You are on page 1of 14

JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICS

Vol. 37, No. 2, March–April 2014

Partial Integrated Guidance and Control for Missiles


with Three-Dimensional Impact Angle Constraints

Xianghua Wang∗ and Jinzhi Wang†


Peking University, 100871 Beijing, People’s Republic of China
DOI: 10.2514/1.60133
A partial integrated guidance and control design for an interception with terminal impact angle constraints in
three-dimensional space is presented in this paper. A three-dimensional nonlinear engagement dynamic is considered
here and full nonlinear six-degrees-of-freedom model of the interceptor with aerodynamic uncertainties has been
accounted for. The partial integrated guidance and control design contains a two-loop controller structure. The outer
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

loop, which generates directly the commanded body rates, is constructed, and then the inner loop is designed to track
the outer loop commands. The main feature of the partial integrated guidance and control design is that it accurately
satisfies terminal impact angle constraints in both azimuth and elevation, in addition to being capable of hitting the
target with high accuracy. Moreover, the aerodynamic uncertainties and the target acceleration, which are assumed
to be bounded, are considered by virtue of the proposed adaptive multiple input multiple output sliding mode control
method. The performance of the proposed scheme is investigated using nonlinear simulation studies. A comparison
among the proposed partial integrated guidance and control method, the conventional method where the guidance
and control loops are designed separately, and the integrated guidance and control method executed in a single loop is
presented to show the effectiveness of the proposed partial integrated guidance and control scheme. Finally, a
Monte Carlo study is conducted to test the robustness to aerodynamic uncertainties.

Nomenclature U, V, W = components of the missile velocity in the


atx , aty , atz = components of the target acceleration in the body frame, m∕s
inertial frame, m∕s2 Vm, Vt = velocity magnitudes of missile and target,
cx , cy , cz , = aerodynamic force and moment coefficients respectively, m∕s
cl , cm , cn Vm , Vt = velocity vectors of missile and target,
cla , clβ , cx0 , = aerodynamic derivatives respectively, m∕s
K, czα , cyβ xt , yt , zt = target position in the inertial frame, m
cnr , cnβ , cme , = aerodynamic derivatives xm , ym , zm = missile position in the inertial frame, m
cm20 , cm10 , cm00 α = angle of attack, deg
Fx , Fy , F z = aerodynamic forces, N β = sideslip angle, deg
g = gravity acceleration, m∕s2 γ = missile flight-path angle, deg
g1 , g2 = latax maneuvers of target in the yaw and γt = target flight-path angle, deg
pitch planes, m∕s2 Δcm , Δcn = aerodynamic uncertainties
Ixx , I yy , I zz , = moments of inertia, kg · m2 δe , δr , δa = actual elevator, rudder, and aileron deflec-
I xy , I yz , I zx tions, deg
kF , kM = constants determined by the missile geom- δec , δrc = elevator and rudder deflection commands, deg
etry, m2 and m3 , respectively θ, ϕ, ψ = pitch, roll, and yaw angles, deg
L, M, N = aerodynamic moments, N · m λE , λA = the elevation and azimuth angles of the line
 T
M, = missile and target, respectively of sight, respectively, deg
Mm = Mach number λE tf, λA tf = terminal values of elevation and azimuth
m = missile mass, kg angles of the line of sight, respectively, deg
o − xI − yI − zI = Cartesian inertial reference frame fixed on λE , λA = desired elevation and azimuth angles of the
the ground line of sight, respectively, deg
P, Q, R = the roll, pitch, and yaw rates, deg ∕s ξ = the damping ratio of the actuator model
Q , R = pitch and yaw rate commands, deg ∕s ρ = atmospheric density, kg∕m3
q1 , q2 , q3 , q4 = quaternions τ m , τt = angles between the line of sight and velocity
r = relative range along the line of sight, m vectors of the missile and target, respec-
rtf = terminal value of the relative range along the tively, deg
line of sight, m ϕref = desired roll angle, deg
tf = final time of the interception, s χ = missile heading angle, deg
χt = target heading angle, deg
ωn = the natural frequency of the actuator model, Hz

Received 27 August 2012; revision received 8 March 2013; accepted for


publication 24 June 2013; published online 14 February 2014. Copyright © I. Introduction
2013 by the American Institute of Aeronautics and Astronautics, Inc. All MANY kinds of missiles are hoped not only to get a minimum miss
rights reserved. Copies of this paper may be made for personal or internal use, distance but also to achieve desired terminal impact angles. Guidance
on condition that the copier pay the $10.00 per-copy fee to the Copyright with impact angle constraints has become a necessity in most modern
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include
the code 1533-3884/14 and $10.00 in correspondence with the CCC. warfare. The motivation for achieving particular terminal impact
*Ph.D. Student, Department of Mechanics and Aerospace Engineering, angles stems from the requirement of increasing the lethality of the
College of Engineering; xianghuaw@pku.edu.cn. warhead that the vehicle carries [1]. For example, for antiship or
† antitank missiles, the terminal impact angle is important for warhead
Professor, Department of Mechanics and Aerospace Engineering, College
of Engineering; jinzhiw@pku.edu.cn (Corresponding Author). effect. Especially, it is well known for missiles that vertical impact on
644
WANG AND WANG 645

the target can reduce the miss distance produced by navigation errors. conventional design [20]. In the PIGC design, the outer loop takes the
It can also be generalized for an unmanned aerial vehicle that has body rates as the virtual control inputs and directly yields the body
flight-path angle constraints depending on its missions [2]. rate commands. Then, the inner loop tracks the body rate commands
A variety of guidance laws have been proposed in the recent with the fin deflections as the control inputs.
literature to cater to impact angle constraints. Many of them, The targets have become more difficult to detect and track based on
however, are primarily two-dimensional (2-D) in nature with one the development of technology. Moreover, because aerodynamic
angle constraint; for example, [3–9]. Three-dimensional (3-D) coefficients are obtained through wind tunnel experiments,
guidance laws have also been proposed in the literature. To the uncertainties inevitably exist in system dynamics, which result in
knowledge of the authors, however, few of such guidance laws allow performance degradation. Hence, robustness about uncertainties of
for terminal impact angle constraints. Oza and Padhi [10] proposed a target and missile model has been a primary check point for modern
suboptimal guidance law achieving terminal impact angle constraints control design, and greater control of the engagement trajectory for
in 3-D space with a model predictive static programming method. A the terminal guidance phase is required. The sliding mode control
3-D variable structure guidance law is presented in [11] based on an methodology is an intuitive and simple robust control technique
adaptive model following control to cater for desired impact angles. addressing highly nonlinear systems with large modeling errors [21].
Sliding mode control is used in [12] to get a 3-D missile guidance law In this paper, an adaptive multiple input multiple output (MIMO)
with impact angle constraints. Oza and Padhi [1] proposed a sliding mode control method is proposed. The target acceleration and
the aerodynamic uncertainties are not required to be known with the
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

