3-D Optimal Evasion of Air-to-Surface Missiles against

Proportionally Navigated Defense Missiles

Sung-Bong Cho*, Chang-Kyung Ryoo*, and Min-Jea Tahk*
* Division of Aerospace Engineering, KAIST, Daejeon, Korea
(Tel : +82-42-869-3794; E-mail: {sbcho, ckryoo, mjtahk}@fdcl.kaist.ac.kr)

Abstract: In this paper, we investigate three dimensional optimal evasive maneuver patterns for air-to-surface attack missiles
against proportionally navigated anti-air defense missiles. Interception error of the defense missile can be generated by evasive
maneuver of the attack missile during the time of flight for which the defense missile intercepts the attack missile. Time varying
weighted sum of the inverse of these interception errors forms a performance index to be minimized. Direct parameter optimization
technique using CFSQP is adopted to get the attack missile’s optimal evasive maneuver patterns according to parameter changes of
both the attack missile and the defense missile such as maneuver limit and time constant of autopilot approximated by the 1st order
lag system. The overall shape of resultant optimal evasive maneuver to enhance the survivability of air-to-surface missiles against
proportionally navigated anti-air missiles is a kind of deformed barrel roll.

Keywords: Evasive maneuver, Survivability, Optimization, Deformed barrel-roll

1. INTRODUCTION 2. FORMULATION OF EQUATIONS OF MOTION

AND INTERCEPTION ERRORS
Rapid development of recent technology of anti-air defense
Let us consider the engagement scenario between a
systems has menaced air-to-surface missiles’ survivability. For
air-to-surface attack missile and a anti-air defense missile as
the enhancement of the survivability of air-to-surface missiles,
shown in Fig. 1. In this scenario, it is assumed that the attack
we can consider a method to augmenting special evasive r
maneuver into the terminal homing phase. Evasive maneuver missile controls acceleration vector am normal to its velocity
of air-to-surface missiles can be defined as a special maneuver to guide to the target. On the other hand, the target
not only to increase survivability but also to minimize terminal continuously launches anti-air missiles to intercept the attack
miss distance without a priori information of the parameters or missile.
flight states of the threats.
The weaving motion[1] in two-dimensional space or the Equations of motion of the attack missile
barrel roll maneuver[2] in three-dimensional can be a The equations of motion of the attack missile in the three
candidate for evasive maneuver. Then, are they optimal? If not, dimensional space are given by
r r r r
what kind of maneuver pattern is the best for evasion? In Ref. r&m = vm , rm (t0 ) = rm 0
[3], an optimal evasive maneuver policy of anti-ship missiles r r r r (1)
against the CIWS(Close-In Weapon System) has been v& = a , v (t ) = v
m m m 0 m0

discussed, where the optimal evasive trajectories are where rm (t ) is denotes the distance of the attack missile
characterized by a kind of barrel roll. However, it may not be
from the target, vm is the velocity of the attack missile. The
the best for the case of homing threats. r
Based on the approach of Ref.[3], in this paper, we guidance command vector ucV is defined in velocity frame as
investigate three dimensional optimal evasive maneuver denoted in Fig. 2;
patterns for air-to-surface attack missiles against r T
proportionally navigated anti-air defense missiles. The optimal ucV =  0 u yaw u pth  (2)
control problem for evasive maneuver of attack missiles r
and must be normal to vm such that
considered in this paper is to find an acceleration command r r
which minimizes the performance index given by the time ucV • vm = 0 (3)
varying weighted sum of the inverse of an interception error of
the defense missile with the terminal constraints of zero miss z
distances. Interception error of the defense missile during the
entire engagement can be calculated from the homing loop
Interception error
adjoint [4] of the defense missile. It is assumed that both
attack and defense missiles are the 1st order lag systems with a ε
m Defense
different time constant. And we assume that only the attack vd missile
missile has a command limit. In this paper, the direct input Attack
Target
missile
parameter optimization technique using CFSQP [5] is used to vm rm
find the optimal solution. Optimization results show that
optimal evasive trajectory also becomes a kind of the barrel
roll whose shape varies according to the time constants of both y
missiles as well as navigation constant of the defense missile. x
The equations of motion of the air-to-surface attack missile
and the interception error of the anti-air defense missile in
three-dimensional space are formulated in section 2. Section 3 Fig. 1 Three dimensional engagement geometry between
deals with the optimal control problems and their numerical the attack missile and the defense missile
solutions for a typical engagement scenario. And section 4 is
conclusion.

