You are on page 1of 47

Journal Pre-proof

Integrated guidance and control for missile with narrow field-of-view


strapdown seeker

Jiayi Tian, Neng Xiong, Shifeng Zhang, Huabo Yang, Zhenyu Jiang

PII: S0019-0578(20)30258-5
DOI: https://doi.org/10.1016/j.isatra.2020.06.012
Reference: ISATRA 3623

To appear in: ISA Transactions

Received date : 18 March 2019


Revised date : 2 June 2020
Accepted date : 17 June 2020

Please cite this article as: J. Tian, N. Xiong, S. Zhang et al., Integrated guidance and control for
missile with narrow field-of-view strapdown seeker. ISA Transactions (2020), doi:
https://doi.org/10.1016/j.isatra.2020.06.012.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the
addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive
version of record. This version will undergo additional copyediting, typesetting and review before it
is published in its final form, but we are providing this version to give early visibility of the article.
Please note that, during the production process, errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier Ltd on behalf of ISA.


Journal Pre-proof

Integrated Guidance and Control for Missile with Narrow


Field-of-View Strapdown Seeker

of
Jiayi Tian

pro
China Aerodynamics Research and Development Center, Mianyang 621000
China

College of Aerospace Science and Engineering, National University of Defense


Technology, Changsha,410073 China

Neng Xiong
re-
China Aerodynamics Research and Development Center, Mianyang 621000
China

Shifeng Zhang, Huabo Yang and Zhenyu Jiang1

College of Aerospace Science and Engineering, National University of Defense


lP
Technology, Changsha,410073 China
rna
Jou

1 Corresponding author: jiang_zhenyu@hotmail.com


Journal Pre-proof

of
Integrated Guidance and Control for Missile with
Narrow Field-of-View Strapdown Seeker

pro
Abstract

Due to the removal of the mechanically stable platform in the conventional


gimbaled seeker, the strapdown seeker’s measurement is coupled with the mis-
re-
sile body attitude motion, such that the inertial line-of-sight (LOS) angular
rate required to implement traditional guidance laws can not be measured, and
the field-of-view (FOV) limit must be considered when designing guidance and
control systems for a strapdown homing missile. To address these practical
lP
problems, an integrated guidance and control (IGC) controller with considering
the FOV limit is proposed in this paper. A novel IGC model is first derived
based on the body-LOS (BLOS) angle that a strapdown seeker can directly mea-
sure, and then an IGC controller is designed using the dynamic surface control
technique. A great merit of this design is that the inertial LOS angle and its
rna

angular rate are not needed, and thus the filters/estimators required to extract
this guidance information in previous studies can also be canceled. Next, by
using the output to input saturation transformation (OIST) technique, the FOV
limit, which is always considered as a state/output constraint, is transformed to
a time-varying boundary limitation on the control input, and then is handled
Jou

simultaneously with the actuator saturation constraint. Finally, extensive nu-


merical simulations against both stationary and moving targets are performed
to fully demonstrate the efficiency of the proposed IGC controller.
Keywords: Integrated guidance and control (IGC), body-LOS (BLOS) angle,
field-of-view (FOV) limit, output to input saturation transformation (OIST)

Preprint submitted to ISA Transactions June 2, 2020


Journal Pre-proof

of
1. Introduction

Replacing the conventional gimbaled seeker with a strapdown seeker brings


the homing system numerous benefits, such as compact structure, high reli-

pro
ability, unlimited tracking rate, and significant cost saving[1]. Whereas, the
5 use of the strapdown seeker makes it more challenging to design guidance and
control systems for a strapdown homing missile. Owing to the lack of a me-
chanically stable platform to isolate the missile body rotation, the strapdown
seeker’s measurement is always contaminated by the missile attitude informa-

10
re-
tion. Thus what can be directly measured and fed to the guidance system by
the strapdown seeker is no longer the commonly-required inertial line-of-sight
(LOS) angular rate but a body-LOS (BLOS) angle[2]. Moreover, compared to
the gimbaled seeker, the strapdown seeker has a relatively narrower total field-
of-view (FOV), which is more likely to cause the target to be lost during the
lP
homing phase, resulting in a larger miss distance or even mission failure[3].
15 To handle the FOV limit, a considerable number of works founded in the
literature have their foundation within the framework of the optimal control
theory. An early study in this category is that of Xin et al.[4], where a closed-
rna

