JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICS

Vol. 37, No. 4, July–August 2014

Nonsingular Terminal Sliding Mode Guidance

with Impact Angle Constraints

Shashi Ranjan Kumar,∗ Sachit Rao,† and Debasish Ghose‡

Indian Institute of Science, Bangalore 560012, India
DOI: 10.2514/1.62737
Guidance laws based on a conventional sliding mode ensures only asymptotic convergence. However, convergence
to the desired impact angle within a finite time is important in most practical guidance applications. These finite time
convergent guidance laws suffer from singularity leading to control saturation. In this paper, guidance laws to
intercept targets at a desired impact angle, from any initial heading angle, without exhibiting any singularity, are
presented. The desired impact angle, which is defined in terms of a desired line-of-sight angle, is achieved in finite time
by selecting the interceptor’s lateral acceleration to enforce nonsingular terminal sliding mode on a switching surface
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

designed using nonlinear engagement dynamics. Numerical simulation results are presented to validate the proposed
guidance laws for different initial engagement geometries and impact angles. Although the guidance laws are designed
for constant speed interceptors, its robustness against the time-varying speed of interceptors is also evaluated through
extensive simulation results.

I. Introduction for guided bombs against stationary targets. The guidance laws
proposed in [7,8] are limited to nonmaneuvering target cases. In this
I N MODERN warfare, intelligent targets are becoming serious
threats and, to counteract them, there is a need to develop guidance
laws based on modern control techniques. In recent years, the sliding
paper, guidance laws having a finite time convergence property along
with the ability to intercept even maneuvering targets at a
mode control, which is a nonlinear control technique, has become prespecified impact angle, based on NTSMC, are proposed.
popular due to its simple design and robustness against uncertainties In the guidance literature, there exist other SMC-based guidance
[1]. It has also been applied to several guidance applications like laws [9–14] that address the impact angle control problem. In [9], an
impact angle control problems. In guidance applications, like air-to- impact angle control guidance law that ensures the interception from
air combat situations, the assurance of interception with a pre- all the engagement scenarios is proposed for nonmaneuvering targets
specified impact angle within a finite time are very important. A using nonlinear engagement dynamics and dual switching surfaces.
variant of the sliding mode control (SMC), known as terminal SMC The switching between these surfaces is based on the conditions
(TSMC), which ensures finite time convergence, has been used in necessary to ensure interception. In [10,11], an SMC-based guidance
missile [2] and unmanned aerial vehicle [3] guidance applications. law to intercept maneuvering targets in head-on, tail-chase, and head-
Guidance laws, which ensure finite time convergence to the pre- pursuit engagement scenarios, which enables imposing a pre-
specified impact angle, based on the terminal sliding mode control, determined interception angle relative to the target’s flight-path
for stationary, constant velocity, and maneuvering targets, are angle, is presented. Analytical conditions for the existence of these
proposed in [4]. Another impact angle control guidance law, based on different engagement geometries are also discussed. In [12], a
the finite time convergence stability theory and sliding mode control, guidance law is derived using a combination of zero-effort miss and
for slow moving targets, is proposed in [5]. Although the guidance impact angle error, for which an estimation of time-to-go tgo is
laws proposed in [4,5] give finite time error convergence, that is, the required. In [13], an integrated guidance and control design using an
interceptor can align itself with the impact angle frame within a finite adaptive SMC algorithm to intercept ground-fixed targets has been
time, both have the drawback of a possible singularity when the error proposed. A nonlinear interceptor model in the pitch plane is used to
being controlled becomes very small, thus leading to control evaluate the performance of the designed guidance law; the
saturation. The same kind of singularity can also be observed in a achievable impact angle range is limited. In [14], guidance laws
guidance law that focuses only on interception with finite time based on high-performance SMC, for stationary and slow moving
convergence [6], where the guidance law has been derived for targets, which increases the target observability along with intercep-
maneuvering targets using finite time stability of nonlinear systems tion at the desired impact angle, are proposed. To derive these
and the Lyapunov scalar differential inequality. To avoid the guidance laws, a nonlinear sliding surface is chosen that varies the
singularity while preserving the finite time convergence property, the damping of the guidance loop system from an initial low to a final
impact angle control guidance laws, based on nonsingular terminal high value during the engagement so that both the objectives can be
sliding mode control (NTSMC) for stationary and constant velocity achieved.
targets, was first proposed in [7]. Similar to [7], a finite time Non-SMC-based guidance laws, for instance, those derived using
convergent impact angle control guidance law is also proposed in [8] optimal control theory, have also been presented in the literature
[15–22]. Because these guidance laws are derived using linearized
engagement dynamics, their performance may degrade under large
Received 9 April 2013; revision received 10 July 2013; accepted for
publication 10 July 2013; published online 14 February 2014. Copyright © heading errors. This is because the linearizations are no longer valid
2013 by the American Institute of Aeronautics and Astronautics, Inc. All under such situations. These guidance laws are derived as the solution
rights reserved. Copies of this paper may be made for personal or internal use, to a linear quadratic optimal control formulation. The guidance laws,
on condition that the copier pay the $10.00 per-copy fee to the Copyright presented in [15–22], except those in [18,21], are designed for
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include stationary targets, whereas in [18,21] they were also designed for
the code 1533-3884/14 and $10.00 in correspondence with the CCC. moving, but nonmaneuvering and maneuvering targets, respectively.
*Doctoral Student, Guidance, Control, and Decision Systems Laboratory, In [23], a generalized impact angle control guidance law is presented
Department of Aerospace Engineering; [email protected].
†
Formerly Postdoctoral Associate, Guidance, Control, and Decision
for stationary targets. In [22,23], analytical solutions of the guidance
Systems Laboratory, Department of Aerospace Engineering; [email protected] laws, expressed as a combination of polynomial, logarithmic, and
.ernet.in. infinite power series functions, are presented to examine the effect of
‡ system lag on their performance. These analytical solutions give
Professor, Guidance, Control, and Decision Systems Laboratory,
Department of Aerospace Engineering; [email protected]. insight into the behavior of the interceptor near the target. In [24], a
1114
 KUMAR, RAO, AND GHOSE 1115

guidance law for stationary targets is derived as the solution to a validate the performance of the proposed guidance laws. Finally, in
nonlinear regulator problem using the state-dependent Riccati Sec. VI, conclusions and possible future work are discussed.
equation technique, where the time-varying state weighting matrix is
assumed to be a function of time-to-go tgo . A backstepping-based
guidance law that requires approximations of tgo , impact angle, and
line-of-sight (LOS) rate has been designed in [25]. A differential
II. Problem Formulation
game-based guidance law is presented in [26], which demands a Consider a planar engagement between an interceptor and a target
considerably higher interceptor lateral acceleration advantage over as shown in Fig. 1a. The interceptor and the target are assumed to
the target. The guidance laws in [25,26] are designed for maneu- have velocities V M and V T , their flight-path angles are γ M and γ T , and
vering targets. A time-optimal guidance law, which aims to minimize their lateral accelerations are denoted by aM and aT , respectively.
the flight time, to intercept stationary targets in the presence of The relative distance between the target and the interceptor is given
bounded lateral acceleration, and a missile autopilot with first-order by r and the LOS angle by θ. Assuming that V M and V T are constants,
lag, is proposed in [27]. In [28], guidance laws, as a polynomial the kinematic engagement equations are
function of tgo whose coefficients are obtained to satisfy the terminal
constraints, are proposed for stationary or slow moving targets. r_
V r
V T cos θT − V M cos θM (1a)
Another guidance law, which increases the target observability from
angle-only measurements along with controlling the impact angle,
based on time-to-go polynomial guidance law, is proposed in [29]. In rθ_
V θ
V T sin θT − V M sin θM (1b)
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