514
 z
Vm
1 1
am ( s )
s s2

γm
y
upth x δ 1 H (τ )
−vcW ( s)
vcτ
ψm
uyaw

Fig. 2 Definition of guidance command vector

The autopilot of the attack missile is assumed as the first order Fig. 3 Generalized homing loop adjoint of
r a defense missile
lag system so that the resultant acceleration a Vm is represented
as The guidance system of the defense missile is represented in
r 1 r r the time domain by W (τ ) . A single-lag guidance system can
a& Vm = (uc − amV ) (4)
Ta be represented by
where Ta denotes the the time constant of the attack missile. N′
r W (s) = (9)
Then, a Vm is transformed into inertial reference frame such s(1 + sTd )
that where Td is the effective time constant of the guidance
r T r
system and N ′ is the navigation constant. And H (τ )
am = CIV  amV (5)
denotes an adjoint signal of interest and calculated as
where CIV is calculated by using azimuth ψ m and elevation 1
angle γ m of the flight path: H (τ ) = ∫ W ( x)[δ (τ − x ) − H (τ − x)]dx (10)
τ
 cγ mcψ m cγ m sψ m sγ m  Converting from the time domain to the frequency domain
  using Laplace transform, we can express Eq. (10) as
C = Ty (−γ m )Tz (ψ m ) =  − sψ m
V
cψ m 0  (6)
− dH ( s )
I
 − sγ mcψ m − sγ m sψ m cγ m  = W ( s )[1 − H ( s )] (11)
ds
where Recall also that
v v − dH ( s )
ψ m = tan −1 y , γ m = tan −1 2 z 2 (7) d
[1 − H ( s )] = (12)
vx vx + v y ds ds
Substitute Eq. (12) into Eq. (11) and take integral to obtain
d (1 − H ( s))
∫ 1 − H (s) = ∫ W (s)ds
Interception error of the defense missile (13)
we assume that neither gravity nor aerodynamic forces
affect the ballistics of the defense missiles. Then, the speed of Then, Eq. (13) becomes
r
the defense missile vd (= vd ) remains in constant during the
entire engagement. And we also assume the attack missile lies
(
1 − H ( s ) = exp ∫ W ( s ) ds ) (14)

on the collision path so that it does not deviate much from the Now let us find the miss due to a step attack missile maneuver
reference x-axis. Then, we approximate the interception time for a single lag guidance system. We can obtain new
for the defense missile to intercept the attack missile as expression of Eq. (14) by substituting Eq. (9) into Eq. (14);
r (t ) r (t )  N′ 
τ (t ) ≈ m = m (8) 1 − H ( s ) = exp  ∫ ds 
vm + vd vc  s (1 + sTd ) 
N′ (15)
where vc denotes the closing velocity between the attack   1 
missile and the defense missile. For homing missiles, it is =  s  s + 
intuitively true that τ is an implicit function of time and has   Td  
Then, from the Fig. 3, the interception error of the defense
its maximum value at t = 0 and then monotonically
missile due to the evasive maneuver of the attack missile is
decreases until becoming 0.
given by
System lag and command limit are the major factors to
1 − H (s)
cause miss distance of the defense missile. In most cases, the MNM ( s ) = am ( s )
command limit of the defense missile can be neglected since it s3
N′ (16)
is enough high compared to that of the attack missile. 1  1 
Therefore, the interception error of the defense missile is = 3  s  s +   am ( s )
caused by the evasive maneuver of the attack missile during s   Td  
the interception time. We can calculate the interception error If F ( s ) is defined as
of the defense missile using the method of adjoint [4]. If the N′
defense missile is guided by PNG(Proportional Navigation 1 − H (s) 1   1 
F ( s) = = 3 s  s +  (17)
Guidance), the homing loop adjoint can be represented as s3 s   Td  
shown in Fig. 3. then, the interception error in time domain can be obtained by

515
taking the inverse Laplace transform of Eq. (16); missile against the PNG guided defense missile will be
MNM (τ ) = L−1 {MNM ( s )} discussed.
= L−1 { F ( s ) am ( s )} 3. 3-D OPTIMAL EVASIVE MANEUVER
τ (18)
= ∫ f ( x)am (τ − x)dx PROBLEM
0