form optimal guidance law considering the FOV limit was designed using the
θ − D nonlinear control method. Based on the optimal control theory with state
20 inequality constraint, Park et al. proposed an optimal guidance law to address
the problem of FOV limit in [5], and this law was further improved in [6] to
make the control energy redistributed along the flight trajectory, thereby zero-
ing the acceleration command at the instant of impact to maximize the warhead
Jou

effect. In [7], an optimal guidance law was assumed as a polynomial function of


25 time-to-go and determined the coefficients of the guidance command based on
the worst homing scenario to guarantee the FOV limit not to be violated. In re-
cent, the varying coefficient of weighted optimal control theory was employed to
handle the FOV limit in [8], where changeable weighting coefficients provided an
additional degree of freedom to shape the missile trajectory while maintaining
30 the lock-on condition. Without claiming optimality, a great number of propor-

2
Journal Pre-proof

of
tional navigation guidance (PNG) laws focusing on the FOV limit were proposed
in previous studies. Enhanced by switching gains/biases, these improved PNG
laws are designed to maintain the constant look angle in the first phase and

pro
achieving the desired impact angle in the second phase. In early studies[9, 10],
35 the navigation gain was selected as N = 1 for the initial phase and N ≥ 2 for
the final phase. Tekin and Erer firstly proposed a numerical algorithm to solve
the values of navigation gains at midcourse and terminal stages in [11]. In [12],
Ratnoo discussed the choice of navigation gains to achieve all possible impact
angles without violating the FOV limit. The pioneering examples of the biased
40
re-
PNG law were those of Park et al., wherein the PNG laws were enhanced with
a bias-shaping addition to intercept stationary[13] and moving[14] targets at
a desired impact angle while satisfying the FOV limit and acceleration limita-
tion. Also, some advanced control methodologies were employed to address this
lP
problem, such as sliding mode control[15], backstepping technique[16], state de-
45 pendent riccati equation method[17], barrier Lyapunov function (BLF)[18], and
nonlinear mapping[19].

LOS to Target
rna

Look Angle Missile Velocity

Angle of Attack
Airframe Centerline
(Optic-axis)
BLOS Angle
Fixed Reference
Jou

Figure 1: Longitudinal seeker angular configuration.

Note that the BLOS angle measured by the strapdown seeker is coupled with
the missile body attitude motion, it is, therefore, necessary to tackle the problem
of the FOV limit with consideration of the attitude control. However, almost all
50 existing studies on this problem just focus on the guidance law design as shown
in [4–19]. To simplify the problem, the angle of attack of the missile is assumed
in these guidance laws to be sufficiently small in the homing phase, thus making

3
Journal Pre-proof

of
it possible to approximate the BLOS angle by a look angle, which is defined as
the angle between the missile velocity and the missile-target LOS as depicted
55 in Fig. 1. In this way, the missile attitude dynamics is neglected and the FOV

pro
limit can be addressed easily just within the process of guidance design[20]. But
this assumption is too ideal in engineering practice[20, 21]. As pointed out by
Paul[22], the tactical missile maneuvering in the atmosphere generally changes
the magnitude and direction of aerodynamic force acting on itself by adjusting
60 the attitude to realize a precise attack against the target, which means that the
angle of attack is the main factor in trajectory shaping and thus cannot remain
re-
small during the entire homing phase. Therefore, despite claiming to satisfy
the FOV limit, these advanced guidance laws may still cause the violence of the
lock-on condition in some cases.
65 Another assumption required to be made to derive the aforementioned guid-
lP
ance laws is the inertial LOS angle and/or angular rate are available. To ensure
that these guidance laws work properly when applied to the strapdown hom-
ing missile, it is, therefore, necessary to add a filter/estimator to the guidance
loop to extract or estimate the inertial guidance information, such as “α −
70 β” filter[23], Extend/Unscented Kalman Filter[24, 25], linear/nonlinear hybrid
rna

differentiator[26], nonlinear tracking differentiator[27], disturbance observer[28]


and extended state observer[29], to name a few. It is worth noting that a ba-
sic prerequisite of implementing these filters/estimators is to ensure sufficient
observability through the homing phase[2]. With few exceptions[2, 30, 31], how-
75 ever, existing guidance laws focus little on the observability. The missile tra-
jectory shaped by these advance guidance laws may be capable of zeroing final
Jou

miss distance[4, 5], realizing specified direction/time attack[16, 20, 32], and sat-
isfying various physical constraints[7, 11, 14], but could not optimize or even
decrease the observability. A good example of this is the PNG law, which gen-
80 erally nullifies inertial LOS angular rate to correct a missile heading toward a
collision course but, maximizing the LOS angular rate is required to optimize
the observability[2, 30]. Hence, it gives rise to the need of designing a novel guid-
ance law that places no reliance on the inertial guidance information, thereby