nonlinear suboptimal guidance law to intercept the ground target,


simultaneously satisfying 3-D impact angle constraints. proposed method, and only the bounds are necessary.
In this paper, we also design control and guidance laws to intercept The exact contribution of the paper over previous works is
a head-on maneuvering target for a skid-to-turn (STT) missile in 3-D concluded as follows. First, to the best of the knowledge of the
space. The terminal constraints are essentially two angles, namely, authors, only a few of the 3-D guidance laws take terminal impact
the elevation and azimuth angles, which need to be satisfied angles into account. In this paper, control laws are designed to cater to
simultaneously, in addition to being capable of hitting the target with two terminal impact angles in the 3-D space in addition to being
high accuracy. capable of hitting the target with high accuracy. Second, PIGC design
The classical missile guidance and control approaches treat the with a two-loop controller structure is used. Note that almost all
guidance and control systems as two separate processes. Typically, literature where an IGC frame is used to handle impact angle
the guidance system is considered as an outer loop, creating an constraint involves a 2-D interception with one terminal impact angle
acceleration to be tracked by an inner-loop autopilot. Although such considered. Hence, it should be the first attempt to achieve two impact
approaches have been proven to be efficient, the cooperation between angle constraints in the 3-D space with the IGC method. Third, an
subsystems is hardly smooth. As a result, guidance and control adaptive MIMO sliding mode control method is proposed, and the
circuits cannot work synergistically, and the performance of overall feature is that it is not required that the disturbances and uncertainties
missile system is not fully exploited [1]. To avoid these shortcomings, are known and only the bounds are necessary.
an integrated guidance and control (IGC) design that can maximize The structure of this paper is organized as follows. The nonlinear
the adjustability of missiles and improve the accuracy of hitting engagement dynamics in the 3-D space, the six degrees of freedom
targets will become dominant in perspective of the guidance and (six-DOF) nonlinear model with aerodynamic uncertainties for an STT
control design of missiles. missile and target dynamics are presented in the following section.
One can find a variety of ideas on IGC for missiles in the recent Problem formulation and design goal are presented in Sec. III. In
literature [13,14]; however, to the best of the knowledge of the Sec. IV, a novel adaptive MIMO sliding mode control method is
authors, few attempts are made on IGC design for missiles with introduced. Partial integrated guidance and control design for an STT
terminal impact angle constraints. Based on the LQR method, Yun missile is given in Sec. V. Simulations on the nonlinear missile model
and Ryoo [15] proposed an IGC design in the 2-D plane with are shown in Sec. VI and conclusions are drawn in the last section.
linearized missile longitudinal dynamics considered. In [16], an
integrated guidance and control system design for a homing missile
against a ground fixed target in the pitch plane was presented, where II. Model Description
an adaptive nonlinear control law was derived. With the assumption A. Engagement Dynamics in the 3-D Space
that the yaw and pitch channels are decoupled, a sliding mode control In this paper, a surface-to-air STT missile is considered to intercept
method was used to design an IGC system for linearized missile a head-on target. A schematic view of the 3-D interception geometry
dynamics in the pitch plane [1]. For a maneuvering target, an IGC is shown in Fig. 1, where o − xI − yI − zI is a Cartesian inertial
design in the vertical plane was developed in [17] based on the active reference frame fixed on the ground. M  and T denote the missile and
disturbance rejection control method. Shin et al. [2] examined the target, respectively; r is the relative range along the line of sight
integrated backstepping design of missile guidance and control to the (LOS); λE and λA are the elevation and azimuth angles of LOS,
engagement of maneuvering targets in a pitch plane with linearized respectively; xm ; ym ; zm  are coordinates of the missile in the o −
missile dynamics. Note that all literature mentioned here presented a xI − yI − zI frame; and the coordinates of target are xt ; yt ; zt . The
2-D IGC design with one terminal impact angle considered and, to engagement dynamics are given by
date, an IGC design with terminal impact angle constraints in the 3-D
space has not been attempted. Therefore, it is of interest to explore an
IGC design for a 3-D interception with two impact angle constraints.
Motivated by the preceding discussion, a partial integrated
guidance and control design similar to [18–20] is used here. For an
IGC framework executed in a single loop, quick maneuvers cause the
faster dynamics of the system to go unstable due to inherent time
scale separation between the faster and slower dynamics. On the
contrary, in the conventional design executed in three loops (the
guidance loop yields the acceleration commands, then an
intermediate loop transforms the acceleration commands to the
body rate ones, which are finally tracked in the innermost loop [18]),
the settling time of the response of different loops will not be able to
match with each other. This causes delay, which will affect the system
performance adversely. However, a partial integrated guidance and
control (PIGC) framework with a two-loop controller structure
overcomes the disadvantages of both the IGC design and the Fig. 1 Three-dimensional interception geometry.
646 WANG AND WANG

8 p
> where the aerodynamic forces and moments are given by
>
> r  xt − xm 2  yt − ym 2  zt − zm 2
>
< y − ym
λA  arctan t 8
xt − xm (1) >
> F  kF ρV 2m cx 
>
> zt − zm < x
>
: λE  arctan p
>  M  kM ρV 2m Cm
xt − xm 2  yt − ym 2 Fy  kF ρV 2m cy
>
> N  kM ρV 2m Cn
:
Fz  kF ρV 2m cz

B. Missile Dynamics
with
For the derivation of missile dynamics, the following assumptions
are made. 8
Assumption 1: Similar to [22], we assume that jatx j ≤ amax tx , >
> c  cx0  Kc2z
jaty j ≤ amax < x 
ty , jatz j ≤ atz where ati , i  x; y; z are components of the
max
cm  cm0 α; Mm   cme δe  Δcm
target acceleration in the inertial frame and amax ti is known > cy  cyβ β cn  cnβ β  cnr δr  Δcn
>
: c  c α;
(i  x; y; z). z zα
Assumption 2: Similar to [23–25], the assumption that the
aerodynamic uncertainties Δci are bounded is made (i  m; n) and
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

the bounds are known. where Δcm , Δcn denote the aerodynamic uncertainties following
Assumption 3: The variations of m, I xx , I yy , and I zz are assumption 2.
negligible [23]. The kinematic equations can be expressed with the use of a
Assumption 4: The missile has Y and Z symmetry, namely, quaternion as [28]
Iyy  I zz , Ixy  I yz  I zx  0 [23].
Assumption 5: The missile is roll stabilized (P  0) and the roll 0 1
q_ 1 0 1
angle ϕ remains constant [23,26]. B q_ 2 C 1 P
Remark 1: Assumptions 3, 4, and 5 are usually made in modeling B C  T1@ Q A (5)
@ q_ 3 A 2
STT missiles. R
Remark 2: STT missiles are roll-position stabilized to remain in the q_ 4
same roll orientation as at launch during the whole flight [26].
Furthermore, a roll stabilization autopilot can be designed easily, as is where
shown next.
The roll dynamics are generally described as [27] 0 1
−q2 −q3 −q4
 B q1 −q4 q3 C
I −I B C
P_  − zzIxx yy QR  kM ρV
2c
T1  @
−q2 A
m l
I xx (2) q4 q1
_ϕ ≈ P −q3 q2 q1

where cl  clβ β  cla δa . Take the roll control fin command as The relationship between the missile velocity components in the
inertial frame and ones in the body frame is given by [28]
 
I xx kM ρV 2m clβ β I zz − Iyy
δa  −  QR − k1 P − k0 ϕ  k0 ϕ ref 0 1 0 1
kM ρV 2m cla I xx I xx x_ m U
@ y_ m A  T 2 @ V A (6)
(3)
z_m W
where ki , i  0; 1 and ϕref are constants. The above roll stabilization
autopilot in Eq. (3) is well defined because cla is not zero. The roll where
dynamics in Eq. (2) with the roll position controller (3) are reduced to
0 1
q21  q22 − q23 − q24 2q2 q4 − q1 q3  2q2 q3  q1 q4 
ϕ  k1 ϕ_  k0 ϕ − ϕref   0 T 2  @ 2q2 q4  q1 q3  q21 − q22 − q23  q24 2q3 q4 − q1 q2  A
2q2 q3 − q1 q4  2q1 q2  q3 q4  q21 − q22  q23 − q24
Hence, appropriate choice of the controller gains ki , i  0; 1 can
assure that the roll angle is kept at the desired roll position ϕref ; that is,
P  0 and ϕ  ϕref . Differentiate Eq. (6) with respect to time once, then

Force equations of the missile are given by [27]



α_  Q − P cos α  R sin α tan β  mV m1cos β f− sin αFx  gx   cos αFz  gz g
β_  P sin α − R cos α  mV1 m f− cos α sin βFx  gx   cos βFy  gy  − sin α sin βFz  gz g

where gx  −mg sin θ, gy  mg cos θ sin ϕ and gz  mg cos θ cos ϕ.

