Paper no.

: 2003-003

Development of an Integrated Fuzzy-Logic-Based Missile

Guidance Law against High Speed Target

Chun-Liang Lin 1 , Hao-Zhen Hung 2 , Yung-Yue Chen 3 and Bor-Sen Chen 3

Abstract—How to develop a novel missile guidance law that can act effectively
against very high speed incoming targets, such as ballistic missiles, has been a major
concern in the defense technique. It has been unveiled recently in the literature that
the most effective way for a lower speed interceptor to successfully engage an
incoming target with very high speed is to limit the aspect angle between the missile
and target flight path to within 180 degree plus or minus few degrees. On the basis of
this requirement, a new integrated fuzzy guidance law is developed in this paper
against very high-speed maneuverable targets on three-dimensional space. An
integrated fuzzy type guidance law is explored for midcourse and terminal phases. It
is shown that the integrated guidance scheme offers excellent tracking performance as
well as performance sensitivity with respect to some parametric variations.

Index Terms—Fuzzy control, Missile, Guidance law, Ballistic target

-------------------------------------------------
1 Department of Electrical Engineering, National Chung Hsing University, Taichung 402, Taiwan
2 Institute of Automatic Control Engineering, Feng Chia University, Taichung 40724, Taiwan
3 Institute of Electrical Engineering, National Tsing Hua University, Hsinchu 325, Taiwan
 I. Introduction

Recently, the engagement of hypersonic ballistic targets has been the new

challenges in the world. When a ballistic target reenters the atmosphere after having

traveled a long distance, its speed is high and remaining time to ground impact is

relatively short. It has been a known design philosophy that the best trajectory for a

defense missile to intercept a hypersonic target is called the nearly head-on [1, 2].

Under this setting, the interceptor can ultimately hit the target without resorting to

huge lateral acceleration.

In the past, the guidance laws based on the above fundamental requirement were

found to be efficient for the targets that are non-maneuverable, and an acceptable miss

distance can be obtained. However, more and more new generation attacking targets

possess higher speed and maneuverability. Under these situations, it is hard to track

these targets by using classical guidance designs [3], since system performance

sensitivity was rarely considered in the design procedure that usually caused an

unacceptable miss distance. In addition, the missile-target dynamics are theoretically

highly nonlinear partly because the equations of motion are best described in an

inertial system, while aerodynamic forces and moments are represented in missile and

target body axis system. Besides, unmodeled dynamics or parametric perturbations

are usually remained in the plant modeling procedure, because of complexity of the

nonlinear guidance design problem, prior approximations or simplifications were

usually required before deriving the analytical guidance gains. Therefore, one does

not know exactly what is the true missile model, and the missile behavior may change

in unpredictable ways. Consequently, optimality of the resulting design cannot be

ensured any more.

Several guidance design techniques such as linear quadratic regulator (LQR) [4],

1
explicit guidance [5], modified proportional guidance [6] and geometric control [7]

have been previously proposed for the implementation of optimal midcourse or

terminal guidance. In particular, linear quadratic regulator [2], modified explicit

guidance [8] and neural networks [8, 9] have recently been applied to treat the ATBM

guidance design problem. However, solving the LQR problem or training the neural

network in real time is often infeasible. Traditionally, midcourse guidance was often

formulated as an optimal control problem to shape the trajectory to maximize the

terminal energy or to minimize the flight time. However, the implementation of the

optimal control midcourse guidance is very difficult since a nonlinear two-point

boundary value problem has to be solved to obtain the optimal trajectory [4]. The

neural network guidance might also be unreliable in practice; if the network was not

well trained, it usually sensibly interpolates input data that are new to the network [8,

9].

It is well known that fuzzy systems have the ability to make use of knowledge

expressed in the form of linguistic rules without completely resorting to the precise

plant models. In control applications, fuzzy logic approaches using if-then rules can

solve complex and practical problems. In recent years, researchers have also

attempted to apply it on missile guidance designs [10-12]. However, only a few

results were presented to support feasibility and performance of their proposed

approaches. Although many applications of fuzzy logic theory on missile guidance

and control have appeared with growing interest, no application to the

three-dimensional (3D) midcourse and terminal guidance problem has been attempted.

The fuzzy logic approach is extended in this paper to deal with the problem. An

integrated midcourse and terminal guidance law of an aerodynamically controlled

missile system to intercept a maneuvering target is developed. The principle aim to

2
the guidance law is to cope with the complex interactions between a missile system

and its changing environment to achieve excellent tracking performance. The design

procedure is systematically and orderly. The fuzzy strategy is robust to the changes of

environments and is closer to the ideal thought of guidance law designers. Simulation

studies are well performed to verify engagement performance, performance sensitivity

and to estimate the defensible volumes under various operating environments.