4
Journal Pre-proof

of
fundamentally avoiding this contradiction.
85 As figure out above, the existing studies on the issue of the FOV limit
mainly focus on the guidance law design and can not guarantee to satisfy the

pro
FOV limit in practice for above unreasonable assumptions. Motivated by these
limitations, an integrated guidance and control (IGC) scheme with considering
the FOV limit is proposed in this paper for the strapdown homing missile.
90 An IGC model that does not require the inertial guidance information but
directly takes the BLOS angle as a system state is derived first, which enables
the FOV limit to be transformed to a state constraint and tactfully avoids to
re-
design a sophisticated filter/estimator appended to the guidance loop. Based
on the dynamic surface control technique, an IGC controller is then designed
95 for the novel IGC model to fully exploit the synergistic relationship between
the guidance and control loops, and thus the unreasonable assumption that
lP
the angle of attack is nearly zero is no longer to be made. Finally, the FOV
limit, which has been modeled as a system constraint, is further transformed to
a time-varying boundary limitation on the nominal control input obtained by
100 the IGC controller using the output to input saturation transformation (OIST)
technique[33], thus guaranteeing a precise attack without violence of the FOV
rna

limit.
The exact contributions of this paper over previous works are concluded as
follows:

105 1. Different from almost all previous studies focusing on the guidance law
design, an IGC scheme is proposed to address the FOV limit in this pa-
Jou

per, thus avoiding the unreasonable assumption that the angle of attack
remains zero during the homing phase.
2. Owing to the novel IGC model, the synthesized IGC law no longer requires
110 the inertial guidance information but just relies on the BLOS angle, thus
allowing to cancel the additional sophisticated filters/estimators appended
to the guidance loop.
3. By employing the OIST technique, the FOV limit is transformed to a

5
Journal Pre-proof

of
time-varying input constraint and then is handled with the saturation
115 constraint on control input in an integrated manner. This distinguishes
the proposed IGC law from a few relevant designs to address this problem

pro
using the IGC technique[21, 34, 35], in which the FOV limit is treated as
a state constraint and then is tackled by the Barrier Lyapunov Function
(BLF).

120 The remainder of this paper is organized as follows. First, a novel detailed
IGC design model is derived in Sec. 2. Then, the proposed IGC law is introduced
re-
in detail in Sec. 3, and the stability proof is also stated. Next, both performance
comparisons and Monte-Carlo analysis are implemented in Sec. 4 to demonstrate
the effectiveness of the proposed IGC law. Finally, this paper is finished with a
125 conclusion in Sec. 5.
lP
2. The IGC Model

Consider a planar engagement between a moving target and a roll stabilized


axisymmetric missile mounted on a strapdown seeker as shown in Fig. 2.

Y VT
rna

Xb T
A
X
T
A LOS
V Y TT
NX O
AX N T
AY K
D R
M O
NY T
Jou

M X
Figure 2: Two dimensional engagement geometry for the strapdown missile.

In inertial coordinate system M XY , R denotes the relative distance along


130 missile-target LOS from the missile M to the target T . AX , AY , and NX , NY
denote the missile acceleration along and perpendicular to its velocity V and

6
Journal Pre-proof

of
−−−→
body axis M Xb , respectively; and the target acceleration along and perpendic-
ular to its velocity VT is denoted as ATX , ATY , respectively. Under the assumption
that the seeker’s optic-axis centerline is aligned with the missile body axis, the

pro
135 LOS makes an angle of λ with respect to the fixed reference known as the iner-
tial LOS angle, and makes an angle of η with respect to the body axis known
as BLOS angle, respectively. Angles ϕ, θ, and α denote missile pitch angle,
flight path angle, and angle of attack, respectively. θT denotes target flight
path angle. All angles are defined as positive in the counterclockwise direction.

