SICE Annual Conference 2008

August 20-22, 2008, The University Electro-Communications, Japan

Time-to-go Estimation Filter for Anti-Ship Missile Application

Ick-Ho Whang1 and Won-Sang Ra2
1
Department of Guidance and Control, Agency for Defense Development, Taejeon, Korea
(Tel : +82-42-821-4413; E-mail: [email protected])
2
Department of Guidance and Control, Agency for Defense Development, Taejeon, Korea
(Tel : +82-42-821-4411; E-mail: [email protected])

Abstract: In this paper, we propose a new time-to-go estimator for conventional PN guided sea-skimming missiles is
proposed. In contrast to the conventional method, our method does not only consider the length of the curved homing
trajectory caused by the initial heading error, but also tried to mitigate the influence of the seeker measurement noises.
In addition, by adopting a new coordinate system moving together with the target in the estimator design, the effect of
target movement on the homing trajectory is efficiently took into account. Simulation results show that the proposed
filter can estimate the time-to-go very accurately with small computational load.

Keywords: time-to-go, estimation, Kalman filter, proportional navigation guidance, anti-ship missiles

1. INTRODUCTION trajectory is not a straight line. To solve the problem,

Tahk et. al. proposed a time-to-go estimation algorithm
In modern anti-ship missile (ASM) design, taking the engagement geometry causing the curved
sea-skimming has been widely accepted as an effective trajectory into account [1]. However, since it does not
way to enhance the missile’s survivability and warhead consider the effect of the sensor measurement errors, the
effectiveness. For the sea-skimming ASM, guidance resultant estimate tends to show unsatisfactory error
laws have been conventionally designed independently performances. In addition, since the algorithm requires
in horizontal plane and in veritcal plane. For horizontal the knowledge of the position and velocity of the target,
plane guidance, proportional navigation (PN) guidance a filter to estimate the target state is additionally
using the microwave seeker line-of-sight (LOS) rate required to realize the algorithm. It makes the estimator
measurements is commonly applied while, for vertical to have complex structures producing large
plane guidance, precision height hold guidance followed computational load.
by terminal vertical homing maneuver is generally In this paper, in order to consider the target moving,
applied. Since the sea-skimming height commands and under the assumption of constant velocity moving target,
the optimal gains for the terminal height maneuver in an inertial coordinate system moving along the target
vertical plane are mainly dependent on the time-to-go, with constant velocity is introduced. And the length of
i.e. the required time for completing the homing the curved PNG trajectory is evaluated using only the
guidance, it is very important to estimate the time-to-go quantities which can be measured by the seeker installed
precisely. Besides, time-to-go estimates can play an in ASM. The time-to-go estimator is designed based on
important role in determination of many guidance the trajectory length and the Kalman filter theory. The
events such as warhead arming, and homing failure performance of the proposed estimator is demonstrated
decision. through a simulation example of a representative ASM
In this paper, we design a time-to-go estimation filter homing engagement condition.
for horizontal PN guidance (PNG) trajectories. In
sea-skimming homing missile cases, most part of the 2. TIME-TO-GO ESTIMATOR DESIGN FOR
vertical trajectory is just to hold low altitudes.
PNG TRAJECTORIES
Consequently, the length of the homing trajectory is
nearly equivalent to that of the horizontal trajectory.
2.1 Trajectory evaluation for horizontal PNG
So it may not cause problems to approximate the
In this section, the two dimensional trajectory and its
time-to-go of the overall 3-dimensional homing
length for PN guided missiles are evaluated. The motion
trajectory with the time-to-go of the horizontal PNG
equations are derived in the same way of [2]. The
trajectory only. As a result, the time-to-go estimator for
standing assumption for the target is that the target ship
the two dimensional PNG trajectories which will be
moves with constant velocity until the homing guidance
proposed here can be also used as an effective
completes. And we will derive the equations in
time-to-go estimator for overall 3-dimensional homing
Li -frame which is defined at every filtering instant of
trajectories.
r t i as follows:
Conventionally, the approximation of where r (L1) The origin of Li -frame locates at the target
Vc
position at t i and it moves with the velocity
and Vc are relative range and closing velocity (or range
equivalent to the target velocity.
rate) respectively, has been widely used as a time-to-go (L2) The X-axis of Li -frame directs along the
estimate. Unfortunately, however, it tends to produce
large estimation errors especially when the homing line-of-sight(LOS) vector at t i . And it holds