0 1
Moment equations of the missile are presented as [27] 0 1 q_ 1 0 1
xm B q_ 2 C U_
@ ym A  2T 3 B C  T 2 @ V_ A (7)
@ q_ 3 A
( zm W_
Q_  − IxxI−I
yy
zz
PR  IMyy q_ 4
I −I (4)
R_  − yy xx PQ  N
I zz I zz where
WANG AND WANG 647

0 1
2q1 U − q3 V  q4 W 2q2 U  q4 V  q3 W −q1 V  q2 W q2 V  q 1 W
T 3  @ q3 U  2q1 V − q2 W q 4 U − q1 W q 1 U  q4 W q2 U  2q4 V  q3 W A
−q4 U  q2 V  2q1 W q 3 U  q1 V q2 U  q4 V  2q3 W −q1 U  q3 V

Substituting Eq. (5) into Eq. (7), III. Problem Formulation and Design Goal
0 1 0 1 0 1 0 1 0 1 A. Problem Formulation
x m P U_ P U_ We consider the engagement scenarios where a surface-to-air STT
B C 1 B C B C B C B C missile is intercepting a head-on target. From remark 2, the roll
B y m C  2T 3 T 1 B Q C  T 2 B V_ C  T 3 T 1 B Q C  T 2 B V_ C
@ A 2 @ A @ A @ A @ A channel is roll-position stabilized. What is more, because STT
zm R W_ R W_ missiles hardly use any roll motion [30], the missile under
0 10 1 0 1 consideration has only two controls: one along the pitch and the other
T 11 T 12 T 13 P U_ along the yaw axis [31]. The head-on interception engagement, as
B CB C B C
B C B C B
≜ @ T 21 T 22 T 23 A@ Q A  T 2 @ V_ C A (8) shown in Fig. 2, has the following definition, where τm and τt denote
the angles between the LOS and the missile and target velocity
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

T 31 T 32 T 33 R W_ vectors V m and V t , respectively. M  is the missile and T is the target.


Definition 1 [9]: Head on is an interception engagement, where a
From assumption 5 and remark 2, we have P  0. Then Eq. (8) can be v > 0 exists satisfying −π∕2 < τm < π∕2 and π∕2 ≤ τt ≤ 3π∕2 for
rewritten as t ∈ v; tf  and tf is the final time of the interception.
Remark 4: We assume that, during the head-on scenario, τm ≠  π2
0 1 0 1 0 1 because, in such cases, the interceptor has no maneuver capability
x m T 12 T 13   U_
@ y m A  @ T 22 Q perpendicular to the LOS [9].
T 23 A  T 2 @ V_ A (9)
R From definition 1, it can be seen that jλA − χj < π∕2 in the
zm T 32 T 33 W_ horizontal plane and jλE − γj < π∕2 in the vertical plane for a head-on
interception in 3-D space.
where
B. Design Goal
T 12  −2q1 q3  q2 q4 U  q22  q23 − q21 − q24 V Denote λA and λE as the desired constant azimuth and elevation
 2q1 q2 − q3 q4 W angles of LOS. Because it is possible to intercept a target when LOS
angle rates λ_ A and λ_ E are sustained as zero [32,33], the design goal of
T 13  2q2 q3 − q1 q4 U  2q1 q2  q3 q4 V this research is summarized as
 q21  q23 − q22 − q24 W 
λE → λE ; λ_ E → 0
(11)
T 22  q21  q22 − q23 − q24 U  2q2 q4 − q1 q3 V λA → λA ; λ_ A → 0
 2q1 q4  q2 q3 W Let eE  λE − λE , eA  λA − λA . Then, the design goal in Eq. (11) is
T 23  0 rewritten as

T 32  0 eE → 0; e_ E → 0
(12)
eA → 0; e_ A → 0
T 33  q23  q24 − q21 − q22 U  2q1 q3 − q2 q4 V
− 2q2 q3  q1 q4 W  −T 22 In the missile engagement, e_A → 0 and e_ E → 0 present the ideal
condition for zero miss distance and terminal impact angle
constraints can be satisfied by eA → 0 and eE → 0.
Remark 3: Because the integration process destroys the
satisfaction of the constraint equation q21  q22  q23  q24  1, a IV. Adaptive Multiple Input Multiple Output Sliding
normalization method is introduced in real time provided the Mode Control Method
unnormalized quaternion is Q  q1  iq2  jq3  kq4 and the
normalized quaternion is P  p1  ip2  jp3  kp4 , where Provided a sliding surface vector is a function vector of the system
q1
p1  NQ q2
, p2  NQ q3
, p3  NQ q4
and p4  NQ with N 2 Q  states x, presented as
2 2 2 2
q1  q2  q3  q4 . s  fx  0; s ∈ Rn (13)

C. Target Dynamics and the first-order derivative of Eq. (13) is given by


In this research, a maneuvering aircraft is chosen as the target. The
following point mass model is used for target modeling [29], s_  Fx; t  Hx; tu  Gx; tΔt (14)

0 1 0
−V t sin γ t
1 where u ∈ Rn is the control input; Δt  Δ1 ; Δ2 ; · · · ; Δm T ∈ Rm
x_t is assumed to be a bounded uncertainty vector, satisfying jΔj j ≤
B y_t C B −V t cos γ t cos χ t C
B C B C
B z_t C B −V t cos γ t sin χ t C
B CB C (10) τt
B V_ t C B 0 C →
B C B g1 C Vt
@ χ_ t A @ Vt
A
g2
γ_ t Vt
T

where xt , yt and zt are the target position in the inertial frame. V t is the Vm τm
velocity magnitude of the target; γ t and χ t are the flight-path angle
Reference Line
and the heading angle of the target, respectively; and g1 and g2 are M
latax maneuvers of target in the yaw and pitch planes [29]. Fig. 2 Head-on interception.
648 WANG AND WANG

Δmax
j and Δmaxj is known (j  1; · · · ; m); Fx; t ∈ Rn×1 , Consider the following Lyapunov function candidate:
Hx; t ∈ Rn×n , and Gx; t ∈ Rn×m are assumed to be known
function matrixes of x and t, and Hx; t is nonsingular. 1
V  sT s (19)
The control law is chosen as 2

u  Hx; t−1 Differentiating Eq. (19) with respect to time and using Eq. (18), we
0 1 obtain
2 3 2 3
js1 jρ sgns1  jG1j jsgns1 
B 6 7 6 7 C X
n m X
X n
B
B 6 .. 7 6 .. 7 C
C V_  sT s_ ≤ −k jsi jρ1  jGij jjsi jΔmax
j
^ max
− σj Δ j 
B 6 . 7 6 . 7 C
B 6 7 X m 6 7 C i1 j1 i1
6 ρ 7 6 7
×B
B −Fx; t − k6 jsi j sgnsi  7 − 6 jGij jsgnsi  7 σ Δ
^ max
j j C
C
_^ max P
B 6 7 j1 6 7 C From Δ j 0 > 0 and Δj
^ max  γ j  ni1 jGij jjsi j ≥ 0, Δ j t ≥
^ max
B 6 . 7 6 .. 7 C
B 6 .. 7 6 . 7 C max
^ j 0 > 0 (t ≥ 0) can be obtained (j  1; · · · ; m). Hence,
Δ
@ 4 5 4 5 A
Δmax − σj Δ^ max
j t ≤ Δj
max − σ Δ ^ max
j j 0. With Δj
max known, choose
jsn jρ sgnsn  jGnj jsgnsn  j
max max
σ j and Δj 0 satisfying σ j Δj 0 ≥ Δj , then Δmax
^ ^ max
j −
(15)
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