140 re-
Remark 1. Nearly all of the previous studies assume that the missile angle of
attack α is small enough and the missile attitude dynamics can be neglected, so
the BLOS angle η can be approximated by the look angle κ between the LOS and
the missile velocity, thereby addressing the problem of FOV limit just within the
process of guidance law design. However, this widely-used assumption is a bit
lP
145 unreasonable for engineering practice despite the fact that it greatly simplifies
this intractable problem. In fact, as pointed out by Paul[22], a tactical missile
heavily depends on the variation of the angle of attack to change the magnitude
and direction of the aerodynamic force acting on itself to respond to the guidance
rna

command. Strictly speaking, therefore, previous studies only addressed the look
150 angle constraint rather than the FOV limit. To satisfy this strict restriction,
the FOV limit must be considered both in guidance and control system design.

According to Fig. 2, the missile kinematics equations are given by

Rλ̇ = VT sin (θT − λ) − V sin (θ − λ) (1)


Jou

Ṙ = VT cos (θT − λ) − V cos (θ − λ) (2)

V̇ = AX V̇T = ATX (3)

V θ̇ = AY VT θ̇T = ATY (4)

Differentiating Eq. (1) and substituting Eqs. (2)-(4) yields

Rλ̈ = −2Ṙλ̇ − AX sin (θ − λ) − AY cos (θ − λ)

+ ATX sin (θT − λ) − ATY cos (θT − λ) (5)

7
Journal Pre-proof

of
Regard the target acceleration as an unknown disturbance dT , for it can
hardly be measured accurately in practice. Thus, Eq. (5) is rewritten as

Rλ̈ = −2Ṙλ̇ − AX sin (θ − λ) − AY cos (θ − λ) + dT (6)

pro
Assumption 1. During the homing phase t ∈ [t0 , tf ], the target acceleration is

always bounded. Let supt∈[t0 ,tf ] ATX (t) = ĀTX , supt∈[t0 ,tf ] ATY (t) = ĀTY , then
dT satisfies

|dT | = ATX sin (θT − λ) − ATY cos (θT − λ) ≤ ĀTX + ĀTY = DT (7)
re-
where DT is a positive constant.

The missile acceleration can be given by

AX = ax − g sin θ (8)
lP
AY = ay − g cos θ (9)

where ax and ay are missile overloads along and perpendicular to the missile
velocity. g is the acceleration of gravity.
As shown in Fig. 2, the missile angle of attack can be expressed in terms of
rna

the missile body axis and flight path angle according to

α=ϕ−θ (10)

Substituting Eqs. (8) to (10) into Eq. (6) yields

Rλ̈ = −2Ṙλ̇ − sin (ϕ − λ) (ax cos α + ay sin α)


Jou

+ cos (ϕ − λ) (ax sin α − ay cos α) + g cos λ + dT (11)

Note that the overloads ax and ay satisfy

ax cos α + ay sin α =nx (12)

ay cos α − ax sin α =ny (13)

155 where nx and ny are missile overloads along and perpendicular to the body axis,
respectively.

8
Journal Pre-proof

of
Combining Eqs. (11) to (13) yields

2Ṙλ̇ nx ny g
λ̈ = − − sin (ϕ − λ) − cos (ϕ − λ) + cos λ + dT (14)
R R R R

pro
In longitudinal plane, the missile attitude dynamics is given by

ϕ̇ =ωz (15)
Mzα qSLα Mzω̄z qSL2 ωz M δz qSLδz
ω̇z = + + z + dω (16)
Iz V Iz Iz

where ωz is the pitch rate and δz is the actuator deflection angle for pitch con-
re-
trol. Structural parameters S, L, and Iz denote the reference area ,the reference
length, and the moment of inertial, respectively. Aerodynamic coefficients Mzα ,
160 Mzω̄z , and Mzδz denote the pitch moment derivatives with respect to the angle
of attack, the non-dimensional pitch rate, and the actuator deflection angle,
1
respectively. q = ρV 2 is the dynamic pressure, where ρ is the air density. dω
lP
2
is a lumped disturbance, possibly including external atmospheric disturbances,
internal parametric perturbations in structural parameters and aerodynamic co-
165 efficients, as well as unmodeled dynamics resulted from neglected cross-coupling
effects and truncated errors introduced by the linearization approximation.
rna

Assumption 2. In practice, uncertainties resulted from external atmospheric


disturbances and internal parametric perturbations are generally bounded. There-
fore, according to Ref.[21], it is reasonable to assume that the lumped disturbance
dω is bounded, namely
|dω | ≤ Dω (17)
Jou

where Dω is a positive constant.