the guidance laws proposed in [15–29], except [27], the estimation of

tgo plays a crucial role. However, these guidance laws require the
estimation of tgo , which is a formidable challenge. A cubic spline- aM aT
based guidance law for stationary targets is presented in [30], where it γ_ M
; γ_ T
(1c)
VM VT
is derived using the inverse method from a cubic spline curve-based
trajectory, which satisfies the terminal constraints. In [31–35], where θT
γ T − θ and θM
γ M − θ.
guidance laws, which intercept nonmaneuvering targets at a pre- Achieving a desired impact angle is an additional objective along
specified impact angle, are presented. These guidance laws are the with the usual requirement of interception. As shown in Fig. 1b, the
variants of proportional navigation guidance (PNG), known as biased impact angle, denoted by θimp, is defined as the angle between the
PNG, in which a biased term is added to the PNG law to satisfy impact velocity vectors of the interceptor and target at the time of
angle constraints. To derive these bias terms, nonlinear engagement interception and is given by
dynamics has been used in [31–35], except in [32], where a small
angle assumption has been made. In [35], a shaping method for the θimp
γ Tf − γ Mf (2)
bias term is also proposed, which enables the interceptor to take into
account the maximum lateral acceleration capability and the limita-
where γ Tf and γ Mf are the flight-path angles of the target and the
tion on the seeker field of view. Other variants of PNG law are found
interceptor, respectively, at the time of interception. The guidance
in [36–38]. In [36,37], two-stage PNG laws with different navigation
laws for intercepting the target at the desired impact angle, within a
constants to intercept nonmaneuvering targets are designed using
finite time, are designed under the assumption of V M > V T . With this
nonlinear engagement dynamics. These laws ensure that the lateral
assumption, target-to-interceptor speed ratio ν satisfies
acceleration remains bounded even in the terminal phase. In [38], an
interception angle control guidance, expressed as a combination of a
VT
PNG command and a feedback of the interception angle error, is ν
<1 (3)
proposed for stationary targets. VM
In this paper, a guidance law, which ensures finite time con-
vergence to the impact angle constraint guidance problem, is first It is possible to establish a relationship between the impact angle and
designed for a general maneuvering target, and then stationary and the LOS angle. It can be shown that, for most engagement scenarios, a
constant velocity targets are treated as special cases. Some pre- unique LOS angle exists for a particular impact angle. This relation is
liminary ideas that lead to the complete development of the guidance established under the assumption that the missile and target are on a
law presented here were discussed in [7]. The paper is organized as collision course as shown in Fig. 1c. When the interceptor and target
follows: In Sec. II, the problem statements and the relation between are on the collision course, the condition
the given impact angle and a corresponding unique final LOS angle
are described. In Sec. III, the basics of nonsingular terminal sliding V M sinγ Mf − θF 
V T sinγ Tf − θF  (4)
mode are described. In Sec. IV, the kinematics-based guidance laws
are designed for stationary, constant velocity, and maneuvering holds, which, from Eq. (1b), corresponds to rθ_
0. Substituting
targets. In Sec. V, numerical simulation results are presented to Eq. (4) into Eq. (6), the relation

Fig. 1 Interceptor and target: a) engagement geometry, b) impact angle, and c) collision course.
 1116 KUMAR, RAO, AND GHOSE


sin θimp To illustrate the design of these algorithms, consider the double
θF
γ Tf − tan−1 (5) integrator system x

u, x, u ∈ R. In the terminal SMC algorithm
cos θimp − ν
[40], to design u, which achieves x
0 within finite time, the
can be derived. From this relation, it can be seen that, when function
θimp
nπ, n
1; 2; : : : , the LOS angle θF is independent of ν.
These cases represent the head-on and tail-chase scenarios. For the s
x_  βjxjα signx; β > 0; 0<α<1 (8)
stationary target case, the impact angle is defined with respect to a
reference. Without loss of generality, γ Tf
0 deg can serve as this is chosen as the switching surface on which the control u is
reference. discontinuous in the form
For the relation between θF and θimp to be meaningful, it is
essential that they have a one-to-one correspondence. This aspect has u
−M signs; M>0 (9)
been discussed in detail in [4].
Now, to select M, evaluate s_
u  αβjxjα−1 x_ and choose
M > αβjxjα−1 jxj,
_ which ensures that the sliding mode occurs within
finite time. Because 0 < α < 1, the bound on the gain M increases to
III. Nonsingular Terminal Sliding Mode Control
very large values when the state x achieves a very small value and thus
The problem of intercepting the target at a prespecified impact generates a singularity. This part will be further elaborated
angle within a finite time is addressed in [4] with the help of terminal subsequently.
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

sliding mode control methodology. However, the guidance law, To avoid this kind of singularity using the nonsingular terminal
proposed in [4] contains a term jθ − θF jα−1 . The value of SMC algorithm [41,42], the switching surface for the same double
jθ − θF jα−1 becomes very small when the impact angle error integrator system can be redefined as
becomes very small. Since 0 < α < 1, it makes jθ − θF jα−1 shoot up
to a very large value. Consequently, the guidance law developed in [4] s
x  βx_ α ; β > 0; 1 < α
p∕q < 2 (10)
exhibits a singularity when the impact angle error becomes very
small, leading to control saturation. In [4], some modifications to the where p and q are odd integers. Note that, if p and q are odd, then it
guidance law are suggested to account for this singularity. However,
will preserve the sign of x_ in x_α and the z1∕q will always have a real
it is preferable to design a guidance law that does not have this
solution if z is real, irrespective of the sign of z. The discontinuous
singularity. Note that the existence of this singularity is not just
control u, which achieves x
0 within a finite time, is chosen as
confined to the guidance law in [4] but is a generic problem for
guidance law based on terminal sliding mode control. In this paper,
u
−M signs; M>0 (11)
the objective is to design a guidance law that is free from singularity
and also has all the nice properties of the guidance law designed in
[4], that is, the robustness and finite time error convergence. Now, to select M, evaluate s_
x_  αβx_ α−1 u and choose
As is well known, SMC theory, developed for control affine M > 1∕αβjxj _ 2−α , which ensures that the sliding mode occurs
dynamic systems, offers a simple design procedure and the within a finite time. During the sliding mode, s
0 and, because
implementation of this algorithm leads to closed-loop dynamics that α
p∕q, the sliding mode dynamics is given by
are insensitive to parameter variations and disturbances. This
algorithm requires the design of functions of the system states and x_
−βx
^ 1∕α ; β^
β−1∕α (12)
controls which are discontinuous on them. For example, consider the
system On integrating Eq. (12), it can be shown that the time required for the
state x to reach 0 is T c ≥ 0 and is given by
x_
fx  bxu; x ∈ Rn ; u ∈ Rm ; m≤n (6)
 
jxT s j1−1∕α p
for which the control u should be designed to drive x → 0. The Tc

jxT s jp−q∕p (13)
1 − 1∕α p−q
application of this algorithm involves the selection of 1) functions
sx ∈ Rm , which are typically linear combinations of the states that where T s is the time instant at which the sliding mode occurs. It can
lead to x → 0 if sx
0, and 2) control u acting discontinuously, also be observed from Eq. (13) that the convergence time T c , which is
componentwise, on sx in the form
the time required to obtain x
0, is finite for all finite values of
  xT s . In the case of terminal SMC also, similar results for time of
ui if si x > 0; convergence can be derived.
ui
(7)
u−i if si x < 0 The discontinuous controller might result in a limit-cycle behavior
of the state s, which is known as chattering [1]. There exist several
where i
1; : : : ; m. The functions sx are denoted as switching techniques by which this phenomenon can be attenuated. One
surfaces and, with the appropriate choice of control magnitudes u i method is by reducing the discontinuous control magnitude, where
and u−i , the states of the system are aligned toward sx, reach it the control input is chosen as a sum of an equivalent controller and a
within a finite time interval, and then continue to remain on it. The discontinuous controller. For example, the control can be chosen as
resulting closed-loop dynamics, now given by sx
0, are known the sum
as sliding modes and are of order m. The features of the sliding mode
that set it apart from other nonlinear control algorithms are that u
ueq  udisc (14a)
sx
0 holds within a finite time interval and that the sliding
mode dynamics are insensitive to variations in system parameters
and external disturbances that might appear in the controllable
subspace [1]. ueq
−1∕αβx_ 2−α ; udisc
−M signs (14b)
There exists a variant of conventional SMC known as terminal
SMC,which gives finite time convergence but has a singularity when As has been shown in [41,42], this choice of control, which leads to
the error being controlled becomes very small [39,40]. In this paper, the dynamics s_
−Mx_ α−1 signs, for any M > 0, will enforce
another variant of the conventional SMC algorithm, known as the sliding mode on s
0, even if x_
0. However, the application of
NTSMC [41,42], is used to design the guidance laws. This algorithm, this idea presumes the exact knowledge of the system being
while retaining the benefits offered by conventional sliding modes, controlled. In the case of terminal SMC also, the control can be
also ensures the finite time convergence of the state being controlled chosen similar to Eq. (14) with the equivalent control component
to its desired value. given by
 KUMAR, RAO, AND GHOSE 1117

ueq
−αβjxjα−1 x_ (15) To design these components, the equivalent control approach [1] is
used. On differentiating Eq. (19), we get the derivative of the
From Eq. (15), it can also be observed that, because 0 < α < 1, ueq switching surface as
shoots up to large values due to the presence of singularity when x
goes to a small value. S_
θ_ − θ_ F   αβθ_ − θ_ F α−1 θ
− θ
F  (21)
Note that in the case of nonsingular terminal SMC, the requirement
of α > 1 comes from Eq. (13), which gives a positive error Substituting Eqs. (16) and (17) in Eq. (21), we get
convergence time, whereas the requirement of α < 2 comes from
   
 
_
Eq. (14b). From Eq. (14b), it can be observed that the exponent of x, a a α−1 2_r θ_ cos θT
which is 2 − α, becomes negative if α > 2 and may lead to a S_
θ_ − T  αβ θ_ − T −  aT
VT VT r r
singularity when x_ ≈ 0 during implementation of the control   
algorithm. Hence, the nonsingular TSMC algorithm imposes a cos θM a_
− aM − T (22)
requirement of 1 < α < 2 to avoid singularity during implementation. r VT
The motivation to use this algorithm comes from [41,42], where this
concept has been demonstrated successfully for the control of Now, to obtain the equivalent control component of the sliding mode
dynamic systems such as n-link rigid manipulators, because it avoids (SM) controller, let S_
0 in Eq. (22) to obtain
the problem of singularities that the terminal SMC algorithm presents
 
  
during implementation. r 2r_ θ_ cos θT a_
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

aeq
M
−  aT − T
cos θM r r VT
 2−α 
1 _ aT
IV. Design of Sliding Mode Guidance Laws  θ− (23)
αβ VT
In this section, the impact angle constrained guidance laws are
developed using the kinematics-based nonlinear engagement and the discontinuous component of the SM controller is selected in
dynamics, given by Eq. (1). First, the guidance law is designed for the form as
a general case of a maneuvering target and then the same will be
derived for the special case of constant velocity and stationary targets. M
adisc
M
signS (24)
From Eq. (1), the relative degree two dynamics between the control signcos θM 
input aM and the LOS angle θ, given by
Next, to determine the bound on the gain of the discontinuous
   