II. Missile and Target Dynamic Models

For problem formulation, consider the 3D translation equations of motion used to

compute the missile trajectory shown as in Figs. 1 and 2. The missile is modeled as a

point mass and the equations of motion are described by

vm = (T cos α − D) / m − g sin γ (1a)

γ = ( L + T sin α )cosφ /( mvm ) − g cosγ / vm (1b)

ψ = ( L + T sin α )sin φ /( mvm cos γ ) (1c)

xm = vm cos γ cosψ (1d)

y m = vm cos γ sinψ (1e)

hm = vm sin γ (1f)

where the aerodynamic forces of the missile are evaluated as the following

expressions:
1 2 1 2
L= ρ vm smCmL , D= ρ vm smCmD (2)
2 2

with CmL = CmLα (α − α0 ) and CmD = CmD 0 + µCmL 2 . The aerodynamic derivatives

CmLα , CmD 0 , and µ are given as functions of Mach number M, which is a function

of vm and the altitude hm . ρ is the air density measured in kg/m3 and was

shown to be accurately approximated exponentially as

3
 ρ = 0.12492(1 − 0.000022557hm ) 4.2561 g (3)

Also, the missile mass m and thrust T are given functions of time t. The initial

conditions of the state variables are denoted by

xm (0) = xm 0 , ym (0) = ym 0 , hm (0) = hm 0 ,

vm (0) = 0, γ (0)=γ 0 , ψ (0)=ψ 0

With regard to the missile-target engagement problem, the angle-of-attack α

and roll angle φ are treated as the control variables for the vertical and horizontal
planes respectively. Their roles are to steer the missile to the pitching and yawing
directions of the predicted intercept point respectively.

For the construction of target trajectories, a mathematical model describing for

the motion of the ballistic target under some simplifying assumptions is established.
Referring to [13], a vehicle in the reentry phase over a flat, nonrotating earth with
constant gravity is illustrated in Fig. 3. Suppose the ballistic target system model in
radar coordinates centered at radar site is expressed as

ρ vt2
vtx = − g cos γ 1 sin γ 2 + atx
2β
ρ vt2
vty = − g cos γ 1 cos γ 2 + aty (4)
2β
ρ vt2
vth = − g sin γ 1 − g + ath
2β

where vtx , vty , and vth denote the velocity components of vt along X, Y, and H axes,

respectively, atx ( t ) , aty ( t ) and ath ( t ) are uncertain accelerations due to model

uncertainties and maneuverability of the ballistic target, γ 1 , γ 2 , and the ballistic

coefficient β are defined as

⎛ vth ⎞
γ 1 ( t ) = tan −1 ⎜ − ⎟
⎜ vtx2 + vty2 + vth2 ⎟⎠
⎝
⎛ vtx ⎞
γ 2 ( t ) = tan −1 ⎜⎜ ⎟⎟ (5)
⎝ vty ⎠
W
β=
st CtD 0

4
where st , W and CtD 0 represent the reference area, weight and zero-lift drag
coefficient of the target, respectively.

III. Counter Attack Strategy

For a lower speed interceptor to successfully engage a high-speed target, an
important guideline is that both should be near head-on geometry while the
interceptor entering the homing phase. This strategy will minimize the line-of-sight
rotational rates and avoid saturating the command signals in the terminal phase. See
Fig. 4 for the engagement scenario.
The guidance scheme possesses three phases: midcourse phase, shaping phase
and terminal phase. To design the midcourse guidance law so that the missile attains
near head-on geometry before it entering the terminal phase, we define the preset
look-on point range as

R p = R + Rlock (6)

where Rlock is the seeker lock-on range which is constant and the relative range
between the missile and target is

R = ( xt − xm )2 + ( yt − ym )2 + ( ht − hm )2 (7)

The required time-to-go for the target attaining the preset look-on point can be
estimated as

RP Rp R
tgop = − = (8)
R P Rmtx vrx + Rmty vry + Rmth vrh

where the relative range and velocity on X, Y, and H coordinates are defined by

Rmtx = xt − xm , vrx = vmx − vtx

Rmty = yt − ym , vry = vmy − vty

Rmth = ht − hm , vrh = vmh − vth

with
vmx = vm cos γ cosψ

vmy = vm cos γ sinψ

5
 vmh = vm sin γ

Based on tgop obtained, the predicted lock-on point can be estimated by

x f = xt + vtx t gop , y f = yt + vty t gop , h f = ht + vth t gop (9)

To attain the near head-on geometry, we specify the desired flight-path angle as

 ⎛ h − hm ⎞
rf = tan −1 ⎜ f
⎜ x − x ⎟⎟
(10)
⎝ f m ⎠

 ⎛ y f − ym ⎞
ψ f = tan −1 ⎜⎜ ⎟⎟ (11)
⎝ x f − xm ⎠

The predicted line-of-slight (LOS) angle with respect to the lock-on point is given by

 ⎛ ht − hm ⎞
θ v = tan ⎜ 
−1
⎟ (12)
⎝ xt − xm ⎠
 
−1 ⎛ y t − y m ⎞
θ h = tan ⎜  ⎟ (13)
⎝ xt − x m ⎠

where the subscripts v and h denote, respectively, the vertical and horizontal planes
and the following estimations are used:

  
xt = xt + vtx t , yt = yt + vty t , ht = ht + vth t

with
R Rlock R
t = − lock = (14)

Rp Rmtx vrx + Rmty vry + Rmth vrh

In practice, target position and velocity information in the Cartesian inertial

frame were the uplink data obtained from a ground-based radar or satellite. For the
estimation of ballistic target trajectories using filters, one can refer, for example, to
[14]. Missile position and velocity information in the Cartesian inertial frame were
obtained from an inertial reference unit.