From Fig. 2, it is observed that the BLOS angle can be expressed in terms
of the LOS angle and the pitch angle according to

η =λ−ϕ (18)

9
Journal Pre-proof

of
Subtracting Eqs. (14) and (16), and then substituting Eq. (18) yields

2Ṙλ̇ nx ny g
η̈ = − + sin η − cos η + cos (ϕ + η) + dT
R R R R

pro
M α qSLα Mzω̄z qSL2 ωz M δz qSLδz
− z − − z − dω (19)
Iz V Iz Iz

Note that the closing velocity Ṙ can also hardly be measured accurately in
−2Ṙλ̇
practice, so the term is regarded as an unknown disturbance dṘ and
R
thus, Eq. (19) is rewritten as

nx ny g M α qSLα
η̈ =
R
sin η −
R
re-
cos η + cos (ϕ + η) − z
R
ω̄z
M qSL ωz
− z
V Iz
2
− z
Iz
δz
M qSLδz
Iz
+ dT − dω + dṘ (20)

Assumption 3. When the missile is close to the target enough, the strapdown
seeker will operate in its dead zone and then the IGC law will stop work. Con-
lP
sequently, only the situation of R ≥ Rf should be considered when designing the
IGC law, where Rf is a small positive constant related to the dead zone of the
strapdown seeker. Hence, it is reasonable to assume that dṘ is bounded, namely
rna

|dṘ | ≤ DṘ (21)

where DṘ is a positive constant.

Because of the existence of mechanical components, the actuator saturation


constraint should be considered when designing the IGC law. Thus, Eq. (20) is
rewritten as
Jou

nx ny g M α qSLα
η̈ = sin η − cos η + cos (ϕ + η) − z
R R R Iz
Mzω̄z qSL2 ωz Mzδz qSL
− − sat(δz ) + dT − dω + dṘ (22)
V Iz Iz

where sat(δz ) denotes the nonlinear actuator saturation characteristic and is of


the form 
sgn(δz )δzm , |δz | ≥ δzm
sat(δz ) =

δz , |δz | < δzm

10
Journal Pre-proof

of
where δzm is the maximum actuator deflection angle.
As shown in Fig. 3, the saturation function sat(δz ) in the IGC model (22)
leads to a sharp corner when |δz | = δzm , thereby making subsequent IGC law

pro
design and stability analysis more complicated. To tackle it, the saturation
function sat(δz ) is approximated by a smooth tangent function T (δz ), which is
given by
 
δz eδz /δzm − e−δz /δzm
T (δz ) = δzm × tanh = δzm (23)
δzm eδz /δzm + e−δz /δzm

re-
lP

Figure 3: Graph of saturation function sat(δz ) and smooth tangent function T (δz ).

The approximation error is defined as d(δz ) = sat(δz ) − T (δz ), which is a


bounded error function
rna

|d(δz )| = | sat(δz ) − T (δz )| ≤ δzm (1 − tanh(1)) = 0.2384δzm (24)

Then Eq. (22) is rewritten as

nx ny g M α qSLα Mzω̄z qSL2 ωz


η̈ = sin η − cos η + cos (ϕ + η) − z −
R R R Iz V Iz
Jou

Mzδz qSL Mzδz qSL


− T (δz ) − d(δz ) + dT − dω + dṘ (25)
Iz Iz

Define a total disturbance as

Mzδz qSL
d = dT − dω + dṘ − d(δz ) (26)
Iz

Based on Assumption 1-3, it is proved that the total disturbance d is bounded

11
Journal Pre-proof

of
as

M δz qSL
|d| = dT − dω + dṘ − z d(δz )
Iz
δ
M z qSL

pro
≤ |dT | + |dω | + |dṘ | + 0.2384 z δzm

Iz (27)
≤ DT + Dω + DṘ + 0.2384∆δzm

=D
δ δ
Mz z qSL Mz z qSL

where ∆ = supt∈[t0 ,tf ] along
170
1 Iz , is the maximum value of Iz
re-
the missile trajectory. D is a positive constant.
Without loss of generality, the following assumption is made

Assumption 4. The rate of change of total disturbance d is bounded, namely



˙
d(t) ≤ Ld (28)
lP
where Ld is a positive constant.