2r_ θ_ cos θT cos θM component of the SM controller, consider the Lyapunov function
θ

−  aT − aM (16) candidate as V
1∕2S2 . On differentiating V and substituting
r r r
Eqs. (20), (23), and (24) and simplifying, we get
is used to design the interceptor lateral acceleration aM to ensure that  
jSjj cos θM j _ aT α−1
the θ
θF condition is met at interception. Note that, because aM is V_
SS_
−αβM θ− (25)
multiplied by the term cos θM , the LOS angle can be controlled only r VT
if jθM j ≠ π∕2. From the dynamics of θM , it can be easily shown that
the jθM j
π∕2 is not an equilibrium point and, hence, the lateral Since 1 < α
p∕q < 2 and p and q are positive odd integers, the
acceleration aM can be used to control θ. term θ_ − aT ∕V T α−1 > 0 for any θ_ ≠ aT ∕V T. Let
 
j cos θM j _ aT α−1
A. Maneuvering Targets ρθ;
_ θM ; aT 
αβM θ− (26)
In this subsection, the guidance law is derived for a general case of r VT
maneuvering targets. For a maneuvering target, it can be seen from
Eq. (5) that the desired LOS angle θF changes with time, due to the As mentioned earlier, θM
π∕2 is not a stable equilibrium, and
change in flight-path angle of the target at interception γ Tf . The rate hence the cos θM
0 condition can occur only momentarily. Hence,
of change of the desired LOS angle θ_ F is given by the analysis can be performed with an assumption that, for most of the
time during engagement, cos θM has nonzero value. From Eqs. (25)
θ_ F
aT ∕V T (17) and (26), it is clear that

In this case, the target maneuver aT is assumed to be governed by V_
−ρθ;

_ θM ; aT jSj (27)
first-order dynamics as
where ρθ;_ θM ; aT  > 0 with any M > 0 for all θ_ ≠ aT ∕V T . So, V_ is
τs a_ T
aCT − aT ; jaCT j ≤ aT max (18) negative definite and hence the condition for Lyapunov stability is
satisfied for all θ_ ≠ aT ∕V T and, consequently, the sliding mode
where τs is the positive time constant of the target dynamics and occurs.
aT max is the maximum target maneuver command. The Lyapunov stability condition is satisfied only for the case of
Now, to design the interceptor lateral acceleration aM using the θ_ − aT ∕V T ≠ 0, but it may go to zero during the reaching phase. To
principles of nonsingular TSMC, the switching surface is selected in achieve sliding mode S
0, within finite time, it is necessary to show
the form that θ_ − aT ∕V T
0 is not an attractor. To show this, analyze the
dynamics of z
θ_ − aT ∕V T . On differentiating z, we get
p
S
θ − θF   βθ_ − θ_ F α ; β > 0; α
; p>q    
q a_ T 2r_ θ_ cos θT cos θM a_
z_
θ −


−  aT − aM − T (28)
(19) VT r r r VT

where p, q are assumed to be odd integers and 1 < α < 2. Next, the Substituting for the control aM from Eqs. (20), (23), and (24) in
control input that will enforce the sliding mode on S
0 is designed Eq. (28) and simplifying, we get
as the sum of an equivalent and a discontinuous controller, in the form
Mj cos θM j
aM
aeq disc z_
−1∕αβz2−α − signS (29)
M  aM (20) r
 1118 KUMAR, RAO, AND GHOSE

  
Now, for z
0, Eq. (29) reduces to signS r a 2−α _ 2−α
M> θ_ − T −θ
j cos θM j αβ VT
Mj cos θM j  
z_
− signS (30) ra_ T
r  aT cos θT −
VT
Because M, r, and j cos θM j are all positive quantities, for S > 0 and
S < 0, it can be seen from Eq. (30) that z_
−η and z_
η, during the engagement. Note that this is a sufficient condition but not
respectively, where η
Mj cos θM j∕r. This shows that z
0 is necessary.
not an attractor and, hence, near the vicinity of z
0, that is, jzj ≤ δ, For the case of a constant maneuvering target, a_ T
0. In addition,
with a small positive δ, the crossing of the trajectory from the if the target acceleration aT is known but not its rate of change a_ T ,
boundary of z
δ to z
−δ for s > 0 and z
−δ to z
δ for S < 0 then during implementation of the guidance laws, a_ T can be treated as
occurs within a finite time. The same conclusion can also be drawn an uncertainty and made equal to zero in the equivalent control
from Fig. 2, in which x1 and x2 correspond to the states (θ − θF ) and component given by Eq. (23). Hence, in both the constant
(θ_ − aT ∕V T ), respectively, which shows the phase plot of the maneuvering and unknown a_ T case, the equivalent control
dynamics of the desired LOS angle error. For the region outside component is given by
z ≤ δ, it is already proved that the Lyapunov stability condition is  
  
1 1 _ aT 2−α
satisfied. Therefore, from any arbitrarily initial point, the sliding aeq
M
−2r_ θ_ aT cos θT  θ−
mode S
0 can be reached within a finite time. cos θM αβ VT
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

As shown in Sec. III, after occurrence of the sliding mode, the time
taken for the state to be controlled, that is, the LOS angle error, to whereas the discontinuous control component adisc
M is still selected as
reach zero with this variant of SMC, given by Eq. (13), is finite. in Eq. (24) with the inequality (33) reduced to
Hence, within a finite time, the impact angle can be achieved with this
guidance law. r signS
M>− a_ (33)
To implement this SM controller, the information of target V T j cos θM j T
acceleration aT and its rate of change a_ T are required in the equivalent
control component given by Eq. (23). But, in practical scenarios, this which further reduces for the case of the constant maneuvering target
information may not be available to the guidance law. In this case, as M > 0. Note that, similar to the case of target information available
the equivalent control component of the SM controller can be to the guidance law, it can also be shown that θ_ − aT ∕V T
0 is not an
implemented by treating the target acceleration and its rate of change attractor for the cases where this information is partially or
as uncertainties, which can be suppressed by the discontinuous completely unknown to the guidance law.
control component of the SM controller. So, the equivalent control
component in this situation is given by B. Stationary and Constant Velocity Targets
 
 In this subsection, the guidance law for the constant velocity and
r 2r_ θ_ 1 stationary target case is derived from the results for a maneuvering
aeq
−  θ_ 2−α (31)
M
cos θM r αβ target case. In both cases, the constant velocity targets and stationary
targets, aT
0 and a_ T
0 and so θ_ F
0. The switching surface for
whereas the discontinuous control component is still selected in the these cases reduces to
form given by Eq. (24).
To determine the bound on the gain of the discontinuous control S
θ − θF   βθ_ α ; β > 0; 1 < α
p∕q < 2 (34)
component M, consider again the Lyapunov function candidate
V
1∕2S2 . On differentiating V and substituting Eqs. (20), (24), where p and q are odd integers. In these cases, the equivalent control
and (31) and simplifying, we get given by Eqs. (23) and (31) reduces to the same expression and is
given by
 

aT α−1 Mj cos θM j  
  
V
SS
−αβjSj θ −
_ _ _ − signS 1 r _ 2−α
VT r aeq
−2_r θ 
_ θ (35)

 2−α    M
cos θM αβ
1 a cos θT a_
× θ_ − T − θ_ 2−α  aT − T
αβ VT r VT whereas the discontinuous control component, selected in the same
(32) form as Eq. (24), with the bound on the gain M, given by the
inequality (33), reduces to zero as in the case of known target
Now, as mentioned before, θ_ − aT ∕V T α−1 > 0 for all acceleration and its derivative.
Note that r appears in the denominator of the control input in the
θ_ ≠ aT ∕V T . Hence, for V_ to be negative definite, the gain M of
dynamics of θ given by Eq. (16). This may appear to lead to a small
the discontinuous control component adiscM should be selected such
control magnitude when r is large and large control magnitude when
that it satisfies the inequality
it is small. The control input aM consists of two components given by
Eq. (20). However, it can be seen from Eqs. (23) and (31), inequality
(33), or Eq. (35) that r, which appears in the numerator of aeq
M , cancels
r in the denominator of aM in Eq. (16) and hence the problem of
sluggish and large control with large and small values of r,
respectively, is not associated with the equivalent control component.
On the other hand, the discontinuous control component, given by
Eq. (24), which contributes to the control significantly before
occurrence of the sliding mode, makes the control sluggish for large
value of r. However, after occurrence of the sliding mode, S ≈ 0,
and because the signum function is replaced by the sigmoid function,
the contribution of discontinuous control is negligible to the con-
troller. Hence, the problem of large control for small r never arises.
In addition, to overcome the difficulty of sluggish control with
Fig. 2 Phase plot of the error dynamic system. the discontinuous component, one may compensate for r in the
 KUMAR, RAO, AND GHOSE 1119

denominator of aM in Eq. (16) by choosing the gain as Mr in place of goes to zero. To ensure that A ≥ 2, the guidance law is modified to
M in Eq. (24).