The 3D space is divided into two planes: vertical plane and horizontal plane.
With this arrangement, the desired guidance laws are designed independently.

A. Vertical Plane Guidance

6
 Referring to Figs. 5 and 6, the predicted velocity angle errors between the present
and final vectors on the vertical and horizontal planes are defined as

δ v = rf − γ (15a)


δ h = ψ f −ψ (15b)

and the heading error angles are defined as


σ v = γ −θv (16a)


σ h = ψ −θh (16b)

Designing of the fuzzy logic-based guidance law is inspired by the framework of

the traditional explicit guidance law [15], which used the velocity error angle and
heading error angle and each term multiplied with certain constant gains to generate
guidance commands.

The rule-based representation of a fuzzy logic controller does not include any
dynamics, and the computational structure of a fuzzy logic-control consisting of
fuzzification, inference, and defuzzification is highly nonlinear. These make a fuzzy
logic-control a natural nonlinear static transfer element like a static controller. A fuzzy
logic system is a kind of gain-scheduling systems which can cover a wide range of
operating regions. The specific feature is appropriate for the preceding guidance
design since the missile-target engagement system is a highly nonlinear model in the
sense. Based on this observation, let the fuzzy midcourse guidance law take the
following form:

α fm = f m (δ v , σ v ) (17)

where f m (⋅ , ⋅) is an input-output mapping of a fuzzy logic system. The variable δ v ,

specifying the predicted terminal speed relative to the current velocity, is contributed
by the velocity error vector; the variable σ v , specifying the terminal position relative
to the current position, is contributed by the position error vector.

When the missile approaches homing at the end of midcourse phase, trajectory
shaping becomes less important whereas minimizing the position error for good

7
accuracy becomes more important. While the relative range of missile and target is
less than an appropriate shaping range prior to seeker lock-on, the shaping guidance
law activates. In this phase, guidance commands issued from midcourse phase should
be smoothly conveyed to terminal phase. As the missile enters the terminal phase, the
role of velocity error becomes minor. It would be better to apply the position error
message σ v to design the terminal guidance law such that the aspect angle between
the missile and target flight path is closed to 180 degree. For the fuzzy terminal
guidance law, it is natural to select the heading error angle σ v and heading error

angle rate σ v as the antecedent variables like a PD-type controller, that is

α ft = ft (σ v ,σ v ) (18)

Figure 7 illustrates the block diagram of the operation of vertical guidance law where

Rshap denotes the appropriate shaping range between missile and target.

B. Horizontal Plane Guidance

As in the case of vertical plane guidance equations, the similar idea can be

adopted for the horizontal plane guidance equation except that the gravity term does

not exist. In addition, α becomes φ which is the roll angle, γ becomes ψ

which is the lift direction angle and θ becomes the inertial line-of-sight angle on the

horizontal plane.

Once the control variables α and φ are determined, the corresponding

acceleration command amp and amy of the pitch and yaw axes can be related

respectively by

1
amp = ( L + T sin α ) cos φ − g cos γ (19a)
m
1
amy = ( L + T sin α ) sin φ (19b)
m

IV. Fuzzy Rule-Based Guidance Law

8
 In the following development, our focus is placed on the vertical plane guidance

law design. Extension to the horizontal plane is similar and straightforward.

Vertical plane guidance

A. Midcourse phase: fuzzy logic-based midcourse guidance (FLMG)

The input and output variables of FLMG, also called the linguistic variables, are

σ v and δ v , and the output variable is α .

Each of the linguistic variables is assumed to take five linguistic sets defined as

LN ≡ large negative, LP ≡ large positive, SN ≡ small negative, SP ≡ small positive,

ZE ≡ zero. For these variables, the corresponding membership functions are adopted

in Fig. 8 where the physical domains are set to cover the operating ranges of all

variables. To simplify the computation in the actual operation, triangular membership

functions are suggested. It has been found that using complex forms of membership

functions, such as bell-shaped functions, cannot bring any advantage over the

triangular ones. For a nonevasive target, tracking errors in midcourse phase are

generally larger than those in terminal phase. To avoid overshooting the control

command, low distinguishable fuzzy sets (compared with those used in terminal

guidance) are suggested.

The rule base contains a collection of rules and forms an integral part of the total

knowledge embedded in the guidance computer. In midcourse phase, it is hoped that

the missile reaches the lock-on point with higher speed as far as possible. Therefore,

the velocity error angle is considered more important than the heading error angle,

and the former is placed with higher weight. As a result, one can expect that the

velocity error angle would converge quicker than the heading error angle during

midcourse phase. The exceptional case is that the situation of the tracking error

9
between the missile and target is very large. This means that the missile is

substantially moving away from the head-on condition. The fuzzy rules are thus

modified so that the heading error angle will not grow too much. A complete set of 25

guidance rules, listed in Table 1, has been applied to meet our purpose.