σj Δ ^ max
j t ≤ 0 (t ≥ 0) holds. Based on this analysis, we have V_ ≤
P
with the adaptive laws −k ni1 jsi jρ1 (t ≥ 0). According to Lemma 2, js1 jρ1  · · ·
X
n  jsn jρ1 2 ≥ js1 j2  · · · jsn j2 ρ1 holds; that is,
_^ max ^ max
Δ j  γ j jG ij jjs j
i ; Δ j 0 > 0; j  1; · · · ; m (16) X
n
i1 jsi jρ1 ≥ js1 j2  · · · jsn j2 ρ1∕2  2Vρ1∕2
where sgnsi  denotes the sign function of si ; k > 0, 0 < ρ < 1, i1
γ j > 0, and σ j > 0 (j  1; · · · ; m) are designed constants; Gij ,
i  1; · · · ; n, j  1; · · · ; m are elements of the matrix Gx; t; and Hence, V_ ≤ −k2ρ1∕2 V ρ1∕2 . From Lemma 1, the sliding surface
Δ^ max is the estimation of Δmax vector converges to zero in finite time ts given by
j j . Then, we have the following result.
Theorem 1: Suppose a sliding surface vector in Eq. (13) has the
first-order derivative of the form as in Eq. (14). Then, using the V01−ρ∕2
ts 
control law in Eq. (15) and the adaptive laws in Eq. (16), it can be k1 − ρ2ρ−1∕2
guaranteed that s will go to zero in finite time.
To prove Theorem 1, the following lemmas are introduced. This completes the proof.
Lemma 1: Assume that a continuous nonnegative function Vt Remark 5: If traditional sliding mode with no adaptation is applied,
satisfies the differential inequality Vt_ ≤ −τV η t, where τ > 0 and large switching gains would result, which may yield large control
0 < η < 1 are constants [34]. Then, for any given t0 , input and large chattering. The proposed method can avoid these
shortcomings by virtue of the adaptive laws. Moreover, γ j can be
V 1−η t ≤ V 1−η t0  − τ1 − ηt − t0 ; t0 ≤ t ≤ tr chosen small enough to suppress the quick growth of Δ ^ max
j .
and Vt  0, t ≥ tr with tr given by Remark 6: The adaptive control law in Eq. (15) is discontinuous for
the use of the signum function, which may lead to large chattering of
V 1−η t0  the control input and demand high rates for control surface
tr  t0 
τ1 − η deflections. To weaken the effects of the chattering, a saturation
function is often used to take the place of the signum function [36–
Lemma 2: Suppose a1 ; a2 ; · · · ; an and 0 < p < 2 are all positive 39]. Here, we may smoothe the signum function with satε s,
numbers. Then, the following inequality holds [35]: expressed as
8
a21  a22  · · · a2n p ≤ ap1  ap2  · · · apn 2 < 1; s > ε
satε s  s∕ε; jsj ≤ ε (20)
:
Proof of Theorem 1: From Eq. (14), −1; s < −ε
sT s_  sT Fx; t  sT Hx; tu  sT Gx; tΔt (17)
where ε is a small positive constant [40]. What is more, a second-
Substituting Eq. (15) into Eq. (17) yields order actuator model and rate saturation of control surface deflections
are also added in the simulation.
sT s_  sT Fx; t  sT Remark 7: If Eq. (20) is used to substitute the signum function, the
0 2 3 2 3 1 control law can only guarantee the bounded motion around the
js1 jρ sgns1  jG1j jsgns1  sliding surface [38,39]. Thus, the control law in Eq. (15) guides s to
B 6 7 6 7 C converge into a boundary layer jsj ≤ ε in finite time.
B 6 .. 7 6 .. 7 C
B 6 . 7 6 . 7 C
B 6 7 X
m 6 7 C
B 6 7 6 jG jsgns  7 ^ max C
×B
B−Fx; t − k6 jsi jρ sgnsi  7 − 6 ij C
i 7σ j Δj C V. PIGC with Impact Angle Constraints
B 6 7 j1 6 7 C
B 6 .. 7 6 .. 7 C In this section, we design the control inputs, the elevator deflection
B 6 . 7 6 7 C
@ 4 5 4 . 5 A δe , and the rudder deflection δr , which achieve the design goal in
jsn jρ sgnsn  jGnj jsgnsn  Eq. (12) in the presence of unknown target maneuvers represented by
target acceleration components atx , aty , and atz . For the inherent
X
n property of time scale separation between the translational and
 sT Gx; tΔt  −k jsi jρ1 rotational dynamics, PIGC with a two-loop controller structure is
i1 used. The outer loop is designed to drive eA , eE , e_A , and e_ E to zero
m X
X n  m X
X n 
using the pitch and yaw rates as virtual control inputs denoted as Q
− ^ max
jGij jjsi j σ j Δ j  G ij i Δj
s and R . The inner loop is constructed to make the actual body rates Q
j1 i1 j1 i1 and R track the body rate commands Q and R from the outer loop,
X
n m X
X n taking the elevator and rudder deflection commands δec and δrc as
≤ −k jsi jρ1  jGij jjsi jΔmax
j − σj Δ
^ max
j  (18) control inputs. The variables δec and δrc are the inputs of the actuator,
i1 j1 i1 which yields the actual fin deflections δe and δr being applied directly
WANG AND WANG 649

Body rate
commands
The 3-D Q* , R * eQ , eR δ ec , δ rc The actuator
δe ,δ r The missile
engagement The outer loop The inner loop
model model
model
Q, R
Actual
body rates

Fig. 3 Relationship between the outer and inner loops.

to the missile. The sketch of the relationship between the outer and 0 1 0 1
      x   x
inner loops is shown in Fig. 3. s_E f1  kE λ_ E g1 @ t A −g1 @ m A
s_1    y t  y m
s_A f2  kA λ_ A g2 −g2
zt zm
A. Outer Loop Design
(25)
Take the pitch and yaw rates as virtual control inputs denoted as Q
Substituting Eq. (9) into Eq. (25), we have
and R . The goal of the outer loop is to find the expressions for Q and
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

R such that eA , eE , e_ A , and e_ E converge to zero. 0 1 0 1


    xt   T 12 T 13  
Differentiate Eq. (1) with respect to time once and twice, f1  kE λ_ E g1 B C −g1 B C Q
respectively. Then, we have s_1   B yt C  B T 22 T 23 C
@ A @ A
f2  kA λ_ A g2 −g2 R
8 zt T 32 T 33
> r_  x_t − x_ m cos λE cos λA y_ t − y_m cos λE sin λA _zt − z_m sin λE 0 1 0 1
>
> y_ − y_ cos λA −x_ t − x_ m sin λA
<   U_   x t
λ_ A  t m −g1 B C Q B C
r cos λE T2B C  C1 B C
>
> z_t − z_m cos λE −y_ t − y_ m sin λE sin λA −x_ t − x_ m sin λE cos λA
 @ V_ A ≜ A1  B1 @ y t A (26)
>_
: λE  −g2 R
r W_ zt
(21)
where
0 1
and U_
   
f1  kE λ_ E −g1 B C
r − rλ_ 2E − rλ_ 2A cos2 λE  x t − xm  cos λE cos λA A1   B
T 2 @ V_ C
A
f2  kA λ_ A −g2
 y t − y m  cos λE sin λA  zt − zm  sin λE (22a) W_
0 1
  T 12 T 13
−g1 B C
B1  B T 22 T 23 C
rλ A cos λE  2r_λ_ A cos λE − 2rλ_ A λ_ E sin λE @ A
−g2
T 32 T 33
 y t − y m  cos λA − xt − x m  sin λA (22b)  
g1
C1 
g2
rλ E  2r_λ_ E  rλ_ 2A sin λE cos λE  zt − zm  cos λE − y t − ym 
Considering x t  atx , yt  aty , zt  atz , where atx , aty , atz are
× sin λE sin λA − x t − x m  sin λE cos λA (22c) acceleration components of the target in the inertial frame, Eq. (26)
can be rewritten as
It is implied that, from Eqs. (22b) and (22c),    atx 
Q
0 1 0 1 s_1  A1  B1  C1 aty (27)
      x   x R
λ E f1 g1 @ t A −g1 @ m A atz
  y t  y m (23)
λ A f2 g2 −g2 Because the pitch and yaw rates are considered as virtual control
zt zm
inputs, substitute Q , R for Q, R in Eq. (27), then
where    atx 
Q
s_1  A1  B1  C1 aty (28)
2r_ R
f1  − λ_ E − λ_ 2A sin λE cos λE ; atz
r
From assumption 1, atx , aty , and atz are bounded. Applying
r_
f2  2λ_ A λ_ E tan λE − 2 λ_ A Theorem 1 to the system in Eq. (28) implies
r
      
sin λE cos λA
Q jsE jρ1 sgnsE 
g1  − r − sin λErsin λA cos λE
r
; −1
 B1 −A1 − K1
R jsA jρ1 sgnsA 
   
λA cos λA jC1 11jsgnsE 
g2  − rsin cos λE r cos λE 0 − σ x a^ max
tx
jC1 21jsgnsA 
    