j_rjθ_ rθ_ 1−α
aeq
M
2 (40)
C. Domain of Sliding Mode Capturability j cos θM j αβj_rj
In this subsection, capturability analysis, based on the states after
occurrence of the sliding mode, is discussed. This is termed as Even with this modification, it can be shown that, while intercepting
“sliding mode capturability analysis” and is different from stationary targets, aeq
M → ∞ as r_ → 0. On the other hand, while
capturability in the conventional sense, which is based on initial intercepting constant velocity targets,
condition. Note that, after occurrence of the nonsingular terminal
sliding mode, the equivalent control component, given by inequality rθ_ 2−α
(33), reduces to the form given by aeq
M

αβj cos θM j
1
aeq
M
a cos θT − 2_r θ
_ (36) when r_
0. But note that, even now, aeqM → ∞ if cos θM → 0. As it
cos θM T can also be shown that jθM j
π∕2 is not an equilibrium point in the
closed loop, during implementation, the lateral acceleration is
and also leads to bounded according to the saturation function

θ
θF ; θ_
θ_ F
aT ∕V T ; θ_ T
0 (37) aM max signaM  if jaM j ≥ aM max
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

aM
(41)
aM if jaM j < aM max
which, as expected, is very similar to the results derived in [4] after
occurrence of the terminal sliding mode. Hence, the analysis where aM max is the maximum allowable interceptor lateral
performed to obtain sliding mode capturability in [4] is applicable for acceleration.
this case also. This shows that, if the range rate V r at the point where
impact angle criterion has been satisfied is negative, then the
trajectory in the relative velocity space does not cross the V θ axis V. Simulation Results
during the engagement and the closing speed remains positive and
thus leads to interception of the target. In this section, the results of the performance of the proposed
guidance laws obtained through numerical simulation are presented.
These results validate the designs and satisfy all the objectives.
D. Interpretation of the Guidance Laws
Simulation results validating the guidance law for the case of
The guidance law from Eqs. (20) and (31) can also be rewritten as stationary targets, constant velocity targets, constant maneuvering
the sum of three components targets, and weaving targets are presented. In all the cases, the desired
 _ 1−α  final LOS angle is computed using Eq. (5), and the lateral
2−_rθ_ tgo θ −_rθ_ acceleration is bounded according to Eq. (41). Further, a continuous
aM
  adisc (38)
cos θM αβ cos θM M approximation of the discontinuous controller, when jsj ≤ ε, is used
to reduce chattering in the system. As has been shown in [1], with
where the conventional approximation of time-to-go tgo is assumed to such an approximation, the deviation from the ideal sliding mode is of
be tgo
r∕−_r [43]. The first component is similar to the PNG law the order ε. Following this technique, the discontinuous function
with an effective navigation constant of two and an additional term in signs is approximated by the sigmoid function
the denominator, which comes from the nonlinear engagement  
dynamics. This component steers the interceptor to the collision 1 1
sgmfs
2 −as − ; a>0 (42)
course, thus enabling it to intercept the target. The second component 1  exp 2
aids the PNG law. Because α
p∕q, 1 − α
q − p∕q and so
θ_ 1−α > 0, ∀ θ._ As a consequence, this component can also be where the constant a is inversely proportional to ε and chosen as 20.
interpreted as the component of the lateral acceleration, which The parameters α and β of the switching surface, given by the
increases the effective navigation constant initially and its effect general form of Eq. (19), determine the transient response
decreases as the time-to-go decreases. The third term corresponds to characteristics of the closed-loop LOS angle dynamics. In this paper,
the feedback component that accounts for uncertainties in the p
5, q
3, and β
1 are chosen, which satisfies the requirement
dynamics. of the NTSMC algorithm on α to avoid singularity during
Note that the guidance law, with the target information available to implementation. The gain M of the discontinuous control component
M is chosen as 3 × 10 . These values, on performing several
adisc
it, can also have similar interpretation. In this case, the equivalent 3
control component is given by Eq. (23), in which the first term is the numerical simulations, were observed to yield satisfactory results.
same as in Eq. (38) and the second term has (θ_ − aT ) in place of θ,_ but
it still adds to the interceptor lateral acceleration before occurrence of A. Constant Speed Interceptors
sliding mode. The third and fourth term in Eq. (23) corresponds to the
In this subsection, simulation results are presented for constant
term that accounts for target acceleration and its rate of change,
speed interceptors against different kinds of targets. The initial
respectively, and the discontinuous component accounts for the
conditions of the interceptor, as well as its speed and lateral
uncertainties.
acceleration bounds aM max , which are common to all types of targets,
are listed as data set 1 in Table 1. The target parameters, applicable to
E. Extension to Large Heading Angle Errors the particular type of target, are also listed in Table 1.
The equivalent control from Eq. (31) can also be rewritten in the In the figures that follow, the point denoted by A always
form corresponds to the initial position of the interceptor. The points

 denoted by other letters of the alphabet correspond to those time
−A_r θ_ rθ_ 1−α instants during the engagement when events such as the occurrence
aeq
M
; A
2 (39) of sliding mode and interception occur.
cos θM −αβr _

It can be seen that, if r_ > 0, then A < 2, which implies that the 1. Stationary Targets
effective navigation constant is less than two. Note that studies on Here, the case of intercepting a stationary target at an impact angle
classical PNG [44] reveal that, for N < 2, the magnitude of the missile of −90 deg is presented first. Without loss of generality, γ T
0 deg
lateral acceleration goes to infinity as the interceptor–target range is used as a reference to determine the desired final LOS angle, given
 1120 KUMAR, RAO, AND GHOSE

Table 1 Simulation parameters for the case of demands on the lateral acceleration, which occur primarily due to the
constant speed interceptor model requirement that the sliding mode should occur much before
Parameters Data set 1 Data set 2 interception, can be reduced by changing the switching surface
parameters as well as reducing the magnitude of the discontinuous
r0 10 km 15 km
θ0 30 deg 0 deg component of the lateral acceleration. These changes, in turn, would
γ M0 60 deg 45 deg change the trajectory of the interceptor. It can also be noted that, after
VM 500 m∕s 2000 m∕s satisfying the impact angle criterion, the interceptor lateral accelera-
γ T0 0, 120 deg 45 deg tion demand remains at zero.
VT 0, 400 m∕s 1600 m∕s The same guidance law is implemented to intercept stationary
aM max 400 m∕s2 Unbounded targets at different impact angles but from the same initial heading
angle. For these cases, the trajectories in the X; Y space and the
corresponding lateral acceleration values are shown in Fig. 4. As can
by Eq. (5), as θF
90 deg. The trajectories of the interceptor in the be seen, the lateral acceleration for each case has a similar trend as
physical space and relative velocity space, the occurrence of sliding that shown in Fig. 3d.
mode on the surface S
0, and the corresponding lateral The results of the performance of the guidance law to intercept at
acceleration that enforces the sliding mode are shown in Fig. 3. an impact angle of −90 deg, which leads to a desired LOS angle of
The points A and F in Fig. 3 correspond to the start and end point. It θF
90 deg, but for different initial interceptor heading angles, are
can also be observed that, at point F, the LOS rate becomes zero. shown in Fig. 5. As can be seen, the controls aM , shown in Fig. 5b, are
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

From Fig. 3b, it can also be seen that the point B corresponds to able to steer the interceptor to the same interception course decided
the point where V θ becomes zero for the first time, but at this time, by the desired LOS angle.
the impact angle criterion has not been satisfied. Hence, to meet the
objective of the impact angle, the interceptor changes its course and 2. Constant Velocity Targets
reaches point C where the relative velocity component V θ is In this case, the target maneuver is assumed to be zero (aT
0).
maximum, as shown in Figs. 3a and 3b. From Fig. 3c, it can be The other parameters used for the simulation studies for this type of
observed that the sliding mode occurs at point D in finite time and target are as given in data set 1 of Table 1. In this case, an impact angle
much before interception. In Fig. 3, points D and E denote the time of 90 deg is desired and the simulation results are shown in Fig. 6.
instants at which sliding mode occurs and the impact angle criterion Similar to the stationary target case, points A and D in Fig. 6
is satisfied, respectively. This observation supports the claim that correspond to the start and end points. It can be observed from Fig. 6b
the desired LOS angle is achieved within a finite time with the that the LOS rate θ_ is very small at the time instant at which the sliding
nonsingular TSMC algorithm after occurrence of the sliding mode. In mode has occurred. The points B and C, shown in Fig. 6b, denote the
Fig. 3b, the points B, E, and F are close to each other because, when time instants at which sliding mode has occurred and the impact angle
the impact angle error is satisfied, the LOS rate is close to zero. In criterion has been satisfied, respectively. It can be seen from Figs. 6a
addition, as can also be seen in Fig. 3a, once the interceptor orients and 6b that it is only at point C, when the interceptor begins to move in
itself in a direction to achieve the desired impact angle at point E, the a straight line, that the desired LOS angle is achieved. In addition, at
LOS rate remains at zero and the interceptor does not change its B, the value V r T B  < 0, which validates the capturability analysis,
course. From Fig. 3d, it can be seen that the interceptor applies a large because interception occurs at the point denoted by D. It can also be
lateral acceleration at two instants of time: the first of which is during observed from Fig. 6a that the points B, C, and D are close to each
the start of the engagement and the second is when S ≈ 0. These large other because, at these instants, the relative velocity V θ is close to

a) b)

c) d)
Fig. 3 Stationary target: a) trajectories in
X;Y space; b) trajectory in
V r ;V θ  space; c) occurrence of sliding mode; d) interceptor lateral
acceleration aM .
 KUMAR, RAO, AND GHOSE 1121

a) b)
Fig. 4 Interception of a stationary target at various impact angles: a) trajectories in
X;Y space; b) interceptor lateral acceleration aM .
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

a) b)
Fig. 5 Interception of a stationary target at a given impact angle, but with different initial heading angles: a) trajectories in
X;Y space; b) interceptor
lateral acceleration aM .

