launch conditions was done
^ such that the evaluation would provide a good sampling of the
Launch Launch of f- weapon's performance, The measurements were corrupted
Launch Launch aspect angle, boresight angle
condition range, ft deg deg with the noise models defined in Sec. VIII. Since the noise
models are random processes, it becomes necessary to fly
1 8,000 0 0 repeated runs with different noise sequences for the same
2 13,000 90 0 launch conditions (Monte Carlo analysis). Through some
3 26,000 180 0
4 7,500 0 40 preliminary simulation analysis it was determined that 10 runs
5 11,000 90 40 provide an accuracy in miss distance within 0.1 ft, which is
6 24,000 180 40 adequate for the purposes of this study. The results presented
7 1,000 45 0 in this section are based on the use of 10 Monte Carlo runs
8 3,000 135 0
9 1,000 45 40 and are illustrated in Table 4. The miss distance is calculated
10 2,500 135 40 as the mean miss distance over the 10 Monte Carlo runs. Filter
1 incorporates the first-order Markov process, filter 2 in-
corporates the second-order Markov process, filter 3 in-
corporates the constant target acceleration model, and filter 4
Table 4 Average miss distance for each engagement incorporates the zero target acceleration model.
Miss distance, ft
XL Results
Launch The results of the estimation error vs time plots for each of
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

condition Filter 1 Filter 2 Filter 3 Filter 4 the candidate filters indicate that they all do a poor job of
418
estimating the states with only passive seeker information in
1 5.73 4.74 4.63
2 7.00 5.83 6.03 192.4 highly dynamic environment. The errors resulting from the
3 4.81 7.46 6.10 256.8 first-order Markov filter are very similar to those from the
4 13.09 9.73 9.90 742.8 constant acceleration filter. Both produce small relative
5 7.85 6.88 11.46 270.1 position and velocity errors until the target begins
6 3.00 10.50 11.45 296.0 maneuvering. Velocity errors begin to grow following the first
7 3.55 17.40 3.61 16.3 9g target maneuver, are corrected in approximately 2 s then
8 11.60 10.25 10.10 8.4 grow agajn, in the opposite direction when the target per-
9 2.01 13.54 2.10 59.3 forms its final 9g dive. The velocity error is again largely
0 4.61 14.49 5.21 7.5 corrected before the end of the engagement. Position errors
Average
miss
for both filters grow as the target begins to maneuver but are
distance 6.33 10.08 7.06 226.8 corrected before the engagement ends. With both the first-
order Markov and constant acceleration filters, the target
acceleration estimates are oscillatory and overestimated when
the target is not maneuvering. When the target beings its
maneuver, target acceleration is initially underestimated. The
filters recover and soon overestimate target acceleration
again. The oscillations continue to grow and are quite large
near the end of the engagement.
1.0 A 2.0 A 3,0 4.01 5.0 6.
The state estimation errors produced by the second-order
Markov filter are quite different from either of two filters
discussed above. Estimates of realtive position and velocity
begin diverging immediately.
The position error peaks at nearly 4000 ft and is entirely
corrected during the last 2 s of the engagement. Velocity
estimate error continues to diverge, however, throughout the
flight. Target acceleration errors are more oscillatory than
those from the Markov filters. It is important to note that for
both the two Markov filters and the constant acceleration
filters, the errors in azimuth and elevation angles are
relatively small until the last 0.5 s of the engagement at which
Fig. 5 Azimuth angle error for first-order Markov filter. point they diverge rapidly.
The velocity and position estimation errors from the zero
acceleration filter are small until the target begins to
maneuver. Even then, these errors never grow as large as for
i.o-i the other filters and they are corrected by the filter before the
engagement ends. However, the errors in azimuth and
1.0 2.0 3.0 4.0 5.0 y| 6.0
elevation angles resulting from this filter become large as the
target begins to maneuver and, as the final target dive is
' W / \rj V
executed, the angle errors diverge rapidly. This results in the
large misses listed in Table 4.
TIME (SEC) These results clearly indicate a deficiency in the EKF's
I ability to estimate relative position, velocity, and target ac-
ui
celeration with passive seeker information. However, the
"]-° filter's true effectiveness is measured in terms of missile
performance using the LQG guidance law coupled with the
EKF. The results in Table 4 indicate that the missile performs
well with all the filters except the one using the zero ac-
-2.0- • celeration model. The differences in missile performance with
the first-order Markov, second-order Markov, and constant
Fig. 6 Elevation angle error for first-order Markov filter. acceleration filters are not substantial, except that missile with
 MAY-JUNE 1984 TARGET ACCELERATION MODELING FOR MISSILE GUIDANCE 321