Construct a sliding surface vector as jC1 12jsgnsE  jC1 13jsgnsE 
− σ y a^ max
ty − σ ^
a max
z tz
    jC1 22jsgnsA  jC1 23jsgnsA 
sE e_ E  kE eE 8 max
s1   (24) > a_^  γ x jC1 11jjsE j  jC1 21jjsA j; a^ max
sA e_ A  kA eA >
< tx tx 0 > 0
_a^ max  γ jC 12jjs j  jC 22jjs j; a^ max 0 > 0 (29)
ty y 1 E 1 A ty
where kE , kA are designed positive constants. Differentiating Eq. (24) >
>
: _ max
with respect to time once yields a^ tz  γ z jC1 13jjsE j  jC1 23jjsA j; a^ tz 0 > 0
max
650 WANG AND WANG

where C1 ij, i  1; 2, j  1; 2; 3 are the elements of matrix C1 , and estimations. Denote σ^ and σ_^ as the estimations of σ and its first-order
K1 > 0, 0 < ρ1 < 1, γ i > 0, σ i > 0, and σ i a^ max ti 0 ≥ ati , i 
max derivative σ_ , respectively. The SMOD then takes the form
x; y; z are all designed constants. 8
Remark 8: U, _ V,
_ W_ are used in the control law in Eq. (29), ^  ρ0 signJ0 
< σt
which are acceleration components of the missile and measurable. U, χ_  bjσ − σ^ tj1∕2 signσ − σt
^ − ajJ0 j1∕2 signJ 0 
V, W are available from Eq. (6), where x_ m , y_ m , z_m are measurable. :
J0  χ  σ − σt^
With the virtual control inputs in Eq. (29), the outer loop can
achieve that eE , eA , e_ E , e_ A converge to zero, namely, the hit-to-kill
interception can be achieved with the desired terminal impact angles Provided the constants a > b > 0 and ρ0 > 0 are chosen
satisfied. appropriately, then σ^ → σ and σ_^ → σ_ are guaranteed in finite time.
Remark 10: The obtained fin deflection commands δic , i  e; r are
the inputs to the fin actuators that give the actual fin deflections δi ,
B. Inner Loop Design i  e; r via a second-order actuator model,
The goal of the inner loop is to design the elevator and rudder
      
deflection commands δec and δrc to make the actual pitch and yaw d δi 0 1 δi 0
rates Q and R track the commanded ones Q and R from the outer   δ (33)
dt δ_ i −ωn −2ξωn
2 δ_ i ω2n ic
loop. The variables δec and δrc are the inputs of the actuator, which
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

yields the actual fin deflections δe and δr being applied directly to the where ξ  0.7, ωn  20 Hz.
missile.
Let eQ  Q − Q , eR  R − R . Construct another sliding
surface vector as VI. Simulations
   R  Performance of the PIGC algorithm just developed is investigated
s e  kQ R eQ
s2  Q  Q (30) in this section via numerical simulations, using 3-D nonlinear
sR eR  kR eR
engagement dynamics and a full nonlinear six-DOF model of the
interceptor with aerodynamic uncertainties. First, we outline
where kQ and kR are designed positive constants.
simulation parameters and test scenario. Then, the effectiveness of
Considering the roll stabilization P  0, Eq. (4) can be rewritten as
the proposed method is verified using a sample run. Next, the
 _ proposed PIGC algorithm is compared with the conventional method
Q  IMyy
and IGC method. Finally, a Monte Carlo study is conducted to test the
R_  N I zz robustness of the proposed method.
Differentiating Eq. (30) with respect to time once and substituting
A. Simulation Scenario and Parameters
δec , δrc for δe, δr yields
In this section, a surface-to-air STT missile is considered in its
s_2  A2  B2 u  C2 Δ (31) terminal homing to engage with a target coming head on.
Using a simple curve fitting scheme based on the least square error
where criterion, the aerodynamic coefficient is approximated by a well-
defined affine function as follows [27], and the nominal values of
 kM ρV 2m  aerodynamic derivatives are listed in Table 1:
I yy cm0  kQ eQ − Q_ 
A2  kM ρV 2m
;
I zz cn β  kR eR − R_ 
β
cm0 α; Mm   cm20 α2  cm10 α1  Mm   cm00
 kM ρV 2m  The simulation parameters are target parameters and interceptor
I yy cme 0
B2  kM ρV 2m
; parameters. The target parameters are given by 1) speed: V t 
0 I zz cnr 600 m∕s, χ t 0  57.3 deg, and γ t 0  57.3 deg; and 2) position:

  Δcm    xt 0  2000 m, yt 0  2000 m, and zt 0  3000 m.
ΔQ I δ Interceptor parameters are given by 1) roll angle: ϕ  ϕref 
Δ  Δcyyn ; u  ec ; C2  kM ρV 2m
ΔR I
δrc 0 deg; 2) speed: U0  400 m∕s, V0  500 m∕s, and
zz
W0  600 m∕s; 3) mass: m  144 kg; 4) moments of inertia
From assumption 2, ΔQ and ΔR are bounded, namely around the body axes: I xx  1.6151 kg · m2 and I yy  I zz 
jΔi j ≤ Δmax ; i  Q; R. Because B2 is always invertible, applying 136.2648 kg · m2 ; 5) geometry constants: kF  0.0143 m2 and
i
Theorem 1 to the system in Eq. (31), the deflection commands are kM  0.0027 m3 ; 6) atmospheric density: ρ  0.2641 kg∕m3 ;
presented as 7) gravity acceleration: g  9.81 m∕s2 ; 8) fin deflection limits:
jδe j ≤ 30 deg and jδr j ≤ 30 deg; 9) fin rate limits: jδ_ e j ≤

δec
   ^ max
ΔQ σ Q sgnsQ 
 
jsQ jρ2 sgnsQ 
 100 deg ∕s and jδ_ r j ≤ 100 deg ∕s; 10) position: xm 0  ym 0 
 B−1
2 −A 2 − C2 − K 2
zm 0  0 m; and 11) aerodynamic uncertainties: Δcm t 
δrc ^ max
Δ R σ R sgnsR  jsR jρ2 sgnsR  Δcn t  0.05 sinπ4 t.
8
<Δ _^ max
Q  γ Q jsQ jkM ρV m ; ΔQ 0 > 0
2 ^ max
B. Effectiveness Verification
(32)
: _^ max ^ max In this section, two cases of desired terminal impact angles are
ΔR  γ R jsR jkM ρV 2m ; Δ R 0 > 0
considered to verify the effectiveness of the proposed method. For
where K 2 > 0, 0 < ρ2 < 1, γ i > 0, σ i > 0, and σ i Δ^ max
i ≥ Δmax
i , i
Q; R are all designed constants. Table 1 Nominal values of aerodynamic
Using the adaptive control law in Eq. (32), the tracking errors eQ , derivatives
eR go to zero. Connecting with the outer loop design, the PIGC
cx0 K czα cyβ
design for the interception of a surface-to-air missile against a head-
on target is completed. 003772 1.6763 0.6203 −0.21
Remark 9: Inspecting Eq. (32), it appears that Q_  and R_  are cnr cla cnβ clβ
−0.584 −0.127 0.08 0.116
needed for the implementation of control. However, the derivative Q_  cme cm20 cm10 cm00
and R_  cannot be directly measured in practice. In this work, a Sliding −0.675 −0.035 0.036617 5.3261  10−6
Mode Observer/Differentiator (SMOD) [41–43] is used to yield their
WANG AND WANG 651

Trajectories

45
45 3000 missile

Elevation angle λ deg


Azimuth angle λA deg
target

E
40 2000
40
1000
35
35
0

2000
30 30 2000
1000 1000
0 1 2 3 0 1 2 3 0 0
time sec time sec
Fig. 4 LOS angles and trajectories for case 1.
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

case 1, λA  30 deg and λE  30 deg. For case 2, λA  20 deg and ^ max ^ max
zt 0  100 m∕s , ΔQ 0  ΔR 0  100, K 1  K 2  1,
a^ max 2

λE  40 deg. The missile model is full nonlinear six-DOF model ρ1  ρ2  3∕5, kA  kE  100, kQ  10, and kR  40. To weaken
presented in Sec. II. The target model is Eq. (10) and it is assumed that the chattering of the control input, Eq. (20) is used in Eqs. (29) and
the target is performing a sinusoid maneuver; that is, g1  g2  (32) with ε  0.1.
55 sin 3t m∕s2 . The actuator model is Eq. (33). The engagement The terminal values of r, λA , and λE are given by
dynamics are Eq. (1). Initial conditions and simulation parameters are rtf  0.7683 m, λA tf  29.97 deg, and λE tf  29.9 deg
the same as presented in the previous part. The simulation curves for case 1 and rtf  0.6217 m, λA tf  20.05 deg, and
including azimuth angle λA , elevation angle λE and trajectories of λE tf  40.02 deg for case 2.
missile and target, pitch rate Q, yaw rate R, elevator deflection δe , As shown in the figures, for the two cases, the proposed
elevator rate δ_ e , rudder deflection δr , rudder rate δ_ r are shown in method achieves not only the hit-to-kill interception, but also the
Figs. 4 and 5 for case 1 and Figs. 6 and 7 for case 2, with σ x  40, desired impact angles and the control inputs are within the allowable
σ y  80, σ z  40, σ Q  0.01, σ R  0.01, a^ max xt 0  a ^ max
yt 0  bounds.