Based on the proposed interception strategy, the guidance rules can be divided

into five distinctive groups:

Group 1: δ v is closed to zero. This means that the current flight path angle γ is

consistent with the desired flight path angle γ f . The control action is thus

intended to correct the heading error angle σ v . However, the amount is

small or closed to zero so that the current δ v will not be altered.

Group 2: δ v is large negative. The control action is intended to significantly reverse

this trend.

Group 3: δ v is small negative. For the situation of σ v being negative or closed to

zero, the control action is impossible to simultaneously compensate for both
of δ v and σ v . Design for the control action is dedicated to compensate for

δ v since it is considered to be more important. For σ v being positive, the

control action is intended to compensate for δ v and σ v , and its amount is
the average of these factors.

Groups 4 & 5: These cases are equivalent to the opposite situations of Groups 3 and 2.

B. Shaping phase: fuzzy logic-based shaping guidance (FLSG)

Since position error dominates the final miss distance in terminal phase, a

feasible approach in this phase is to keep applying FLMG and linearly reduce the

value of the predicted velocity angle δ v and the seeker lock-on range Rlock to zero

so that the fuzzy inferenced command is gradually dominated by the position error.

Other operations in this phase remain invariant as those in midcourse phase.

10
C. Terminal phase: fuzzy logic-based terminal guidance (FLTG)
In those cases where a terminal seeker is locked onto the target and providing

reliable tracking data then the terminal guidance activates. During this phase, the

position error dominates the final miss distance. The influence resulting from the

velocity error becomes less important. The navigation term contributed by the heading

error angle generates effective acceleration command to reduce the final miss

distance.

For the linguistic variables, five fuzzy sets (LN, SN, ZE, SP, LP) have also been

considered where the corresponding membership functions are chosen as in Fig. 9. In

terminal phase, the aspect angle is supposed to near 180 degree. For an incoming

target the tracking error should be less than that in midcourse guidance. Thus highly

distinguishable fuzzy sets should be used to effectively increase the control sensitivity.

Also, to fully exert the maneuverability the control variable is presumed to cover a

wider operating domain than that in the midcourse course. A rule table analogous to

the standard PD-type fuzzy rule base is constructed in Table 2.

Referring to Table 2, the set of 25 guidance rules are divided into five groups:

Group 1: Both σ v and σ v are very small or closed to zero. This means that the

current flight path angle γ is closed to the LOS angle θ . The amount of
control action is also small and is intended to correct small deviations from
θv .

Group 2: σ v is positive or closed to zero. When σ v is also positive, it means that

the missile flying direction is moving away from the target incoming
direction. The control action is thus intended to significantly reverse this
trend.

Group 3: σ v is negative. When σ v is positive, it means that the missile flying

direction is moving towards the target incoming direction. The control action

11
 is thus intended to speed up the convergence of σ v and σ v to zero. When

σ v is small negative, the control action is set to zero to avoid overshooting

the response.

Groups 4 & 5: These cases are equivalent to opposite situation of Groups 3 and 2.

The max-min (Mamdani type) inference is used to generate the best possible

conclusion. This type of inference is computationally easy and effective; thus it is

appropriate for real-time control applications. The crisp guidance command is

calculated here using the center-of-gravity (COG) defuzzification. The criterion

provides the defuzzified output with better continuity. For a plant that is highly

sensitive to the command quality such as the missile autopilot, this will be appropriate

than other defuzzification methods. Structure of the fuzzy logic based guidance law is

shown as in Fig. 10.

Horizontal plane guidance

The design philosophy for the horizontal guidance law is quite similar to that of

the vertical plane guidance except that the control variable α is replaced by φ and

its major concern is to minimize the lateral position error. For these adopted variables

in midcourse and terminal phases, the corresponding membership functions are

chosen as those adopted in the vertical plane guidance. The only difference is that the

output variable of the horizontal plane guidance was chosen to have a wider range

than that of the vertical plane guidance.

V. Results and Analyses

A. Defensible volume
The allowable upper limit of final miss distance (MD) is supposed to be 5 m so as

to meet the mission requirement. The nominal physical parameters of the missile used

12
in the simulation studies are listed in Table 3. Also, we suppose that thrust of the

missile vanishes at 7.5 seconds. The tactical ballistic target with the incoming speed

of 1800 m/s and different reentry angles of 45, 60, and 75 degree were respectively

considered in the simulation studies. For the target, related parameters were given as

β =2440 kg , CtD 0 = 1.81 and M t =1200 kg. Suppose that the ground radar has
m2

detected the tactical ballistic target after reentry at a range of 50.0 km. The target

evasive accelerations aty ( t ) and ath ( t ) were generated by two first-order

Gauss-Markov processes to statistically represent 5g lateral accelerations [3]; for the

axial direction, atx ( t ) = 0 .