a) b) c)
Fig. 6 Constant velocity target: a) trajectories in
X;Y space; b) occurrence of sliding mode; c) interceptor lateral acceleration aM .

zero. From Fig. 6c, it can also be observed that the interceptor lateral acceleration is selected to enforce the sliding mode on the manifold
acceleration demand has a trend similar to the stationary target case. given by Eq. (19). For the simulation, an impact angle at the time of
Note that the analysis of the trajectories in V r ; V θ  space similar to interception is selected to be 60 deg. In this case, since aT ≠ 0, the
Fig. 3b can also be done here. desired LOS angle satisfies the relations in Eq. (17).
The performance of guidance laws are also evaluated for the case The simulation results are shown in Fig. 8, in which the points A
where the intercept is required with the target at an impact angle of and D correspond to the start and end of the engagement. It can also
90 deg but for different initial interceptor heading angles. Results of be observed that sliding mode occurs in finite time, denoted by point
these simulations are shown in Fig. 7. Based on these results, B, as shown in Fig. 8b. Note that, due to the inherent finite time
conclusions, similar to the stationary target case, can be drawn. convergence property of the NTSMC algorithm, the impact angle
criterion has been satisfied at point C, as shown in Fig. 8c, within a
finite time after occurrence of the sliding mode. From Fig. 8b, it can
3. Constant Maneuvering Targets be noted that the interceptor lateral acceleration profile is smooth and
In the earlier subsections, it has been shown that the proposed applies a large lateral acceleration at two instants: first, at the start of
guidance laws are able to intercept the targets at a desired impact engagement and second, when S comes close to zero, that is, just
angle within a finite time. Here, performance of the same guidance before occurrence of the sliding mode. These large demands on the
laws for a target, executing a constant maneuver, is evaluated through lateral acceleration, which occur primarily due to the requirement that
numerical simulation. For this case, the simulation conditions are as the sliding mode should occur much before interception, can be
given in data set 1 in Table 1. In this subsection, the target maneuver is reduced by changing the switching surface parameters, as well as by
assumed to be aT
30 m∕s2 . For this type of target, the lateral reducing the magnitude of the discontinuous component of the lateral
 1122 KUMAR, RAO, AND GHOSE

a) b)
Fig. 7 Interception of a constant velocity at a given impact angle, but with different initial heading angles: a) trajectories in
X;Y space; b) interceptor
lateral acceleration aM .
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

a) b) c)
Fig. 8 Constant maneuver target: a) trajectories in
X;Y space; b) occurrence of sliding mode and required interceptor lateral acceleration aM ;
c) tracking of LOS and its rate.

acceleration. From Fig. 8b, it can be noted that, even after occurrence To evaluate the performance of the proposed guidance law,
of the sliding mode, the interceptor lateral acceleration demand does simulations with different values of constant target maneuver aT , but
not reduce to zero. After occurrence of the sliding mode, the at an impact angle of 90 deg, have been carried out and the results are
equivalent control component given by inequality (33) reduces to shown in Fig. 10. It can be observed that the interceptor is able to
   intercept the target executing different constant maneuvers ranging
aT Vr from 1 to 5g. From Fig. 10, it can be seen that the engagement time
aeq
M
cos θT − 2 increases with a decrease in the maneuvering capability of the target.
cos θM VT
This is due to the requirement of intercepting the target at an angle of
Since aT ≠ 0 and V r T C  < 0, Eq. (43) gives a nonzero value of the 90 deg. The profile of interceptor lateral acceleration demand, as
equivalent control component. Here also, the trajectories can be shown in Fig. 10b, shows a similar trend to the previous cases. It can
analyzed in V r ; V θ  space similar to Fig. 3b. also be noted that, in the case of aT
5g, the interceptor lateral
The performance of the guidance law is also evaluated for the acceleration demand is saturated at 40g due to the sharp turn required
scenario of intercepting the target at an impact angle of 90 deg but for to meet the impact angle requirement after occurrence of the sliding
different initial interceptor heading angles. Results of these simu- mode. Similar to the case of the target’s anticlockwise maneuvers, it
lations are shown in Fig. 9. Based on these results, conclusions can also be shown that the interceptor can intercept the target
similar to the stationary and constant velocity target case can also executing clockwise maneuvers of magnitude 2.8g with the same
be drawn. choice of gain M of the guidance law. By increasing the magnitude of

a) b)
Fig. 9 Interception of a constant maneuvering target at a given impact angle, but with different initial heading angles: a) trajectories in
X;Y space;
b) interceptor lateral acceleration aM .
 KUMAR, RAO, AND GHOSE 1123

a) b)
Fig. 10 Constant maneuver with different positive value of constant maneuver aT : a) trajectories in
X;Y space; b) interceptor lateral acceleration aM .

gain M, it can be shown that the proposed guidance law can intercept To evaluate the robustness of the guidance laws in the absence of
the targets with higher maneuver levels. information on target acceleration aT for a constant maneuvering
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

To demonstrate the effect of gain selection on the performance of target, given by Eqs. (20), (24), and (31), simulations were performed
the guidance laws, some simulations were carried out with a different with the same initial condition as given in data set 1 in Table 1 but with
value of gain M of the discontinuous control component adiscM , and the a desired impact angle of 30 deg. In addition, these results are
results are shown in Fig. 11. For these simulations, three different compared with the results for the case where it is assumed that the
values of gain M
0.3 × 104 , 0.7 × 104 , and 1.1 × 104 are selected target information is available to the guidance laws but with the same
and the results are shown in Fig. 11. It can be observed that, although simulation conditions, and results are shown in Fig. 12. The analysis
the guidance law with higher gain enforces the sliding mode earlier in similar to Fig. 3b can be done for the trajectories of interceptor and
the system, as shown in Fig. 11b, and makes it insensitive to dis- target in V r ; V θ  space. From Fig. 12a, it can be observed that, in
turbances, it leads to saturation in the interceptor lateral accelera- both the cases, the interceptor is able to intercept the target with the
tion as shown in Fig. 11c. From Fig. 11a, it can also be noted that, as a desired impact angle. Note that, in the present engagement scenario,
consequence of the saturation in lateral acceleration at the start of the the interceptor takes less time to intercept the target in the case of
engagement, the interceptor changes its course such that it deviates absence of target information than in the case where it is available to
from its path and, to correct the same, it needs a very sharp turn the guidance law. From Fig. 12b, it can be noted that sliding mode
when S ≈ 0, leading to saturation again. On the other hand, the lower occurs earlier in the case where aT is known. The interceptor lateral
value of gain M gives a smooth trajectory but enforces the sliding acceleration demand profile in both cases, as shown in Fig. 12b, has a
mode later. similar trend.

a) b) c)
Fig. 11 Constant maneuver with different value of gain M of discontinuous control component adisc
M : a) trajectories in
X;Y space; b) occurrence of
sliding mode; c) interceptor lateral acceleration aM .

a) b)
Fig. 12 Comparison for constant velocity targets with and without target acceleration information: a) trajectories in
X;Y space; b) occurrence of
sliding mode and required interceptor lateral acceleration aM .
 1124 KUMAR, RAO, AND GHOSE