15 40 30
Elevator deflection δ deg
10
Pitch rate Q deg/s

30 20
Yaw rate R deg/s

5
20 10
0
10 0
−5

−10 0 −10

−15 −10 −20


0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

100
Rudder deflection δr deg

100
Elevator rate δ. deg/s

Rudder rate δ.r deg/s

50 20
50
e

0 0 0

−50
−50 −20
−100
−100
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 5 Body rates, fin deflections and rates for case 1.

Trajectories
50 48

45 missile
3000
Elevation angle λE deg
Azimuth angle λ deg

46 target
40
A

2000
35 44

30 1000
42
25 0
20 2000
40 2000
1000 1000
15
0 1 2 3 0 1 2 3 0 0
time sec time sec
Fig. 6 LOS angles and trajectories for case 2.
652 WANG AND WANG

15 50 20

Elevator deflection δe deg


40

Pitch rate Q deg/s


15

Yaw rate R deg/s


10
30
10
5 20
5
10
0
0 0

−5 −10 −5
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

100

Rudder deflection δr deg


100

Rudder rate δ.r deg/s


Elevator rate δ.e deg/s

50 20
50

0 0 0
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

−50 −50
−20
−100
−100
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 7 Body rates, fin deflections and rates for case 2.

C. Comparison of Interception Performance commands to the body rate ones, which are finally tracked in the
In this part, a comparison among PIGC, the conventional method, innermost loop [18]). The settling time of the response of different
and the IGC method is carried out. The conventional design is loops will not be able to match with each other, which may affect the
executed in three loops (the guidance loop yields the acceleration system performance adversely. The IGC design is executed in a
commands, then an intermediate loop transforms the acceleration single loop; that is, the fin deflections are directly deducted without

Trajectories
48 48
46 46 missile
target
44 44 3000
Elevation angle λE deg
Azimuth angle λA deg

42 42
2000
40 40
38 38 1000
36 36
0
34 34
2000
32 32
2000
30 30 1000
1000
28 28 0 0
0 1 2 3 0 1 2 3
time sec time sec
Fig. 8 LOS angles and trajectories for case 1 with conventional method.

800 800 20
Elevator deflection δe deg

600 600
Pitch rate Q deg/s

15
Yaw rate R deg/s

400 400
10
200 200
5
0 0

−200 −200 0

−400 −400 −5
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

100
100
Rudder defletion δ deg

30
Elevator rate δe deg/s

Rudder rate δr deg/s


r

50 20 50
.

10
0 0
0
−50
−50
−10
−100
−100 −20
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 9 Body rates, fin deflections and rates for case 1 with conventional method.
WANG AND WANG 653

Trajectories
46 48

44 46 missile
44 target
42 3000

Elevation angle λ deg


Azimuth angle λ deg
42
40

E
A
2000
40
38
38 1000
36
36
34 0
34
32 32 2000 2000
30 30 1000 1000
28 28
0 1 2 3 0 1 2 3 0 0
time sec time sec
Fig. 10 LOS angles and trajectories for case 1 with IGC method.
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

30 30

Elevator deflection δe deg


20 10 20
Pitch rate Q deg/s

Yaw rate R deg/s

10 10

0 5 0

−10 −10

−20 0 −20

−30 −30
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

25
Rudder deflection δr deg

100
Elevator rate δe deg/s

40

Rudder rate δ.r deg/s


20
50
15
.

20
0 10

5
−50 0
0
−100
−5 −20
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 11 Body rates, fin deflections and rates for case 1 with IGC method.

an introduction of a virtual control input. However, quick maneuvers is Eq. (33). The engagement dynamics are Eq. (1). The desired
may cause the faster dynamics of the system to become unstable due terminal impact angles are λA  30 deg and λE  30 deg.
to inherent time scale separation between the faster and slower Two cases of target maneuvering are considered:
dynamics. PIGC is the method adopted in this paper, which contains a 1) Case 1 is a sinusoid maneuvering target; that is, g1  g2 
two-loop controller structure. The outer loop yields directly the body 55 sin 3t m∕s2 .
rate commands to make the tracking error converge to zero and the 2) Case 2 is a square maneuvering target, namely g1 and g2 are
inner loop is designed to track the outer loop commands. Hence, squares with an amplitude of 55 m∕s2 , a period of 1 s, and a phase
the PIGC method cannot only preserve the inherent time scale delay of 0.1 s.
separation, but also helps to reduce the relative degree of the The simulation curves including azimuth angle λA , elevation angle
IGC frame. λE and trajectories of missile and target, pitch rate Q, yaw rate R,
In all methods, the missile model is a full nonlinear six-DOF model elevator deflection δe , elevator rate δ_ e , rudder deflection δr , rudder
presented in Sec. II. The target model is Eq. (10). The actuator model rate δ_ r for case 1 are shown in Figs. 4 and 5 with the PIGC method,

Trajectories

45
3000
45
missile
Elevation angle λ deg
Azimuth angle λ deg

2000 target
E

40
A

40
1000

35 0
35
2000

30 30 1000
2000
0 1000
0 1 2 3 0 1 2 3 0
time sec time sec
Fig. 12 LOS angles and trajectories for case 2 with PIGC method.
654 WANG AND WANG

Elevator deflection δe deg


15 40 20

Pitch rate Q deg/s

Yaw rate R deg/s


10 30
10
5
20
0 0
10
−5
−10
−10 0

−15 −10 −20


0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

Rudder deflection δr deg


100
Elevator rate δe deg/s

Rudder rate δr deg/s


100
50 20
50
.

.
0 0 0

−50 −50
−20
−100
−100
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 13 Body rates, fin deflections and rates for case 2 with PIGC method.

Trajectories

45 45 missile
3000
Elevation angle λ deg
Azimuth angle λ deg

target
E
A

2000
40 40
1000

35 35 0
2000

30 30 2000
1000
1000
0 1 2 3 0 1 2 3 0 0
time sec time sec
Fig. 14 LOS angles and trajectories for case 2 with conventional method.

Figs. 8 and 9 with the conventional method, and Figs. 10 and 11 with performing. The conventional method can also complete the design
the IGC method. For case 2, the simulation results are shown in goal in this simulation example. However, the conventional approach
Figs. 12 and 13 with the PIGC method, Figs. 14 and 15 with the separates the guidance loop from the control loop. Although such
conventional method, and Figs. 16 and 17 with the IGC method. approaches have been proved to be efficient, the cooperation between
Ternimal values rtf, λA tf, and λE tf with the PIGC method, subsystems is hardly smooth. As a result, the settling time of the
conventional method and IGC method are shown in Tables 2–4, response of different loops will not be able to match with each other,
respectively. and the performance of the overall missile system is not fully
It can be seen from the figures and tables that the proposed method exploited. Especially, when the time lag between two loops is large,
can achieve good performances whatever maneuvers the target is the interception may be not achieved. Moreover, it is clear from

800 800 20
Elevator deflection δe deg

600 600
Pitch rate Q deg/s

15
Yaw rate R deg/s

400 400
10
200 200
5
0 0

−200 −200 0

−400 −400 −5
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

100
100
Rudder defletion δr deg

30
Elevator rate δe deg/s

Rudder rate δr deg/s

50 20 50
.