= f * am
Now, let us consider following optimal control problem;
where f is the inverse Laplace transform of F and “ * ”
denotes the convolution operator. And, τ is the adjoint time r
Find uc which minimizes
of the defense missile and can be interpreted the time of flight
tf τ 2 (t )
for the defense missile to intercept the attack missile. J =∫ dt (28)
For each navigation constant, F ( s ) and its inverse 0 ε (t )
Laplace transform f ( x) are given by subject to Eq. (1)
with terminal constraint
For N ′ = 3 r
rm (t f ) = 0 (29)
1
F N ′= 3 ( s ) = 3
(19) and inequality input constraint
 1 r
s +  uc ≤ U c (30)
 Td 
f N ′= 3
( x) = 0.5 x 2e − x Td (20) where U c denotes the maximum permissible acceleration
For N ′ = 4 command. Since the interception error ε converges to 0 as
s the defense missile approaches to the attack missile, the cost
F N ′= 4
(s) = 4
(21) function becomes so large in the terminal flight phase that a
 1
s +  lot of numerical effort to minimize the cost may be
 Td  concentrated more on this phase than on initial/midcourse
 phase. For the stability of the solution finding algorithm,
x 
f N ′= 4
( x) = x 2e − x Td  0.5 −  (22) τ 2 (t ) is considered into the performance index to reduce the
 6Td 
large weighting effect due to the inverse of the interception
For N ′ = 5 error in the terminal phase. The closed-form solutions of this
s2 optimal control problem might not be easily derived due to the
F N ′= 5
( s) = 5
(23) nonlinearities included in the performance index and the
 1
s +  inequality constraints. To find the policy of three dimensional
 Td  evasive maneuver of the air-to-surface missile against
proportionally navigated the anti-air missile, we should adopt
2 − x Td  x x2 
f
N ′= 5
( x ) = x e  0.5 − +  (24) numerical optimization techniques to solve the problem. This
 3T 24Td 2  optimal control problem is converted into parameter
Finally, by substituting Eq. (20), (22) and (24) into Eq. optimization problem with unknown parameter vector
(18), we can obtain the miss distances for adjoint time τ , composed of discretized control and flight time such that
T
which is caused by the evasive maneuver of the attack missile; X = u yaw (i ), u pth (i ), t f  , i = 0,..., N (31)
ε N ′ = 3 (τ ) = MNM N ′ = 3 (τ )
(25) Therefore, the number of unknown parameter is 2( N + 1) + 1 .
= ∫ ( 0.5 x 2e − x Td )am (τ − x)dx
τ

0
As a tool for parameter optimization, CFSQP [5] which is an
open code for constrained optimization problems based on
ε N ′=4
(τ ) = MNM N ′=4
(τ ) sequential quadratic programming is used. Integration of
τ 
 x (26) equations of motion to evaluate the value of performance
= ∫  x 2e − x Td  0.5 −  am (τ − x )dx index and the violation of terminal constraints is performed by
0 
  6T d  the 4th order Runge-Kutta method. During integration, the
ε N ′ = 5 (τ ) = MNM N ′ = 5 (τ ) controls u yaw (t ) and u pth (t ) are assumed to be linearly

τ
changed between adjacent nodes.
 x x2   (27)
= ∫  x 2e − x Td  0.5 − + 2 
am (τ − x) dx The initial conditions of engagement scenario between the
0  
  3Td 24Td   attack missile and the defense missile are given as
Note that the interception error ε becomes 0 as τ goes 1) For the attack missile;
to 0 and if the attack missile does not maneuver, then the At the beginning of the evasive maneuver, it is assumed that
interception error is always zero. We also note that for the attack missile lies on the near collision path to target so
non-zero am (t ) for 0 ≤ t ≤ t f , by using the convolution that the initial acceleration is very small.
r
rm (t0 ) = [ 4000 0 0] (m)
T
integral of Eqs. (25), (26) and (27) we can calculate the
r
vm (t0 ) = [ −300 0.1 0.1] (m s )
interception errors for the adjoint time without multiple run of T

nonlinear simulation. Here, t f denotes the flight time of the

r
aVm (t0 ) = [ 0 0.1 0.1] (m s )
T 2
attack missile to intercept the target. These analytic
expressions on the interception error, then, can be used to Guidance command is highly limited and realized by the 1st
assess the evasive performance or to evaluate the cost of order lag approximation;
survivability of the attack missile. In following section, an U c = 30 (m s 2 ) ( ≈ 3g )
optimal control problem for evasive maneuver of the attack

516
 Ta = 1.0 (rad s )
2) For the defense missile;
We assume that the defense missile launches at the target and
its initial velocity is more than twice when it is compared to
50
that of the attack missile. It is also assumed that velocity of the
defense missile does not vary during the entire engagement. 40
r
rd (t0 ) = [ 0 0 0] ( m)
T

30
r

z (m)
vd = 700 (m s )
20
Time constant of the defense missile is the same as the attack
10
missile.
Td = 1.0 (rad s ) 0
4000
3500
It is assumed that the defense missile has perfect measurement 3000
-140
on the attack missile and no command limit. 2500 -120
2000 -100
Based on the above engagement conditions, we investigate 1500
-80
-60
the three-dimensional optimal evasion of the attack missile x (m1000 -40 m)
and its evasive performance for the case of that the navigation ) -20 y(
500
constant of the defense missile is 3 and 4, i.e., N ′ = 3 and
0
0 20