- 247 -
PR0001/08/0000-0247 ¥400 © 2008 SICE
 until the homing guidance completes.
y (0) 0 , y (0) Vrel\ 0 y 0 (5)
Note that the Li -frame is inertial since it moves
without acceleration and rotation. Since Li -frame Integrating Eq.(4) and applying the initial conditions of
moves together with the target, the trajectory formed in (5) reach the following results.
Li -frame will be relative movement of the missile to the
target. y 0 ª§ t ·
N 1
º
y (t ) t go «¨  1»
go
¸¸ (6)
1  N «©¨ t f ¹ »
€ ¬ ¼

y 0 ª º
N 1
§t ·
}™Œ“ y (t ) «1  N ¨ go ¸¸ » (7)
1 N « ¨ tf »
¬ © ¹ ¼
ɒ tšš“ŒG
In the above equations, t f is guidance ending time, and
Ʌ t go tf  t .
{ˆ™ŽŒ› 
§ 1· dy
From the equations of (2) and (7), y ¨  ¸
Fig. 1 Engagement geometry in Li -frame © Vc ¹ dx
holds. And therefore the length of the homing trajectory
Fig. 1 shows the engagement situation of the missile can be evaluated as follows.
and target in Li -frame at t k ( k t i ). In the figure,
2 2
Vrel , \ , and O are relative velocity, heading error 0 § dy · 0 1 § dy ·
angle, and LOS angle at the current time t k ,
li ³
Vct f
1  ¨ ¸ dx |
© dx ¹
³V t 1  2 ¨© dx ¸¹ dx
c f

respectively.
Let’s assume that the missile flies around the collision 0
ª 1 § y ·2 º
triangle. The approximated relative motion of the ³
tf
«1  ¨ ¸ » Vc dt go
«¬ 2 © Vc ¹ »¼
missile to the target will be derived in the similar way of (8)
[2]. In this case, the LOS angle can be approximated by § 1 § y 0 · ·
2

x0 ¨1  ¨ ¸ ¸
¨ 2(2 N  1) © Vc ¹ ¸
y © ¹
O| (1)
x § \0 2
·
x0 ¨ 1  ¸
© 2(2 N  1) ¹
where
In the above equation, the subscript ‘0’ means the
x | Vc t go (2) quantity at t i .
Since, in most cases, the seeker for ASM can measure
and, x , y , and Vc are X and Y elements of the the relative range r (i ) , range rate r(i ) , and LOS rate
relative position vector, and closing velocity, O (i ) , we can construct the quantities in Eq.(8) using the
respectively. In addition, t go means time-to-go.
following relations.
Let the dot located over a quantity represent its time
derivative. In PNG homing case, the acceleration is x0 r (i ) (9)
expressed as follows[2].
y 0 r (i ) O (i )
\0 (10)
y |  NVc O
 (3) Vc r(i )
Vc r(i ) (11)
Integrating Eq.(3) becomes
It means that we can evaluate the length of the
y  NVc O  c1 . (4) homing trajectory in Li -frame using the seeker
measurements and some filters.
In the above equations, N is effective navigation ratio
and c1 is integration constant. 2.2 Time-to-go Estimator design
Since the above differential equations are time In this section, the new time-to-go estimator is
invariant, we may make the following initial conditions derived based on the results of the section 2.1. The basic
by regarding the time instance of t i as the initial concept of the design is as follows:
time. (C1) Every filtering instant t k , a new Lk -frame

- 248 -
 is constructed and the trajectory length lk is
evaluated in the frame through Eq.(8)-(11) using lm ( k ) 1
zkTgo | t go (k )  vkTgo (19)
the seeker measurements. Sk Sk
(C2) Total speed S k of the missile observed in
Lk -frame is evaluated in some appropriate way. For the above equation, S k can be chosen in many
lk ways such that (i) desired homing speed of the ASM,
(C3) Time-to-go is estimated by t go | .
Sk
 
2
(ii) rm (k ) 2  rm (k ) Om (k ) or its filtered value, or (iii)

The basic structure of the time-to-go estimator is some other estimation method in consideration of total
based on Kalman filter theory[3]. In order to apply the ASM trajectory including the vertical maneuvers.
Kalman filter, we reference the fact that the time-to-go With the system model (12) and measurement model
should decrease by 't , the sampling interval, every (19), the final time-to-go estimator will be easily
filtering time instant. Accordingly, the system model for derived by the Kalman filter theory[3]. The resultant
the time-to-go dynamics will be estimator is summarized in Table 1.

Table 1 Time-to-go Estimator.

t go (k  1) t go (k )  't  wkTgo , wkTg  N  0; QkTgo  (12)
System Propagation:
where N (m; R ) means the Gaussian distribution with PTgo (k  1| k ) PTgo (k | k )  QkTgo ,
mean m and variance R . tˆgo (k  1| k ) tˆgo ( k | k )  't
The time-to-go measurement model for applying the
Measurement Update:
Kalman filter is derived from Eq.(8)-(11) under the 1
concept of (C1)-(C3) stated above. Derivation of the KTgo (k  1) PTgo (k  1| k ) ª¬ PTgo ( k  1| k )  RkTgo º
1 ¼
time-to-go measurement model begins with the
assumption that the seeker measurements are obtained PTgo (k  1| k  1) ª¬ I  KTgo (k  1) º¼ PTgo (k  1| k )
along the following models. tˆgo (k  1| k  1) tˆgo (k  1| k )

rm (i ) r (i )  vir , vir  N  0; Rir  (13)  KTgo (k  1) zkTgo

1 tˆgo (k  1| k ) 
1 l
rm (i ) r(i )  vir , vir  N  0; Rir  (14) where RkTgo Rk .
Sk2
Om (i ) O (i )  viO , viO  N  0; RiO 
  