10
0 0
0
−50
−50
−10
−100
−100 −20
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 15 Body rates, fin deflections and rates for case 2 with conventional method.
WANG AND WANG 655

Trajectories
46 48

44 46 missile
3000
44 target
42

Elevation angle λE deg


Azimuth angle λA deg
42 2000
40
40
38 1000
38
36
36
0
34
34
32 2000
32
30 30 1000 2000
1000
28 28
0 1 2 3 0 1 2 3 0 0
time sec time sec
Fig. 16 LOS angles and trajectories for case 2 with IGC method.
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

30 15 30

Elevator deflection δ deg


20 20
Yaw rate R deg/s

e
Pitch rate Q deg/s

10
10 10

0 5 0

−10 −10
0
−20 −20

−30 −5 −30
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec

60
Rudder deflection δ deg

100
Elevator rate δe deg/s

20 Rudder rate δ. deg/s 40


r

50
r
.

10 20
0
0
−50 0
−20
−100
−10 −40
0 1 2 3 0 1 2 3 0 1 2 3
time sec time sec time sec
Fig. 17 Body rates, fin deflections and rates for case 2 with IGC method.

Figs. 9 and 15 that larger chattering appears with the conventional values. The desired terminal impact angles are chosen as λA 
method and the pitch and yaw rates are larger than other methods. 30 deg and λE  30 deg. Assume that the target is performing a
Although the interception is also achieved with the IGC method, sinusoid maneuvering, namely g1  g2  55 sin 3t m∕s2. The
there are larger LOS angular errors in two cases of target maneuvers. terminal values of relative range and LOS angles rtf, λA tf, and
What is more, the second-order derivatives of LOS angles λ E and λ A λE tf for 100 runs are shown in Fig. 18. The means and standard
that are not available in practice are needed to construct an IGC deviations for rtf, λA tf, and λE tf are shown in Table 5.
controller. In summary, the proposed PIGC method is better than the It can be seen from Fig. 18 that, for 100 runs, rtf lies in [0 m,
conventional and IGC methods. 1 m], λA tf lies in [29.9 deg, 30.15 deg], and λA tf lies in [29.9 deg,
30.1 deg]. The values in Table 5 also reveal that the proposed adaptive
D. Monte Carlo Simulation
As shown in the preceding sections where the aerodynamic
derivatives of the missile take the nominal values listed in Table 1, the Table 3 Terminal values with the
PIGC design yields encouraging results. It is seldom the case that the conventional method
actual system will have exactly the same nominal model. To evaluate Case rtf, m λA tf, deg λE tf, deg
the robustness to the aerodynamic uncertainties, a Monte Carlo 1 0.3788 30.08 29.85
simulation study consisting of 100 sample runs is carried out. In the 2 0.2528 30.09 29.57
simulation, the random variables are the aerodynamic derivatives
listed in Table 1, deviating uniformly by 20% from their nominal

Table 2 Terminal values with the Table 4 Terminal values with the
PIGC method IGC method

Case rtf, m λA tf, deg λE tf, deg Case rtf, m λA tf, deg λE tf, deg
1 0.7683 29.97 29.9 1 0.3676 29.87 29.94
2 0.9403 30.05 30.05 2 0.7466 29.8 29.79
656 WANG AND WANG

1 30.15

0.8 30.1

λA(tf) deg
0.6 30.05
r(tf) m
0.4 30

0.2 29.95

0 29.9
0 20 40 60 80 100 0 20 40 60 80 100
times times

30.1

30.05
λ (tf) deg

30
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

29.95

29.9
0 20 40 60 80 100
times
Fig. 18 Results of rtf, λA tf, and λE tf for 100 runs.

Table 5 Means and standard deviations for rtf, [4] Kim, M., and Grider, K. V., “Terminal Guidance for Impact Attitude
λA tf, and λE tf Angle Constrained Flight Trajectories,” IEEE Transactions on Aerospace
and Electronic Systems, Vol. AES-9, No. 6, 1973, pp. 852–859.
rtf, m λA tf, deg λE tf, deg doi:10.1109/TAES.1973.309659
Mean 0.5141 30.0127 29.9903 [5] Kim, B. S., Lee, J. G., and Han, H. S., “Biased PNG Law for Impact with
Standard deviation 0.3046 0.0386 0.0299 Angular Constraint,” IEEE Transactions on Aerospace and Electronic
Systems, Vol. 34, No. 1, Jan. 1998, pp. 277–288.
[6] Ryoo, C. K., Cho, H., and Tahk, M. J., “Optimal Guidance Laws with
Terminal Impact Angle Constraint,” Journal of Guidance, Control, and
MIMO sliding mode control PIGC method can achieve good Dynamics, Vol. 28, No. 4, 2005, pp. 724–732.
robustness to the aerodynamic uncertainties. doi:10.2514/1.8392
[7] Rao, S., and Ghose, D., “Sliding Mode Control Based Terminal Impact
Angle Constrained Guidance Laws Using Dual Sliding Surfaces,”
VII. Conclusions Proceedings of the 12th IEEE Workshop on Variable Structure Systems,
IEEE Publ., Piscataway, NJ, 2012, pp. 325–330.
In this paper, a surface-to-air STT missile is considered to intercept [8] Erer, K. S., and Merttopcuoglu, O., “Indirect Impact-Angle-Control
a target coming head-on in the 3-D space. The six-DOF nonlinear Against Stationary Targets Using Biased Pure Proportional Navigation,”
STT missile model with aerodynamic uncertainties and the 3-D Journal of Guidance, Control, and Dynamics, Vol. 35, No. 2, 2012,
nonlinear engagement dynamics are used. Considering that PIGC can pp. 700–703.
preserve the inherent time scale separation between the slower and doi:10.2514/1.52105
faster dynamics, a controller with a two-loop structure is designed, [9] Shima, T., “Intercept-Angle Guidance,” Journal of Guidance, Control,
and Dynamics, Vol. 34, No. 2, March–April 2011, pp. 484–492.
which can accurately satisfy terminal impact angle constraints in both
doi:10.2514/1.51026
azimuth and elevation, in addition to being capable of hitting the [10] Oza, H. B., and Padhi, R., “Impact Angle Constrained Suboptimal
target accurately. Taking the target acceleration and aerodynamic Model Predictive Static Programming Guidance of Air to Ground
uncertainties as disturbances and assuming that these disturbances Missiles,” Journal of Guidance, Control, and Dynamics, Vol. 35, No. 1,
are bounded, the adaptive MIMO sliding mode control method is 2012, pp. 153–164.
proposed, which makes the PIGC design with better robustness. doi:10.2514/1.53647
Finally, the simulations on the nonlinear model are conducted and the [11] Sun, W. M., and Zheng, Z. Q., “3D Variable Structure Guidance Law
results show the effectiveness of the proposed PIGC design. Based on Adaptive Model-Following Control with Impact Angular
Constraints,” Proceedings of the 26th Chinese Control Conference,
IEEE Publ., Piscataway, NJ, 2007, pp. 61–66.
Acknowledgment [12] Gu, W. J., Yu, J. Y., and Zhang, R. C., “A Three-Dimensional Missile
Guidance Law with Angle Constraint Based on Sliding Mode Control,”
This work is supported by the National Nature Science Foundation Proceedings of the 2007 IEEE International Conference on Control and
of China under Grants 61074026 and 90916003. Automation, IEEE Publ., Piscataway, NJ, 2007, pp. 299–302.
[13] Idan, M., Shima, T., and Golan, O. M., “Integrated Sliding Mode
Autopilot-Guidance for Dual-Control Missiles,” Journal of Guidance,
References Control, and Dynamics, Vol. 30, No. 4, 2007, pp. 1081–1089.
[1] Oza, H. B., and Padhi, R., “A Nonlinear Suboptimal Guidance Law with doi:10.2514/1.24953
3D Impact Angle Constraints for Ground Targets,” AIAA Paper 2010- [14] Shkolnikov, I., Shtessel, Y., and Lianos, D., “Integrated Guidance-
8185, 2–5 Aug. 2010. Control System of a Homing Interceptor-Sliding Mode Approach,”
[2] Shin, H. S., Hwang, T. W., Tsourdos, A., White, B. A., and Tahk, M. J., AIAA Paper 2001-4218, 2001.
“Integrated Intercept Missile Guidance and Control with Terminal [15] Yun, J., and Ryoo, C. K., “Integrated Guidance and Control Law with
Angle Constraint,” Proceedings of the 26th International Congress of Impact Angle Constraint,” Proceedings of the 2011 11th International
the Aeronautical Sciences, IEEE Publ., Piscataway, NJ, 2008. Conference on Control, Automation and Systems, IEEE Publ.,
[3] Kumar, S. R., Rao, S., and Ghose, D., “Nonsingular Terminal Sliding Piscataway, NJ, 2011, pp. 1239–1243.
Mode Guidance and Control with Terminal Angle Constraints for [16] Guo, J. G., and Zhou, J., “Integrated Guidance and Control of Homing
Nonmaneuvering Targets,” Proceedings of the 12th IEEE Workshop on Missile with Impact Angular Constraint,” Proceedings of the 2010
Variable Structure Systems, IEEE Publ., Piscataway, NJ, 2012, pp. 291– International Conference on Measuring Technology and Mechatronics
296. Automation, IEEE Publ., Piscataway, NJ, 2010, pp. 480–483.
WANG AND WANG 657