Fig. 11 shows three effectively defensible volumes for the missile against the

target with various reentry angles. Fig. 12 illustrates the corresponding sectional

drawing. It can be found that the defensible volumes decrease as the target with larger

reentry angles. The target with a larger reentry angle yields higher vertical speed. For

that case, the available time for the interceptor to successfully engage the target is

relatively short. As a result, the defensible volume is comparatively smaller. It is also

observed that the defensible volumes shrink at the highest and lowest altitudes. The

former is due to the fact that maneuverability of the missile reduces with increasing

flight altitudes (owing to the lower air density). On the other hand, if the interceptor

engages the target in a lower altitude, the total flight time might not be enough for it

to build up speed. This unavoidably degrades the engagement performance.

With regard to the guidance performance, related tracking responses, control

histories and missile velocity corresponding to point A of Fig. 12(a) were shown in

Fig. 13. It is found that the profiles of the velocity error angle δ and heading error

angle σ that the missile has successfully achieved the head-on condition before it

13
entered into the terminal phase. Note that, during this phase, the guidance commands

are not exerted until the missile velocity exceeds 1.2 Mach.

Requirements imposed on the terminal phase are usually more stringent because

all tracking errors have accumulated in a very short period. Thus, modification of the

traditional PD-type fuzzy rule table was adopted herein. From the transient responses

of the heading error angle, we can find that the PD rule table has the potential to yield

larger terminal speed and provide higher maneuverability. This shows that remarkable

engagement performance can be achieved, if expertise was appropriately incorporated

in the rule design.

Simulation results also show the convergence of heading error angle for the

conventional proportional navigation guidance (PNG) and the integrated fuzzy logic

guidance (IFLG) with the engagement corresponding to the point A in Fig. 12(a). The

PNG gains were adjusted such that the resulting MD is approximately equal to the

current design, i.e. around 3.7 m. From Fig. 14, it can be observed that the tracking

error of PNG is comparatively larger and converges slower than the proposed design.

Therefore, with the aim to gain the same engagement performance, PNG will

consume more control energy.

Figs. 15(a) and 15(b) characterize the defensible volume of the proposed fuzzy

guidance law against targets with velocities 1600 m/s and 2000 m/s and the reentry

angle 60 degree. It is clearly seen that the defensible volume with respect to the lower

speed target is far wider than the higher one.

B. Performance Sensitivity
Sensitivity of the fuzzy logic-based guidance design to variations in aerodynamic

coefficients was also examined. We intentionally changed the drag coefficient CmD

14
and the lift coefficient CmL to 50% and 150% of their nominal values respectively

and considered the defensible points B and C denoted in Fig. 12, where the incoming

target speed was 1800m/s and the initial positions of the incoming target ( xt , yt , zt )

were (20000, 24000, 36000) (m) (with the azimuth angle 50 degree) and (17000,

16000, 36000) (m) (with the azimuth angle 43 degree) respectively. Compared with

other defensible points, these points correspond to the scenarios where the targets

were relatively difficult to be engaged. Figs. 16 and 17 show the transient responses

of the heading error angle. It can be observed that performance robustness of the

fuzzy guidance design degrades in higher engagement altitudes. This is because

maneuverability of the missile decreases gradually in higher altitudes and thus

weakens its tracking capability. However, the fuzzy logic-based design still offers

satisfactory performance while there are large aerodynamic variations.

VI. Conclusions
An integrated fuzzy logic-based guidance scheme is developed to guide a

defense missile engaging the incoming ballistic target with very high speed on 3D

space. In this novel guidance design scheme, a fuzzy midcourse guidance law is first

designed with the aim to provide a superior counterattack attitude, to lessen control

energy consumption and to increase the terminal speed at lock-on point. For the

second part, a terminal fuzzy guidance law is developed with the aim to eliminate the

engagement error caused by the target evasive accelerations and to fulfill the

engagement missions. Extensive numerical verifications and examinations show that

the proposed design provides satisfactory performance from the viewpoint of energy

consumption, defense volume as well as sensitivity on the miss distance performance.

15
 Acknowledgment

This research was sponsored by National Science Council, Taiwan, R.O.C. under the

grant NSC 90-2213-E-035-008.

Nomenclature
CmD = drag coefficient
CmD 0 = zero drag coefficient
CmL = lift coefficient
CmLα = ∂C L / ∂α
D = drag force
g = gravity acceleration
L = lift force
m = missile mass
mt = target mass
T = thrust
vm = missile speed
vt = target speed
xm , ym , hm = coordinates of the missile
xt , yt , ht = coordinates of the target
α = angle-of-attack
γ = flight-path angle
γt = reentry flight-path angle
δv = velocity angle error in vertical plane
δh = velocity angle error in horizontal plane
φ = roll angle
θ = inertial line-of-sight angle
µ = induced drag coefficient
ρ = atmospheric density
σv = heading error angle in vertical plane
σh = heading error angle in horizontal plane
ψ = azimuth angle

References
[1] T., Kuroda, and F., Imado, “Advanced missile guidance system against very high
speed target,” AIAA Paper 88-4092, Aug. 1988.
[2] F., Imado, and T., Kuroda, “Optimal guidance system against a hypersonic
targets,” AIAA Paper 92-4531, Aug. 1992.