(15)
3. SIMULATION RESULTS
Inserting Eqs.(13)-(15) into Eq.(8) followed by
discarding high order terms becomes the trajectory Computer simulations are carried out to demonstrate
length measurement model of Eq.(16). the performance of the proposed time-to-go estimator.
The simulation conditions are carefully designed to
§ describe the representative sea-skimming ASM
§ rm (k )Om (k ) · ·
2
1 engagement situations. Simulation conditions are as
lm (k ) rm (k ) ¨1  ¨ ¸ ¸
¨ 2(2 N  1) ¨© rm (k ) ¸¹ ¸ (16) follows:
© ¹
| lk  vkl , vkl  N  bk ; Rkl  - Initial relative range = 5 km
ª15º
In the above, - Target velocity = « » [m/sec]
¬15¼
­°§ 1 · 2 ªcos 30D º
· r ½°
2
§ rk · O § rk Ok
2
rk - Initial missile velocity = 250 u « D »
[m/sec]
bk ®¨ ¸ Rk  ¨ ¸ Rk  ¨¨ 2
r
¸¸ Rk ¾ (17) ¬ sin 30 ¼
2(2 N  1) °© rk ¹
¯ © rk ¹ © rk ¹ °¿ - Missile speed = 250 [m/s]
­ \k \ k2 ½
2
- PNG effective navigation ratio( N ) = 3
Rkl | ®   1¾ Rkr - Seeker range meas. error s. d. = 50 m
¯ 2 N  1 2(2 N  1) ¿
. (18) - Seeker range rate meas. error s. d. = 20 m
2 2
­ \ k rk2 ½ O ­ \ k2 rk ½ r - Seeker LOS rate meas. error s. d. = 0.5 [deg/sec]
® ¾ Rk  ® ¾ Rk
¯ (2 N  1)rk ¿ ¯ (2 N  1)rk ¿ The time-to-go estimator design parameters are as
follows:
Since the measurement error bias bk is very small,
we can obtain the following time-to-go measurement - Kalman filter process noise covariance:
model. QkTgo 0.012

- 249 -
 - Kalman filter measurement noise covariance: r
method of , (ii) time-to-go measurements of Eq.(19),
RkTgo 0.252 (It is equivalent to the RkTgo in the Vc
case that range of r 2000 and heading error and (iii) proposed time-to-go estimator using Kalman
of \ 20D .) filter. As we can see in the figure, the newly proposed
- S k for zkTgo evaluation(see Eq.(19) and estimator shows the best performance. As a results, it
can be concluded that the proposed filter can estimate
descriptions below it): the second method is used time-to-go with negligible computational load.
 
2
i.e. rm (k ) 2  rm (k ) Om (k ) .
4. CONCLUSION

Since we use the constant QkTgo and RkTgo in In many conventional sea-skimming ASM
Kalman filter design, the resultant time-to-go estimator applications, time-to-go has been one of the most useful
becomes steady state Kalman filter of Table 1. parameters. However, there are few intensive studies on
Therefore, the required computations of the proposed the time-to-go estimation. In this paper, a new
time-to-go Kalman filter are negligible because we can time-to-go estimator for ASM guided by PNG is
use the constant steady state Kalman gain. proposed based on the Kalman filter theory. Based on
Fig. 2 shows the horizontal trajectory of the missile the fact that the time-to-go decreases by sampling
and the target. In the figure blue line and red line interval, the system model is introduced. A new
represent missile and target trajectories, respectively. measurement model for time-to-go is derived based on
the approximated length of the flight trajectory in
Homing Trajctory Lk -frame. The Lk -frame is an inertial frame moving
700
with the target ship. Thanks to using the Lk -frame, the
600
proposed estimator easily takes the effect of the target
movement on the homing trajectory into account.
500
Simulation results show that the proposed estimator is
400 very effective for conventional sea-skimming ASM
applications.
y-pos.

300

200
REFERENCES
100
[1] M. J. Tahk, C. K. Ryoo, and H. Cho, “Recursive
0
0 1000 2000 3000 4000 5000 6000
Time-To-Go Estimation for Homing Guidance
x-pos. Missiles,” IEEE Trans. Aerosp. & Electr. Syst., Vol. 38,
No.1, Jan. 2002
Fig. 2 Homing Trajectory [2] Paul Zarchan, Tactical and Strategic Missile
Guidance, 2nd ed., Vol. 157, Progress in Astronautics
RMS Time-to-Go Estimation Errors and Aeronautics, A Volume in the AISS Tactical Missile
6
R / Vc
Series, 1994.
Length / Vmsl
proposed K.F.
[3] Frank L. Lewis, Optimal Estimation, John Wiley &
5
Sons, 1986.

4
RMS Tgo Errors

0
0 5 10 15 20 25
time

Fig. 3 RMS errors produced by various time-to-go

estimation methods.

Fig. 3 compares the performances of the three

time-to-go estimation methods – (i) conventional

- 250 -