4. Weaving Targets Figs. 4, 7, and 8 in [10]. It can be observed that, during the start of the
Now, to evaluate the performance of the proposed guidance laws engagement, unlike in the deviated pursuit scheme, the interceptor
against more complex maneuvers, simulations are carried out for lateral acceleration demand has a value, as shown in Fig. 15c. In the
weaving targets (targets executing sinusoidal maneuvers) [43] and vicinity of S ≈ 0, although the interceptor lateral acceleration
the results are shown in Fig. 13. In this case, the simulation conditions demand is high as in the deviated pursuit guidance law, it should be
are the same as those given in the data set 1 of Table 1 but with the noted that the interceptor lateral acceleration demand increases
target executing a maneuver given by aT
30 sinπt∕10 m∕s2. In gradually and there is no sudden increase in demand as in the deviated
this case, it is assumed that a_ T is unknown to the guidance law and is pursuit case. It can be observed that, in both the schemes, the lateral
hence made equal to zero in the equivalent control given by Eq. (23). acceleration demand becomes constant toward the end of the
In Fig. 13, points A and D denote the start and end points of the engagement. In the present guidance law, by selecting switching gain
engagement. Note that points B and C denote the time instants at of discontinuous control component adisc M appropriately, we can
which sliding mode occurs and the impact angle criterion has been reduce the maximum value of interceptor lateral acceleration
satisfied, respectively. It can also be seen from Fig. 13b that the demand. Hence, it can be concluded that the results with the present
interceptor lateral acceleration demand shows a smooth behavior but guidance scheme are comparable to the deviated pursuit guidance
reaches its maximum limit for some time. scheme, and it also gives a lower value at the start of the engagement
To evaluate the performance of the proposed guidance law against and gradual increase in the lateral acceleration demand before
a weaving maneuver of the target, simulations were done with satisfying the impact angle criterion.
different values of the amplitude of the weave maneuver and the
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

results are shown in Fig. 14. It can be observed that the interceptor is B. Realistic Interceptors
able to intercept the target at the desired angle, as shown in Fig. 14a. Although in the previous subsection the simulation results are
The occurrence of sliding mode and the interceptor lateral presented for constant speed interceptors, performance will be shown
acceleration profile show similar trends as in Fig. 13. to be equally good for interceptors with varying speeds when we
consider realistic interceptors. This is due to the inherent robustness
5. Comparison with Deviated Pursuit Guidance Scheme property of the sliding mode algorithms. To evaluate the effectiveness
In this subsection, comparison of the guidance law with an existing of the proposed guidance law, numerical simulations have been
guidance law in the literature, which is based on the conventional performed with a realistic interceptor model [45] for different kinds
sliding mode control algorithm and a deviated pursuit strategy [10], is of targets. In this paper, while deriving the guidance law, an assump-
given. The simulation conditions for the tail-chase scenario, as given tion of constant speed ratio of target to the interceptor was made. But,
in [10], are listed as data set 2 in Table 1, with the desired impact angle in this section, it will be shown that, due to the inherent robustness
of 0 deg. In this case, the target is assumed to execute a constant property of the guidance law, it can actually achieve the same
maneuver of aT
−50 m∕s2 . Figure 15 shows the simulation results objective even for time-varying speed ratio. Because the proposed
with these initial conditions. Figures 15a–15c show interceptor– guidance law is designed for planar engagement geometry, an inter-
target trajectories, occurrence of sliding mode, and the lateral ceptor model in the pitch plane is considered to validate the per-
acceleration demand, respectively. These can be compared with formance. The thrust is assumed as a prescribed function of time.

a) b) c)
Fig. 13 Weaving target: a) trajectories in
X;Y space; b) occurrence of sliding mode and required interceptor lateral acceleration aM ; c) tracking of
LOS and its rate.

a) b) c)
Fig. 14 Weaving target with different amplitude of weave maneuver: a) trajectories in
X;Y space; b) occurrence of sliding mode; c) interceptor lateral
acceleration aM .
 KUMAR, RAO, AND GHOSE 1125

a) b) c)
Fig. 15 Simulations for comparison with deviated pursuit guidance scheme [10]: a) trajectory; b) occurrence of sliding mode; c) interceptor lateral
acceleration aM .

8
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

The equations of motion of a point mass flying over a flat, nonrotating >
> 33; 000 0 ≤ t ≤ 1.5
Earth are given by <
T
7500 1.5 < t ≤ 8.5 (49)
>
>
X_ M
V M cos γ M ; Y_ M
V M sin γ M :
(43a) 0 t ≥ 8.5

and the mass variation of the vehicle is given by

T−D a − g cos γ M
V_ M
− g sin γ M ; γ_ M
M (43b) 8
m VM >
> 135 − 14.53t 0 ≤ t ≤ 1.5
<
m
113.205 − 3.31t − 1.5 1.5 ≤ t ≤ 8.5 (50)
where m, V M , and γ M are the mass, speed, and flight-path angle of >
>
:
the missile; T and D are thrust and drag on the interceptor; g is the 90.035 t ≥ 8.5
acceleration due to gravity; XM and Y M denote the position of the
interceptor; and aM is the interceptor lateral acceleration. In this where t is time in seconds. The atmosphere density ρ as a function of
realistic interceptor model, the aerodynamic drag is modeled as height above sea level h is given by

Km2 a2M ρ
1.15579 − 1.058 × 10−4 h  3.725 × 10−9 h2 − 6 × 10−14 h3

D
D0  Di ; D0
CD0 Qs; Di
;
Qs (51)
1 1
K
; Q
ρV 2 (44) The simulation conditions are the same as those listed in Table 1.
πAr e 2

where D0 and Di denote the zero-lift drag and induced drag, 1. Stationary Target
respectively, and Cd0 , K, Ar , e, ρ, s, and Q are the zero-lift drag First, the problem of intercepting a stationary target at an impact
coefficient, induced drag coefficient, aspect ratio, efficiency factor, angle of −60 deg, with the simulation condition listed in data set 1 of
atmosphere density, reference area, and dynamic pressure, respec- Table 1, is presented. Similar to Sec. V.A.1, here also γ T
0 deg is
tively. The zero-lift drag coefficient and induced drag coefficient are assumed as a reference to determine the desired final LOS angle. In
given by this case, the desired final LOS angle is given by θF
60 deg. In
fact, in the stationary target case, the desired LOS angle θF , given by
8
> 0.02 M < 0.93 Eq. (5), remains constant during the engagement irrespective of the
>
>
>
> speed of the interceptor. The trajectories of the interceptor in the
< 0.02  0.2M − 0.93 M < 1.03 physical space and relative velocity space, the occurrence of sliding
CD0
(45)
>
> 0.04  0.06M − 1.03 M < 1.10 mode, the corresponding lateral acceleration which enforces sliding
>
> mode, variation of the interceptor speed, and drag acting on the
>
:
0.0442 − 0.007M − 1.10 M ≥ 1.10 interceptor are all shown in Fig. 16. The points A and H in Fig. 16
correspond to the start and the end point of the engagement,
respectively. Note that the boost phase of the interceptor begins with

0.2 M < 1.15 the start of the engagement itself and is denoted by point A in Fig. 16,
K
(46)
0.2  0.246M − 1.15 M ≥ 1.15 where point B denotes the time instant at which boost phase ends and
the sustain phase begins, whereas point E denotes the end of the
where M is the Mach number and is calculated as sustain phase and beginning of the terminal phase of the interceptor.
Note that, during the boost phase, the guidance law was not applied
VM and it starts only at point B where the sustain phase begins. It can also
M
p
; RC
288 J∕K kg (47) be seen from Fig. 16b that, at point C, V θ
0 for the first time but the
1.4RC T P impact angle criterion has not been satisfied. Hence, the interceptor
changes its course and reaches point D where the relative velocity
where T P , the temperature at height h above sea level, is modeled by component V θ becomes maximum. Note that, during the endgame
 phase of the engagement, denoted by the segment EH in the
288.16 − 0.0065Y M Y M ≤ 11000 trajectories in physical and relative velocity space, sliding mode
TP
(48)
216.66 Y M ≥ 11000 occurs at point F, and, as a consequence, the impact angle criterion
is satisfied at point G, as shown in Fig. 16c. Note that, at point G, V r
The reference area s, used in Eq. (44), is assumed to be 1 m2 and the is negative, which satisfies the requirement of sliding mode cap-
thrust profile of the vehicle is given by turability and hence interception is ensured. It can also be observed
 1126 KUMAR, RAO, AND GHOSE

a) b) c)
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

d) e) f)
Fig. 16 Realistic interceptor against stationary target: a) trajectories in
X;Y space; b) trajectory in
V r ;V θ  space; c) occurrence of sliding mode;
d) interceptor lateral acceleration aM ; e) speed of interceptor; f) drag on interceptor.

a) b)
Fig. 17 Realistic interceptor against interception of a stationary target at various impact angles: a) trajectories in
X;Y space; b) interceptor lateral
acceleration aM .

a) b)
Fig. 18 Realistic interceptor against interception of a stationary target at a given impact angle, but with different initial heading angles: a) trajectories in

X;Y space; b) interceptor lateral acceleration aM .
 KUMAR, RAO, AND GHOSE 1127
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

Fig. 19 Realistic interceptor against constant velocity target: a) trajectories in
X;Y space; b) trajectory in
V r ;V θ  space; c) occurrence of sliding mode;
d) interceptor lateral acceleration aM ; e) speed of interceptor; f) drag on interceptor.