16
[3] C.F., Lin, Advanced Control Systems Design, Prentice-Hall, Englewood Cliffs, NJ,
1994.
[4] A. E., Bryson, and Y. C., Ho, Applied Optimal Control, 2nd ed., Blaisdell
Publishing Co., Waltham, Mass., 1975.
[5] C. F., Lin, and L. L., Tsai, “Analytical solution of optimum trajectory-shaping
guidance,” Journal of Guidance, Control, and Dynamics, Vol. 10, No. 1, pp.
61-66, 1987.
[6] B., Newman, “Strategic intercept midcourse guidance using modified zero effort
miss steering,” Journal of Guidance, Control, and Dynamics, Vol. 19, No. 1, pp.
107-112, 1996.
[7] S., Bezick, and W. S., Gray “Guidance of a homing missile via nonlinear
geometric control methods,” Journal of Guidance, Control, and Dynamics, Vol.
18, No. 3, pp. 441-448, 1995.
[8] C. L., Lin, and Y. Y., Chen, “Design of an advanced guidance law against high
speed attacking target,” Proceedings of the National Science Council, R.O.C.,
Part A, Vol. 23, No. 1, pp. 60-74, 1998.
[9] E. J., Song, and M. J., Tahk, “Three-dimensional midcourse guidance using neural
networks for interception of ballistic targets,” IEEE Transactions on Aerospace
and Electronic Systems, Vol. 38, No. 3, pp. 19-24, 2002.
[10] K., Mishra, I. G., Sarma, and K. N., Swamy, “Performance evaluation of two
fuzzy-logic-based homing guidance schemes,＂Journal of Guidance, Control,
and Dynamics, Vol. 17, No. 6, pp. 1389-1391, 1994.
[11] C. M., Lin, and Y. J., Mon, “Fuzzy-logic-based guidance law design for missile
systems,” in Proc. IEEE Control Applications Conf., pp. 421-426, 1999.
[12] B.S. Chen, Y.Y. Chen and C.L. Lin, “Nonlinear fuzzy H ∞ guidance law with
saturation of actuators against maneuvering targets,” IEEE Transactions on
Control Systems Technology, Vol. 10, No. 6, pp. 769-779, 2002.
[13] P., Zarchan, Tactical and Strategic Missile Guidance, 2nd ed., AIAA Inc.,
Washington, DC, 1994.

[14] A., Farina, B., Ristic, and D., Benvenuti, “Tracking a ballistic target: comparison
of several nonlinear filter,” IEEE Trans. Aerospace and Electronic Systems, Vol.
38, No. 3, pp. 854-867, 2002.
[15] G. W., Cherry, “A general explicit, optimizing guidance law for rocket-propellant
spacecraft,” AIAA Paper 64-638, Aug. 1964.

17
 List of Figures
Fig. 1 3-D intercept geometry
Fig. 2 Definitions of angle-of-attack α and rolling angle φ

Fig. 3 Target model

Fig. 4 Engagement scenario

Fig. 5 Definitions of σ and δ on the vertical plane

Fig. 6 Definitions of σ and δ on the horizontal plane

Fig. 7 Operational diagram of the vertical guidance law

Fig. 8 Membership functions used for the vertical plane guidance in the midcourse
phase (a) δ v , (b) σ v , (c) α

Fig. 9 Membership functions used for the vertical plane guidance in the terminal
phase (a) σ v , (b) σ v , (c) α

Fig. 10 Structure of the fuzzy guidance law

Fig. 11 Defensible volumes against the target with reentry angles (a) 45 degree, (b) 60
degree and (c) 75 degree

Fig. 12 Defensible areas against the target with reentry angles (a) 45 degree, (b) 60
degree and (c) 75 degree

Fig. 13 Control histories corresponding to the defensible point A; (a) midcourse phase
of the vertical plane, (b) terminal phase of the vertical plane, (c) midcourse
phase of the horizontal plane, (d) terminal phase of the horizontal plane, (e)
missile and target velocities
Fig. 14 Comparison of the heading error angle for the defensible point A

Fig. 15 Defensible volumes for the target descending at velocities (a) 1600 m/s and (b)
2000 m/s

Figure 16 Heading error angle to drag coefficient variation (a) and lift coefficient
variation (b) for the defensible point B

Figure 17 Heading error angle to drag coefficient variation (a) and lift coefficient
variation (d) for the defensible point C