the second-order Markov filter does not perform as well as the "the" answer to the tactical missile guidance problem.
other two for close-in shots. The zero acceleration filter Estimating position, velocity and acceleration from angle-
results in very poor performance except under near head-on, only information is a very difficult problem. While Kalman
short launch range conditions. filtering offers the most effective near term solution, the
accuracy of the state estimates obtained to date leave much
XII. Conclusions room for improvement. Even so, previous studies have shown
The relative merits of the acceleration models are based not that a linear quadratic guidance law coupled with an extended
only on the performance qf the resulting filter but also upon Kalman filter working an angle only information provides a
their complexity and resulting computional burden. While the signiciant missile performance improvements over con-
performance of the first-order Markov and constant ac- ventional propertional navigation in highly dynamic
celeration filters are similar, the first-order Markov model engagements.
results in more complex state transition matrix. This becomes
a factor for weapon systems require variable sampling References
rates because the state transition matrix must then be
^astrick, N.L., Seltzer, S.M., and Warren, M.E., "Guidance
calculated online. The second-order Markov filter performs Laws for Short-Range Tactical Missiles," Journal of Guidance and
slightly worse than either of the above two and imposes a Control, Vol. 20, March-April 1981, pp. 98-108.
much more severe computational burden for two reasons. 2
McClendon, J.R., and Vergez, P.L., "Applications of Modern
First, a time varying state transition matrix results which must Control and Estimation Theory to the Guidance of Tactical Air-to-Air
be computed at each sample period. Second, additional Missiles," Proceedings of the Second Meeting of the the Coordinating
complexity is required to prevent singularity when target Group on Modern Control Theory, Aberdeen Proving Grounds, Md.,
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

turning rate approaches zero. The inability of the zero ac- Dec. 1980.
3
celeration model to maintain proper position and velocity Sammons, J.M., Balakrishnan, S., Speyer, J.L., and Hull, D.G.,
angles and the resulting poor missile performance make it "Development and Comparison of Optimal Filters," AFATL-TR-79-
87, Oct. 1979.
unacceptable for use against a highly maneuver able target. 4
Anderson, G.M., "Guidance and Control Law Methodology,"
This study was conducted to evaluate several methods for AFATL-TR-79-86, Oct. 1979.
modeling target acceleration in an extended Kalman filter, 5
Fuller, J.W. et al., "Tactical Missile Guidance and Control Law
and, by no means, are any,of the four filters considered to be Methodology," AFATL-TR-80-144, Oct. 1980.

From the AIAA Progress in Astronautics and Aeronautics Series..,

EXPLORATION OF THE OUTER SOLAR SYSTEM—v. 50
Edited by Eugene W. Greenstadt, Murray Dryer, andDevrieS. Intriligator

During the past decade, propelled by the growing capability of the advanced nations of the world to rocket-launch space
vehicles on precise interplanetary paths beyond Earth, strong scientific interest has developed in reaching the outer solar
system in order to explore in detail many important physical features that simply cannot be determined by conventional
astrophysical observation from Earth. The, scientifically exciting exploration strategy for the outer solar system—planets
beyond Mars, comets, and the interplanetary medium—has been outlined by NASA for the next decade that includes ten or
more planet fly-bys, orbiters, and entry vehicles launched to reach Jupiter, Saturn, and Uranus; and still more launchings are
in the initial planning stages.
This volume of the AIAA Progress in Astronautics and Aeronautics series offers a collection of original articles on the first
results of such outer solar system exploration. It encompasses three distinct fields of inquiry: the major planets and
satellites beyond Mars, comets entering the solar system, and the interplanetary medium containing mainly the particle
emanations from the Sun. i;
Astrophysicists interested in outer solar system phenomena and astronautical engineers concerned with advanced scientific
spacecraft will find the book worthy of study. It is recommended also as background to those who will participate in the
planning of future solar system missions, particularly as the advent of the forthcoming Space Shuttle opens up new
capabilities for such space explorations.