N ′ = 4 . Fig. 4 and 5 show the optimal trajectories of the

attack missile to maximize its survivability for both navigation Fig. 4. Optimal trajectory of the attack missile against the
constants of the defense missile, respectively. Both trajectories defense missile with PNG with N ′ = 3
have kinds of deformed barrel roll due to homing
characteristics of the attack missile in the last part of the flight.
The evasive maneuver of the attack missile for N ′ = 4 is
more radical and forms larger lateral displacement. Although
lateral trajectory displacement is larger, the interception error
for the defense missile with N ′ = 4 is rather small as shown
in Fig. 9 and 10. For both cases, the interception error 20

approaches to zero as the attack missile approaches to the 0

target. Performance index for both navigation constants is -20

given as
-40
J |N ′= 3 = 1.170e3 , J |N ′= 4 = 3.513e3
z (m)

-60
From the optimization results, it is obvious that the
-80
survivability of the attack missile is weakened for the attack
-100
missile with higher navigation constant. In general, PNG
shows that higher navigation constant is more effective for -120
4500
maneuvering targets. However, PNG with higher navigation -1404000
3500
-150
3000
constant tends to generate larger guidance command. Not 2500 -100
2000
considered in this paper, if the guidance command of the x ( 1500
m 1000 -50
m)
) 500 y(
defense missile is limited, the interception error will be 0 0
-500
increased. In this case, we cannot assert that higher navigation
constant of the defense missile always deteriorates the
survivability of the attack missile. Fig. 5. Optimal trajectory of the attack missile against the
defense missile with PNG with N ′ = 4
4. CONCLUSION

In this paper, three-dimensional optimal evasive maneuver

of air-to-surface attack missiles against proportionally 30
N=3
navigated anti-air defense missiles are investigated. We use N=4
20
the homing loop adjoint of the defense missiles to generate the
Yaw command (m/s )

interception error of the defense missiles. And then we apply

2

10
the direct input parameter optimization technique using
CFSQP to minimize the performance index given by the time 0
varying weighted sum of the inverse of the interception error.
Numerical results for typical navigation constants of the -10
defense missile show that the survivability of the attack
missile is weakened as the navigation constant of the defense -20

missile is increased. Optimal evasive trajectories of the attack

-30
missile are kinds of deformed barrel roll shape due to the
homing characteristics. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Optimization for various combinations of time constant of Time (sec)
both attack and defense missile under the consideration of the
command limit of the defense missile should be carried out as Fig. 6. Yaw command histories of the attack missile
a further study. against the defense missile with PNG

517
 35 12
30 11
N=3 N=3
25 N=4 10 N=4
20

Total interception error (m)

9
15
Pitch command (m/s )

8
2

10
7
5
6
0
5
-5
4
-10
3
-15
2
-20
1
-25
-30 0

-35 -1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Time (sec)
Time (sec)

Fig. 7. Pitch command histories of the attack missile Fig. 10 Time history of the interception errors of
against the defense missile with PNG the defense missile

REFERENCES
30
[1] P. Zarchan, “Proportional Navigation and Weaving
25
Targets,” Journal of Guidance, Control, and Dynamics,
N=3
Vol. 18, No. 5, Sep.-Oct., 1995, pp. 969-974.
Pitch-axix acceleration (m/s )

20
2

15
N=4 [2] F. Imado and S. Uehara, “High-g Barrel Roll Maneuvers
10
against Proportional Navigation from Optimal Control
5
Viewpoint,” Journal of Guidance, Control, and
0
Dynamics, Vol. 21, No. 6, Nov.-Dec., 1998, pp. 876-881
[3] C. K. Ryoo, Ick-Ho Whang, and M. J. Tahk, “3-D
-5
Evasive Maneuver Policy for Anti-Ship Missiles against
-10
Close-In Weapon Systems,” AIAA Guidance,
-15
Navigation, and Control Conference, AIAA 2003-5653,
-20
Austin, USA, Aug. 2003.
-25
[4] P. Zarchan, Tactical and Strategic Missile Guidance,
-20 -10 0 10 20 30
2
Third Edition, AIAA Progress Series in Astronautics and
Yaw-axis acceleration (m/s )
Aeronautics, Vol. 176, Washington DC, 1998.
[5] C. Lawrence, J. L. Zhou, and A. L. Tits, User’s Guide
Fig. 8 Projected acceleration patterns of the attack missile for CFSQP Version 2.5 : A C Code for Solving (Large
against the defense missile with PNG Scale) Constrained Nonlinear (Minmax) Optimization
Problems, Generating Iterates Satisfying All Inequality
Constraints, Institute for Systems Research, University
of Maryland, 1997.
4

0
z-error (m)

-2

-4
N=3
N=4
-6
-12.5 -10.0 -7.5 -5.0 -2.5 0.0 2.5 5.0 7.5
y-error (m)

Fig. 9 Projected interception errors of the defense missile

518