18
 List of Tables
Table 1 Suggested rule table for the fuzzy midcourse guidance

Table 2 Suggested rule table for the fuzzy terminal guidance

Table 3 Nominal parameters of the defense missile

19
 at

H-Axis
vt T ( xt , yt , ht )
am vm
R

M ( xm , ym , hm ) θ

X-Axis
is
-Ax
Y

Fig. 1 3-D intercept geometry

H-Axis

φ T
α vm

γ X-Axis
is ψ
Ax
Y-

Fig. 2 Definitions of angle-of-attack α and rolling angle φ

Altitude

g
ht Drag

RV vtx
vty
H-Axis

γ2
γ1
−vth
vt

O
X-Axis xt
Downrange
A xis
Y-
yt Offrange

Fig. 3 Target model

20
 target

predicted-interception
point

lock-on point terminal phase

aspect angle
missile
H-Axis

-A xis shaping phase

Y
midcourse phase

laucher

X-
Ax
is

Fig. 4 Engagement scenario

21
 H
v vf
T ( xt ,ht )

δv

σv

θ
r rf v
X
M ( x ,h )

Fig. 5 Definitions of σ and δ on the vertical plane

M ( x, h)
X

ψ
v
ψf
δh
σh
θh
vf
T ( xt , yt )
Y

Fig. 6 Definitions of σ and δ on the horizontal plane

22
 Ballistic Target

Target kinematic data

Missile kinematic data

Counter Attack
Missile Airframe
Strategy

Information of relative
Actuator comtrol geometry
commands

Estimate the
Relative
Range

if R ≥ Rshap if Rshap ≥ R ≥ Rlock if R ≤ Rlock

Fuzzy Logic Based Fuzzy-Logic Based Fuzzy-Logic Based

Autopilot
Midcourse Shaping Terminal
Guidance Law Guidance Law Guidance Law
α fm = f m (δv , σ v ) α fs = fs (σ v ) α ft = ft (σ v , σv )

Acceleration commands Guidance commands

Guidance command
transformation

Fig. 7 Operational diagram of the vertical guidance law

23
 1
LN LP

ZE

SN SP
0
-20 -16 -12 -8 -4 0 4 8 12 16 20 (Degree)
δv
(a)
1
LN LP

ZE

SN SP
0
-20 -16 -12 -8 -4 0 4 8 12 16 20 (Degree)
σv
(b)
1
LN LP

ZE

SN SP
0
-30 -24 -18 -12 -6 0 6 12 18 24 30 (Degree)
α
(c)
Fig. 8 Membership functions used for the vertical plane guidance in the midcourse
phase (a) δ v , (b) σ v , (c) α

24
 1

LN SN ZE SP LP

0
-5 -4 -3 -2 -1 0 1 2 3 4 5 (Degree)
σv
(a)

LN SN ZE SP LP

0
-50 -40 -30 -20 -10 0 10 20 30 40 50 (Degree/s)

1 σv
(b)

LN LP

ZE

SN SP
0
-60 -48 -36 -24 -12 0 12 24 36 48 60 (Degree)
α
(c)
Fig. 9 Membership functions used for the vertical plane guidance in the terminal
phase (a) σ v , (b) σ v , (c) α

25
 guidance
rule bases
(Tables 1&2)

σ v , σ v , δ v
fuzzification max-min COG defuzzifi-
αf
(Figs. 8&9) inference cation

Fig. 10 Structure of the fuzzy guidance law

H (m)

16000 H (m)

14000
14000

12000
12000

10000
10000

8000 Y (m) 8000

10000
7000 Y (m)
5000
6000 3500
6000
0 0
X (m)
X (m)
2000 4000 6000 8000 10000 2000 4000 6000 8000

(a) (b)

H (m)

12000

10000

8000

6000

4500 Y (m)
4000
0
X (m)
2000 4000 6000

(c)
Fig. 11 Defensible volumes against the target with reentry angles (a) 45 degree, (b) 60
degree and (c) 75 degree

26
 (a)

(b)

(c)
Fig. 12 Defensible areas against the target with reentry angles (a) 45 degree, (b) 60
degree and (c) 75 degree

27
 30 angle of attack(deg)
angle of attack(deg) 50 heading error angle(deg)
velocity angle error(deg) heading error angular rate(deg/sec)
heading error angle(deg)
20 40

10 30
control histor y

20
0

10
-10

-20

-10

-30
0 2 4 6 8 10 11.5 12 12.5 13 13.5 14 14.5 15 15.5
time(sec)

(a) (b)

roll angle(deg) roll angle(deg)

velocity angle error(deg) 80 heading error angle(deg)
10
heading error angle(deg) heading error angular rate(deg/sec)

60
0
40

-10
control histor y

20

-20 0

-30 -20

-40
-40

-60
-50
0 2 4 6 8 10 11.5 12 12.5 13 13.5 14 14.5 15 15.5
time(sec)

(c) (d)
missile velocity(m/s)
1500 target velcoity(m/s)

1000

500

-500

-1000

-1500

0 2 4 6 8 10 12 14 16

(e)
Fig. 13 Control histories corresponding to the defensible point A; (a) midcourse phase
of the vertical plane, (b) terminal phase of the vertical plane, (c) midcourse phase of
the horizontal plane, (d) terminal phase of the horizontal plane, (e) missile and target
velocities