251 pp., 6x9, illus., $15.00Member $24.00List

TO ORDER WRITE: Publications Order Dept., AIAA, 1633 Broadway, New York, N.Y. 10019
 This article has been cited by:

1. B. Kada. 2014. Arbitrary-Order Sliding-Mode-Based Homing-Missile Guidance for Intercepting Highly Maneuverable
Targets. Journal of Guidance, Control, and Dynamics 37:6, 1999-2013. [Abstract] [Full Text] [PDF] [PDF Plus]
2. Robert Chen, Jason Speyer, Dimitrios LianosHoming Missile Guidance and Estimation for Three-Dimensional
Intercept . [Citation] [PDF] [PDF Plus]
3. Robert H. Chen, Jason L. Speyer, Dimitrios Lianos. 2007. Homing Missile Guidance and Estimation Under Agile
Target Acceleration. Journal of Guidance, Control, and Dynamics 30:6, 1577-1589. [Citation] [PDF] [PDF Plus]
4. Robert Chen, Jason Speyer, Dimitrios LianosHoming Missile Guidance and Estimation Under Agile Target
Acceleration . [Citation] [PDF] [PDF Plus]
5. Yong-an ZHANG, Di ZHOU, Guang-ren DUAN. 2004. Multiple Model Filtering in the Presence of Gaussian Mixture
Measurement Noises. Chinese Journal of Aeronautics 17, 229-234. [CrossRef]
6. Sanjay Talole, Vinay Upreti, S PhadkePredictive Estimation and Control Based Missile Autopilot Design . [Citation]
[PDF] [PDF Plus]
7. P. Gurfil, N.J. Kasdin. 2004. Improving Missile Guidance Performance by In-Flight Two-Step Nonlinear Estimation of
Radome Aberration. IEEE Transactions on Control Systems Technology 12, 532-541. [CrossRef]
Downloaded by RYERSON UNIVERSITY on December 1, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.19861

8. Pini Gurfil, Jeremy KasdinImproving Missile Guidance Performance by In-Flight Two-Step Nonlinear Estimation of
Radome Aberration . [Citation] [PDF] [PDF Plus]
9. Pini Gurfil, N. Jeremy KasdinOptimal Passive and Active Target Tracking Using the Two-Step Estimator . [Citation]
[PDF] [PDF Plus]
10. M. Mehrandezh, M. N. Sela, R. G. Fenton, B. Benhabib. 2000. Proportional navigation guidance for robotic interception
of moving objects. Journal of Robotic Systems 17:10.1002/(SICI)1097-4563(200006)17:6&lt;&gt;1.0.CO;2-Z, 321-340.
[CrossRef]
11. Di Zhou, Chundi Mu, Wenli Xu. 1999. Adaptive Two-Step Filter with Applications to Bearings-Only Measurement
Problem. Journal of Guidance, Control, and Dynamics 22:5, 726-728. [Citation] [PDF] [PDF Plus]
12. STEPHAN HEPNER, LUCA BAGNASCHI, HANS GEERINGTowards robust tracking of maneuvering targets .
[Citation] [PDF] [PDF Plus]
13. STEPHAN A. R. HEPNER, HANS P. GEERING. 1990. Observability analysis for target maneuver estimation via
bearing-only and bearing-rate-only measurements. Journal of Guidance, Control, and Dynamics 13:6, 977-983. [Citation]
[PDF] [PDF Plus]
14. JASON L. SPEYER, KEVIN D. KIM, MINJEA TAHK. 1990. Passive homing missile guidance law based on new target
maneuver models. Journal of Guidance, Control, and Dynamics 13:5, 803-812. [Citation] [PDF] [PDF Plus]
15. JASON SPEYER, KEVIN KIM, MINJEA TAHKA PASSIVE HOMING MISSILE GUIDANCE LAW BASED ON
NEW TARGET MANEUVER MODELS . [Citation] [PDF] [PDF Plus]
16. STEPHAN HEPNER, HANS GEERINGOn the observability of target maneuvers via bearing-only and bearing-rate-
only measurements . [Citation] [PDF] [PDF Plus]
17. J.-L. DURIEUX. 1986. Comparison of angular and metric guidance laws for tactical missiles. Journal of Guidance,
Control, and Dynamics 9:4, 505-507. [Citation] [PDF] [PDF Plus]
18. S. B. SKAAR, L. TANG, I. YALDA-MOOSHABAD. 1986. On-off attitude control of flexible satellites. Journal of
Guidance, Control, and Dynamics 9:4, 507-510. [Citation] [PDF] [PDF Plus]