from Fig. 16b that, after satisfying the impact angle criterion, V r sustain phase, a sudden increase in the drag profile, as shown in
decreases due to the decrease in speed of interceptor, as shown in Fig. 16f, can be observed, which is due to the induced drag, given by
Fig. 16e, as engagement proceeds. From Fig. 16e, it can be seen that, Eq. (44), which occurs when the guidance loop is closed. In the
during the boost phase, speed of the interceptor changes at a higher sustain phase, drag on the interceptor is larger than the available
rate because of availability of larger thrust compared with the drag thrust and hence speed of the interceptor decreases as shown in
acting on the interceptor as shown in Fig. 16f. At the start of the Fig. 16e. Because the rate of decrease of the speed of interceptor,

a) b) c)

d) e) f)
Fig. 20 Realistic interceptor against constant maneuvering target: a) trajectories in
X;Y space; b) trajectory in
V r ;V θ  space; c) occurrence of sliding
mode; d) interceptor lateral acceleration aM ; e) speed of interceptor; f) drag on interceptor.
 1128 KUMAR, RAO, AND GHOSE

a) b)
Fig. 21 Constant maneuver with different positive value of constant maneuver aT : a) trajectories in
X;Y space; b) interceptor lateral acceleration aM .
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

given by Eq. (43b), is governed by the difference between thrust and tions are shown in Fig. 18. Based on these results, conclusions,
drag, and the drag on the interceptor decreases due to the decrease in similar to the stationary and constant velocity target case, can also be
the lateral acceleration during the sustain phase, a corresponding drawn. Note that, in this case, the heading angle error is considered
variation in V M can be observed in Fig. 16e. Figure 16d shows the only between 0 and 180 deg.
interceptor lateral acceleration demand required to enforce sliding
mode and it can be noted that it requires a higher value at two instants: 2. Constant Velocity Targets
one at the start and the other just before achieving the impact angle In this subsection, the simulation results for the constant velocity
criterion. Note that a kink in the interceptor lateral acceleration targets, that is, aT
0, are presented. The simulation parameters
demand can be observed in Fig. 16d at the point where the sustain used for this case are those listed in data set 1 in Table 1 except for the
phase of the interceptor ends, that is, at t
8.5 s. It can also be noted initial LOS angle θ0
0 deg. In this case, an impact angle of
that, after satisfying the impact angle criterion, the interceptor lateral 135 deg is selected and the simulation results are shown in Fig. 19.
acceleration demand remains at zero. The points A and F in Fig. 19 denote the start and end of the
The same guidance law is implemented to intercept stationary engagement, respectively. Similar to the stationary target case, in
targets at different impact angles but with the same initial heading Fig. 19, the points A, B, and C denote the time instants at which boost,
angle and the results are shown in Fig. 17. As can be seen, the lateral sustain, and endgame phase of the engagement starts. From Fig. 19c,
acceleration for each case has a similar trend as that shown in it can be observed that, although the flight-path angle of the target
Fig. 17b. remains constant during the engagement, the desired LOS angle θF ,
The performance of the guidance law is also evaluated for the given by Eq. (5), changes due to the change in speed of the
scenario of intercepting the target at an impact angle of 0 deg but for interceptor. From Fig. 19c, it can also be noted that sliding mode
different initial interceptor heading angles. Results of these simula- occurs at point D and then impact angle criterion is satisfied at point

Fig. 22 Realistic interceptor against weaving target: a) trajectories in
X;Y space; b) trajectory in
V r ;V θ  space; c) occurrence of sliding mode;
d) interceptor lateral acceleration aM ; e) speed of interceptor; f) drag on interceptor.
 KUMAR, RAO, AND GHOSE 1129

E. Because V r < 0 at point E, interception is ensured according to the [4] Kumar, S. R., Rao, S., and Ghose, D., “Sliding Mode Guidance and
results of sliding mode capturability. Interceptor lateral acceleration, Control for All-Aspect Interceptor with Terminal Angle Constraints,”
speed of the interceptor, and drag acting on the interceptor, as shown Journal of Guidance, Control, and Dynamics, Vol. 35, No. 4, 2012,
in Figs. 19d–19f, respectively, show similar trends as in Figs. 16d– pp. 1230–1246.
doi:10.2514/1.55242
16f, respectively. [5] Zhang, Y., Sun, M., and Chen, Z., “Finite-Time Convergent Guidance
Law with Impact Angle Constraint Based on Sliding-Mode Control,”
3. Constant Maneuvering Targets Nonlinear Dynamics, Vol. 70, No. 1, 2012, pp. 619–625.
In this subsection, simulation results for constant maneuvering doi:10.1007/s11071-012-0482-3
targets are presented. The simulation parameters used for this case are [6] Di, Z., Sun, S., and Teo, K. L., “Guidance Laws with Finite Time
those listed in data set 1 in Table 1, but with aT
30 m∕s2 . In this Convergence,” Journal of Guidance, Control, and Dynamics, Vol. 32,
No. 6, 2009, pp. 1838–1846.
case, an impact angle of 90 deg is selected and the simulation results doi:10.2514/1.42976
are shown in Fig. 20. The points A and F in Fig. 20 denote the start [7] Kumar, S. R., Rao, S., and Ghose, D., “Non-Singular Terminal Sliding
and end of the engagement, respectively. Similar to the previous Mode Guidance and Control with Terminal Angle Constraints for Non-
cases, in Fig. 20, the points A, B, and C denote the time instants at Maneuvering Targets,” Proceedings of the 12th International Workshop
which boost, sustain, and endgame phases of the engagement starts. on Variable Structure Systems, IEEE, Piscataway, NJ, Jan. 2012,
From Fig. 20c, it can also be noted that sliding mode occurs at point D pp. 291–296.
and impact angle criterion is satisfied at point E. Because V r < 0 at [8] Qiaozhen, S., and Xiuyun, M., “Design and Simulation of Guidance
point E, interception is ensured according to the results of sliding Law with Angular Constraint Based on Non-Singular Terminal Sliding
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

mode capturability. The lateral acceleration profile, speed of the Mode,” Physics Procedia, Vol. 25, 2012, pp. 1197–1204.
[9] Rao, S., and Ghose, D., “Sliding Mode Control Based Terminal Impact
interceptor, and drag acting on the interceptor, as shown in Figs. 20d– Angle Constrained Guidance Laws Using Dual Sliding Surfaces,”
20f, show a similar trend to Figs. 16d–16f, respectively. Proceedings of the 12th International Workshop on Variable Structure
To evaluate robustness of the guidance law with a realistic Systems, IEEE, Piscataway, NJ, Jan. 2012, pp. 325–330.
interceptor model against target maneuvers, simulation with different [10] Shima, T., “Deviated Velocity Pursuit,” Proceedings of AIAA Guidance,
values of aT , given by aT
1, 3, 5, and 7g, has been carried out and Navigation, and Control Conference, AIAA Paper 2007-6782,
results are shown in Fig. 21. In this case, an impact angle of 120 deg is Aug. 2007, pp. 4364–4379.
selected. Figure 21a shows the trajectories of the interceptor and the [11] Shima, T., “Intercept-Angle Guidance,” Journal of Guidance, Control,
target in physical space, whereas Fig. 21b shows the required and Dynamics, Vol. 34, No. 2, 2011, pp. 484–492.
doi:10.2514/1.51026
interceptor lateral acceleration, which shows a trend similar to the
[12] Li, Z., Weimeng, S., and Zhiqiang, Z., “Control of Terminal
previous cases. From Fig. 21a, it can be seen that, even in presence of Engagement Geometry Using Variable Structure Guidance Law with
aT
7g, the guidance law is able to intercept the target with the Impact Angular Constraint,” Proceedings of the Second International
desired impact angle. Symposium on Systems and Control in Aerospace and Astronautics,
IEEE, Piscataway, NJ, Dec. 2008, pp. 1–4.
4. Weaving Targets [13] Hou, M., and Duan, G., “Integrated Guidance and Control of Homing
Missiles Against Ground Fixed Targets,” Chinese Journal of
To evaluate performance against weaving targets, simulation has
Aeronautics, Vol. 21, No. 2, April 2008, pp. 162–168.
been carried out with the initial condition listed in data set 1 in Table 1 [14] Lee, C. H., Kim, T. H., and Tahk, M. J., “Design of Impact Angle Control
except for the initial LOS angle θ0
0 deg and target executing Guidance Laws via High-Performance Sliding Mode Control,”
maneuver given by aT
30 sinπt∕10. In this case, an impact angle Proceedings of the Institution of Mechanical Engineers, Part G:
of 135 deg is selected and the results of the simulation are shown in Journal of Aerospace Engineering, Vol. 227, No. 2, 2013, pp. 235–253.
Fig. 22. Conclusions similar to the previous cases can also be drawn [15] Kim, M., and Grider, K. V., “Terminal Guidance for Impact Attitude
in this case. Angle Constrained Flight Trajectories,” IEEE Transactions on
Aerospace and Electronic Systems, Vol. 9, No. 6, 1973, pp. 852–859.
doi:10.1109/TAES.1973.309659
VI. Conclusions [16] Ryoo, C. K., Cho, H., and Tahk, M. J., “Optimal Guidance Laws with
Terminal Impact Angle Constraint,” Journal of Guidance, Control, and
In this paper, the finite time convergent guidance laws, which do
Dynamics, Vol. 28, No. 4, 2005, pp. 724–732.
not exhibit any singularity during implementation and are able to doi:10.2514/1.8392
intercept stationary, constant velocity, and maneuvering targets at [17] Ryoo, C. K., Cho, H., and Tahk, M. J., “Time-to-Go Weighted Optimal
desired impact angles with negligible miss distance, are proposed. Guidance with Impact Angle Constraints,” IEEE Transactions on
The proposed guidance laws are derived based on nonsingular Control Systems Technology, Vol. 14, No. 3, 2006, pp. 483–492.
terminal sliding mode control theory. In addition, it has been shown [18] Xu, X. Y., and Cai, Y. L., “Optimal Guidance Law and Control of Impact
that interception does not depend on the interceptor launch condition, Angle for the Kinetic Kill Vehicle,” Proceedings of the Institution of
that is, the initial heading angle error. To show effectiveness of the Mechanical Engineers, Part G: Journal of Aerospace Engineering,
guidance law, simulations with a realistic interceptor model, in which Vol. 225, No. 9, 2011, pp. 1027–1036.
doi:10.1177/0954410011402760
aerodynamic variations and gravity effect has been taken into
[19] Lee, Y. I., Kim, S. H., and Tahk, M. J., “Optimality of Linear Time-
consideration, is done. The extension of the proposed guidance laws to Varying Guidance for Impact Angle Control” IEEE Transactions on
a three-dimensional engagement scenario and inclusion of the Aerospace and Electronic Systems, Vol. 48, No. 4, 2012 pp. 2802–2817.
interceptor dynamics in the guidance law are possible areas of further doi:10.1109/TAES.2012.6324662
research. Additional performance measures such as impact time can [20] Park, B. G., Kim, T. H., and Tahk, M. J., “Optimal Impact Angle Control
also be included, which could prove useful in salvo attack applications. Guidance Law Considering the Seeker’s Field-of-View Limits,”
Proceedings of the Institution of Mechanical Engineers, Part G: Journal
of Aerospace Engineering, Vol. 227, No. 8, 2013, pp. 1347–1364.
References doi:10.1177/0954410012452367
[1] Utkin, V., Guldner, J., and Shi, J., Sliding Mode Control in Electro- [21] Taub, I., and Shima, T., “Intercept Angle Missile Guidance Under Time
Mechanical Systems, 2nd ed., Automation and Control Engineering Varying Acceleration Bounds,” Journal of Guidance, Control, and
Series, CRC Press, Boca Raton, FL, 2009. Dynamics, Vol. 36, No. 3, 2013, pp. 686–699.
[2] Yu, J., Zhang, Y., and Gu, W., “Approach to Integrated Guidance/ doi:10.2514/1.59139
Autopilot Design for Missiles Based on Terminal Sliding Mode Control,” [22] Lee, Y. I., Kim, S. H., and Tahk, M. J., “Analytic Solutions of Optimal
Proceedings of International Conference on Machine Learning and Angularly Constrained Guidance for First-Order Lag System,”
Cybernetics, IEEE, Piscataway, NJ, Aug. 2004, pp. 610–615. Proceedings of the Institution of Mechanical Engineers, Part G:
[3] Yu, J., Xu, Q., and Zhi, Y., “TSM Control Scheme of Integrated Journal of Aerospace Engineering, Vol. 227, No. 5, 2013, pp. 827–837.
Guidance/Autopilot Design for UAV,” Proceedings of 3rd International doi:10.1177/0954410012442617
Conference on Computer Research and Development, IEEE, [23] Lee, Y. I., Kim, S. H., Lee, J. I., and Tahk, M. J., “Analytic Solutions of
Piscataway, NJ, March 2011, pp. 431–435. Generalized Impact-Angle-Control Guidance Law for First-Order Lag
 1130 KUMAR, RAO, AND GHOSE