28
 IFLG
5 PNG

0
heading error angle(deg)

-5

-10

-15

-20

-25

0 2 4 6 8 10 12 14 16 18 20
time(sec)

Fig. 14 Comparison of the heading error angle for the defensible point A

29
 H (m)

16000

H (m)

14000 12000

12000
10000

10000
8000

8000
Y (m)
8000 6000
Y (m)
4000 5000
6000
0 0
X (m) X (m)
2000 4000 6000 8000 10000 2000 4000 6000

(a) (b)

Fig. 15 Defensible volumes for the target descending at velocities (a) 1600 m/s and (b)
2000 m/s

30
 nominal
drag coefficient*50%
5 drag coefficient*150%

0
hea ding error an gle(de g)

-5

-10

-15

-20

-25

0 5 10 15
time(sec)

(a)

nominal
5 lift coefficient*50%
lift coefficient*150%

0
hea ding error an gle(de g)

-5

-10

-15

-20

-25

0 5 10 15
time(sec)

(b)
Figure 16 Heading error angle to drag coefficient variation (a) and lift coefficient
variation (b) for the defensible point B

31
 10
nominal
drag coefficient*50%
drag coefficient*150%
5

0
hea ding error an gle(de g)

-5

-10

-15

-20

-25

-30

0 5 10 15
time(sec)

(a)
10
nominal
lift coefficient*50%
5 lift coefficient*150%

-5
hea ding error an gle(de g)

-10

-15

-20

-25

-30

0 5 10 15
time(sec)

(b)
Figure 17 Heading error angle to drag coefficient variation (a) and lift coefficient
variation (d) for the defensible point C

32
 Table 1 Suggested rule table for the fuzzy midcourse guidance
: group 1
σ
α fm
: group 2
LN SN ZE SP LP

LN SN LN LN LN LN : group 3
SN SN SN SN SN LN
δ ZE SP SP ZE SN SN : group 4
SP LP SP SP SP SP
LP LP LP LP LP SP : group 5

Table 2 Suggested rule table for the fuzzy terminal guidance

: group 1
σ
α ft
: group 2
LN SN ZE SP LP

LN LP LP LP LP LP : group 3
SN LP SP SP ZE ZE
σ ZE LP SP ZE SN LN : group 4
SP ZE ZE SN SN LN
LP LN LN LN LN LN : group 5

Table 3 Nominal parameters of the defense missile

Initial weight M m = 600 kg

Final weight M m = 280 kg
Max thrust T = 100630 N
Min thrust T =0N
Reference area sm = 0.086 m 2
Zero drag coefficient CmD 0 = 0.45 − (0.04 / 3) Mach
CmLα = 2.93 + 0.34008 Mach + 0.2615 Mach 2
Lift coefficient
+ 0.0108 Mach 3
Induced drag coefficient µ = 0.053

33
 Biography

IEEE Trans. on Fuzzy Systems

Chun-Liang Lin was born in Tainan, Taiwan, R.O.C. in 1958.
He received his Ph.D. in aeronautical and astronautical
engineering from the National Cheng Kung University, Taiwan
in 1991. He was an associate professor, and professor in the
Department of Automatic Control Engineering at Feng Chia
University, Taichung, Taiwan from 1995 to 2003. He is now a
professor of Department of Electrical Engineering at National
Chung Hsing University, Taichung, Taiwan. His research
interests include guidance and control, intelligent control and
robust control. Dr. Lin received the Distinguished Research
Award from National Science Council of Taiwan in 2000. Dr.
Lin is a senior member of IEEE.

Hao-Zhen Hung was born in Yunlin, Taiwan, R.O.C., in 1979.

He received the B.S. degree in electrical engineering and and
M.S. degree in automatic control engineering from the Feng
Chia University, Taichung, Taiwan, in 2001 and 2003,
respectively. His research interests are fuzzy logic control and
automotive control.

Yung-Yue Chen was born in Pingtong, Taiwan, R.O.C. He

received the B.S. and M.S. degrees from Feng Chia University,
Taiwan, R.O.C., in 1996 and 1998, respectively, and is
pursuing the Ph.D. degree in electrical engineering at National
Tsing Hua University, Hsinchu, Taiwan. His research interests
include guidance design, intelligent control and robust control.

34
Bor-Sen Chen (M’82–F’01) received the B.S. degree from
Tatung Institute of Technology, Taipei, R.O.C., in 1970, the
M.S. degree from National Central University, Taiwan, R.O.C.,
in 1973, and the Ph.D. degree from the University of Southern
California, Los Angeles, in 1982. From 1973 to 1987, he was a
Lecturer, Associate Professor, and Professor at Tatung Institute
of Technology. He is now a Professor at National Tsing Hua
University, Hsinchu, Taiwan, R.O.C. His current research
interests include control and signal processing.
Dr. Chen has received the Distinguished Research Award from
National Science Council of Taiwan four times. He is a
Research Fellow of the National Science Council and the Chair
of the Outstanding Scholarship Foundation.

35