[17] Zhao, C. Z., and Huang, Y., “ADRC Based Integrated Guidance and [31] Vaddi, S. S., Menon, P. K., and Ohlmeyer, E. J., “Numerical SDRE
Control Scheme for the Interception of Maneuvering Targets with Approach for Missile Integrated Guidance-Control,” AIAA Paper
Desired LOS Angle,” Proceedings of the 29th Chinese Control 2007-6672, 2007.
Conference, IEEE Publ., Piscataway, NJ, 2010, pp. 6192–6196. [32] Wu, P., and Yang, M., “Integrated Guidance and Control Design for
[18] Das, P. G., Chawla, C., and Padhi, R., “Robust Partial Integrated Guidance Missile with Terminal Impact Angle Constraint Based on Sliding Mode
and Control of Interceptors in Terminal Phase,” AIAA Paper 2009-6275, Control,” Journal of Systems Engineering and Electronics, Vol. 21,
2009. No. 4, 2010, pp. 623–628.
[19] Shtessel, Y. B., Shkolnikov, I. A., and Levant, A., “Guidance and [33] Gu, W., Zhang, R., and Yu, J., “AThree-Dimensional Missile Guidance
Control of Missile Interceptor Using Second-Order Sliding modes,” Law with Angle Constraint Based on Sliding Mode Control,”
IEEE Transactions on Aerospace and Electronic Systems, Vol. 45, Proceedings of the 2007 IEEE International Conference on Control and
No. 1, 2009, pp. 110–124. Automation, IEEE Publ., Piscataway, NJ, 2007, pp. 299–302.
doi:10.1109/TAES.2009.4805267 [34] Wei, X. J., and Lei, G., “Composite Disturbance-Observer-Based
[20] Chawla, C., and Padhi, R., “Partially Integrated Guidance and Control of Control and Terminal Sliding Mode Control for Non-Linear Systems
UAVs for Reactive Collision Avoidance,” TR AOARD-104014, with Disturbances,” International Journal of Control, Vol. 82, No. 6,
Department of Aerospace Engineering, Indian Institute of Science, 2009, pp. 1082–1098.
Bangalore, India, 2011. doi:10.1080/00207170802455339
[21] Shima, T., Idan, M., and Golan, O. M., “Sliding Mode Control for [35] Yu, S., Yu, X., Shirinzadeh, B., and Man, Z., “Continuous Finite-Time
Integrated Missile Autopilot Guidance,” Journal of Guidance, Control, Control for Robotic Manipulators with Terminal Sliding Mode,”
Downloaded by LOCKHEED MARTIN AERO FORT WORTH on December 1, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.60133

and Dynamics, Vol. 29, No. 2, 2006, pp. 250–260. Automatica, Vol. 41, No. 11, 2005, pp. 1957–1964.
doi:10.2514/1.14951 doi:10.1016/j.automatica.2005.07.001
[22] Harl, N., Balakrishnan, S. N., and Phillips, C., “Sliding Mode Integrated [36] Chen, S. Y., and Lin, F. J., “Robust Nonsingular Terminal Sliding-Mode
Missile Guidance and Control,” AIAA Paper 2010-7741, 2– Control for Nonlinear Magnetic Bearing System,” IEEE Transactions
5 Aug. 2010. on Control Systems Technology, Vol. 19, No. 3, 2011, pp. 636–643.
[23] Chwa, D., and Choi, J. Y., “New Parametric Affine Modeling and doi:10.1109/TCST.2010.2050484
Control for Skid-to-Turn Missiles,” IEEE Transactions on Control [37] Neila, M. B. R., and Tarak, D., “Adaptive Terminal Sliding Mode
Systems Technology, Vol. 9, No. 2, March 2001, pp. 335–347. Control for Rigid Robotic Manipulators,” International Journal of
[24] Hou, M. Z., and Duan, G. R., “Integrated Guidance and Control of Automation and Computing, Vol. 8, No. 2, 2011, pp. 215–220.
Homing Missiles Against Ground Fixed Targets,” Chinese Journal of doi:10.1007/s11633-011-0576-2
Aeronautics, Vol. 21, No. 2, 2008, pp. 162–168. [38] Man, Z. H., Paplinski, A. P., and Wu, H. R., “A Robust MIMO Terminal
[25] Wei, Y. Y., Hou, M. Z., and Duan, G. R., “Adaptive Multiple Sliding Sliding Mode Control Scheme for Rigid Robotic Manipulators,” IEEE
Surface Control for Integrated Missile Guidance and Autopilot with Transactions on Automatic Control, Vol. 39, No. 12, 1994, pp. 2464–
Terminal Angular Constraint,” Proceedings of the 29th Chinese Control 2469.
Conference, IEEE Publ., Piscataway, NJ, 2010, pp. 2162–2166. [39] Zhu, Z., Xia, Y., and Fu, M., “Adaptive Sliding Mode Control for
[26] Lee, J., and Ha, I. J., “Autopilot Design for Highly Maneuvering STT Attitude Stabilization with Actuator Saturation,” IEEE Transactions on
Misssiles via Singular Perturbation-Like Technique,” IEEE Transactions Industrial Electronics, Vol. 58, No. 10, 2011, pp. 4898–4907.
on Control Systems Technology, Vol. 7, No. 5, 1999, pp. 527–541. doi:10.1109/TIE.2011.2107719
[27] Huang, J., and Lin, C. F., “Application of Sliding Mode Control to Bank- [40] Zhou, D., and Sun, S., “Guidance Laws with Finite Time Convergence,”
to-Turn Missile Systems,” Proceedings of the First IEEE Regional Journal of Guidance, Control, and Dynamics, Vol. 32, No. 6, 2009,
Conference on Aerospace Control Systems, IEEE Publ., Piscataway, NJ, pp. 1838–1846.
1993, pp. 569–573. doi:10.2514/1.42976
[28] Qin, L., Zhang, W., and Fan, F., “Quaternion Method for Kinematics [41] Shtessel, Y. B., and Tournes, C. H., “Integrated Higher-Order Sliding
Modeling of High Attack-Angle Flying Carrier,” Journal of North Mode Guidance and Autopilot for Dual-Control Missiles,” Journal of
University of China (Natural Science Edition), Vol. 27, No. 3, 2006, Guidance, Control, and Dynamics, Vol. 32, No. 1, 2009, pp. 79–94.
pp. 276–279. doi:10.2514/1.36961
[29] Dwivedi, P. N., Tiwari, S. N., Bhattacharya, A., and Padhi, R., “A ZEM [42] Tournes, C. H., and Shtessel, Y. B., “Integrated Guidance and Autopilot
Based Effective Integrated Estimation and Guidance of Interceptor in for Dual Controlled Missiles Using Higher Order Sliding Mode
Terminal Phase,” AIAA Paper 2010-8318, 2–5 Aug. 2010. Controllers and Observers,” AIAA Paper 2008-7433, 2008.
[30] Yamasaki, T., Balakrishnan, S. N., and Takano, H., “Integrated [43] Zhurbal, A., and Idan, M., “Effect of Estimation on the Performance of
Guidance and Autopilot Design for a Chasing UAV via High-Order an Integrated Missile Guidance and Control System,” AIAA Paper
Sliding Modes,” Journal of the Franklin Institute, Vol. 349, No. 2, 2012, 2008-7458, 2008.
pp. 531–558.
doi:10.1016/j.jfranklin.2011.08.004

You might also like