System,” Journal of Guidance, Control, and Dynamics, Vol. 36, No. 1, American Control Conference, IEEE Publ., Piscataway, NJ, June 2012,
2013, pp. 96–112. pp. 950–955.
doi:10.2514/1.57454 [35] Kim, T. H., Park, B. G., and Tahk, M. J., “Bias-Shaping Method for Biased
[24] Ratnoo, A., and Ghose, D., “State-Dependent Riccati-Equation-Based Proportional Navigation with Terminal-Angle Constraint,” Journal of
Guidance Law for Impact-Angle-Constrained Trajectories,” Journal of Guidance, Control, and Dynamics, Vol. 36, No. 6, 2013, pp. 1810–1816.
Guidance, Control, and Dynamics, Vol. 32, No. 1, 2009, pp. 320–326. doi:10.2514/1.59252
doi:10.2514/1.37876 [36] Ratnoo, A., and Ghose, D., “Impact Angle Constrained Interception of
[25] Kim, K. S., Jung, B., and Kim, Y., “Practical Guidance Law Controlling Stationary Targets,” Journal of Guidance, Control, and Dynamics,
Impact Angle,” Proceedings of the Institution of Mechanical Engineers, Vol. 31, No. 6, 2008, pp. 1816–1821.
Part G: Journal of Aerospace Engineering, Vol. 221, No. 1, 2007, doi:10.2514/1.37864
pp. 29–36. [37] Ratnoo, A., and Ghose, D., “Impact Angle Constrained Guidance
[26] Shaferman, V., and Shima, T., “Linear Quadratic Guidance Laws for Against Nonstationary Nonmaneuvering Targets,” Journal of
Imposing a Terminal Intercept Angle,” Journal of Guidance, Control, Guidance, Control, and Dynamics, Vol. 33, No. 1, 2010, pp. 269–275.
and Dynamics, Vol. 31, No. 5, 2008, pp. 1400–1412. doi:10.2514/1.45026
doi:10.2514/1.32836 [38] Lee, C. H., Kim, T. H., and Tahk, M. J., “Interception Angle Control
[27] Song, T. L., and Shin, S. J., “Time-Optimal Impact Angle Control for Guidance Using Proportional Navigation with Error Feedback,” Journal
Vertical Plane Engagements,” IEEE Transactions on Aerospace and of Guidance, Control, and Dynamics, Vol. 36, No. 5, 2013, pp. 1556–
Electronic Systems, Vol. 35, No. 2, 1999, pp. 738–742. 1561.
[28] Lee, C. H., Kim, T. H., Tahk, M. J., and Whang, I. H., “Polynomial doi:10.2514/1.58454
Guidance Laws Considering Terminal Impact Angle and Acceleration [39] Yu, X., and Man, Z., “Variable Structure Systems with Terminal Sliding
Downloaded by UNIVERSITY OF NEWCASTLE on December 27, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.62737

Constraints,” IEEE Transactions on Aerospace and Electronic Systems, Modes,” Variable Structure Systems: Towards the 21st Century, edited
Vol. 49, No. 1, 2013, pp. 74–92. by Yu, X., and Xu, J. X., Vol. 274, Lecture Notes in Control and
doi:10.1109/TAES.2013.6404092 Information Sciences, Springer, Berlin, 2002, pp. 109–127.
[29] Kim, T. H., Lee, C. H., and Tahk, M. J., “Time-to-Go Polynomial [40] Zhao, D., Li, S., and Gao, F., “New Terminal Sliding Mode Control for
Guidance with Trajectory Modulation for Observability Enhancement,” Robotic Manipulators,” International Journal of Control, Vol. 82,
IEEE Transactions on Aerospace and Electronic Systems, Vol. 49, No. 10, 2009, pp. 1804–1813.
No. 1, 2013, pp. 55–73. doi:10.1080/00207170902769928
doi:10.1109/TAES.2013.6404091 [41] Yu, X., Zhihong, M., Feng, Y., and Guan, Z., “Nonsingular Terminal
[30] Dhabale, A., and Ghose, D., “Impact Angle Constraint Guidance Law Sliding Mode Control of a Class of Nonlinear Dynamical Systems,”
Using Cubic Splines for Intercepting Stationary Targets,” Proceedings Proceedings of 15th Triennial World Congress of the International
of AIAA Guidance, Navigation, and Control Conference and Exhibit, Federation of Automatic Control, Elsevier Science Ltd., Kidlington,
AIAA Paper 2012-4689, Aug. 2012. Oxford, U.K., July 2002, pp. 431–435.
[31] Kim, B. S., Lee, J. G., and Han, H. S., “Biased PNG Law for Impact with [42] Feng, Y., Yu, X., and Man, Z., “Non-Singular Terminal Sliding Mode
Angular Constraint,” IEEE Transactions on Aerospace and Electronic Control of Rigid Manipulators,” Automatica, Vol. 38, No. 12, 2002,
Systems, Vol. 34, No. 1, 1998, pp. 277–288. pp. 2159–2167.
[32] Jeong, S. K., Cho, S. J., and Kim, E. G., “Angle Constraint Biased PNG,” doi:10.1016/S0005-1098(02)00147-4
Proceedings of the Fifth Asian Control Conference, IEEE, Piscataway, [43] Zarchan, P., Tactical and Strategic Missile Guidance, 5th ed., Progress
NJ, July 2004, pp. 1849–1854. in Astronautics and Aeronautics Series, AIAA, Reston, VA, 2007,
[33] Erer, K. S., and Merttopçuoğlu, O., “Indirect Control of Impact Angle pp. 21, 433.
Against Stationary Targets Using Biased Pure Proportional Navigation,” [44] Shneydor, N. A., Missile Guidance and Pursuit, Ellis Horwood,
Journal of Guidance, Control, and Dynamics, Vol. 35, No. 2, 2012, Chichester, England, U.K, 1998, pp. 104–105.
pp. 700–703. [45] Kee, P. E., Dong, L., and Siong, C. J., “Near Optimal Midcourse
doi:10.2514/1.52105 Guidance Law for Flight Vehicle,” Proceedings of AIAA 36th Aerospace
[34] Akhil, G., and Ghose, D., “Biased PN Based Impact Angle Constrained Sciences Meeting and Exhibit, AIAA, Reston, VA, Jan. 1998, pp. 1–11;
Guidance Using a Nonlinear Engagement Model,” Proceedings of IEEE also AIAA Paper 1998-0583.