Remark 2. Although Assumption 4 seems too strong, it is realistic from a


175 practical point of view. As stated in Ref. [36], the actual controlled physi-
rna

cal quantities, for example, the target acceleration (dT ), the angular rate (dṘ
when considering the dead zone of the strapdown seeker), and the control in-
put (d(δz )), are all physically bounded in practice. Also, the possible external
disturbance dω is usually modeled as bounded and Lipschitz continuous func-
180 tion. Therefore, it is reasonable to assume that the time derivative of the total
disturbance d is bounded in Assumption 4.
Jou

With Eq. (26), Eq. (25) is further rewritten as

nx ny g M α qSLα
η̈ = sin η − cos η + cos (ϕ + η) − z
R R R Iz
ω̄z 2
M qSL ωz M δz qSL
− z − z T (δz ) + d (29)
V Iz Iz

Choosing the system state x1 = η, x2 = η̇ and control input u = δz , the

12
Journal Pre-proof

of
final IGC model according to Eq. (29) is obtained as



ẋ1 = x2


(30)

pro
ẋ2 = f (x) + g(x)T (u) + d




 u̇ = −cu + w

where c is a positive constant and w is an auxiliary control signal to be designed.


Nonlinear functions f (x), g(x) are

nx ny g M α qSLα Mzω̄z qSL2 ωz


f (x) = sin η − cos η + cos (ϕ + η) − z −
R R R Iz V Iz

g(x) = −
Mzδz qSL
Iz
re-
Inspired by Ref. [37], the third equation is artificially introduced to handle
the saturation function sat(δz ) using the Nussbaum function.
Considering the FOV limit, the BLOS angle (namely system state x1 ) during
lP
the homing phase should satisfy

|x1 | ≤ kc (31)

where kc is the maximum BLOS angle of the strapdown seeker.


rna

185 Generally, the target should be locked by the strapdown seeker at the be-
ginning of the homing phase. Hence, it is reasonable to make the following
assumption.

Assumption 5. The initial value of system state x1 satisfies

|x1 (t0 )| ≤ kc (32)


Jou

Hereto, the novel IGC model (30) with considering the FOV limit and the
actuator saturation is derived.

190 Remark 3. Different from preexisting IGC models[21, 38], this novel IGC
model (30) first takes both the gravity and varying missile velocity into account.
Also, commonly-used simplified linear aerodynamic force model L = CLα qSα in
previous studies is replaced by the missile axial and normal overloads nx , ny

13
Journal Pre-proof

of
that can be directly and precisely measured in reality, thereby allowing a pre-
195 cise modeling of forces imposed on the missile body. Moreover, although the
angle-of-attack is always selected as a system state to build the IGC model with

pro
a strict feedback structure, the proposed IGC model (30) does not follow this
tradition due to the fact that the attack of attack is not measurable in practice,
and only uses its nominal value to compute the pitching moment. Owing to
200 these improvements, the proposed IGC model is no longer limited to only the
unpowered flight phase where aerodynamic forces change slowly, but is applica-
ble to the entire homing phase. This distinguishes the IGC model (30) from all
existing IGC models.
re-
Remark 4. Note that the IGC model (30) places no reliance on the inertial
205 LOS angle and/or its angular rate. Thus no filter/estimator needs to be de-
signed to distill this inertial guidance information. Also, more advanced control
lP
methods can be applied to the IGC model (30) without considering the target
observability.

3. A Novel IGC Law Design and Analysis


rna

210 3.1. The control objective

The zero-effort miss (ZEM) is defined to be the distance a missile would


miss a target if the target continued along its present course and the missile
made no further corrective maneuvers. According to Ref. [39], the ZEM can be
described as
R2 λ̇
Jou

ZEM = p (33)
Ṙ2 + R2 λ̇2
It can be founded from Eq. (33) that a sufficient condition for zeroing the
ZEM is to nullify the inertial LOS angular rate λ̇. This is also the guidance
principle of the parallel approaching law. From Fig. 2, the inertial LOS angle
can be expressed in terms of the BLOS angle and pitch angle according to

λ=η+ϕ (34)

14
Journal Pre-proof

of
In longitudinal plane, the missile pitch angle satisfies

ϕ̇ = ωz (35)

Combining Eqs. (34) and (35), the sufficient condition λ̇ → 0 can be equally

pro
transformed to
η̇ → −ωz (36)

Hence, the control objective of this paper is summarized as: for the IGC
model (30) 

 ẋ1 = x2



re-




ẋ2 = f (x) + g(x)T (u) + d
 u̇ = −cu + w
(37)

subject to the FOV limit (31)

|x1 | ≤ kc (38)
lP
and the control input saturation

|u| ≤ δzm (39)

find an actuator defection angle u(t) on the time interval t ∈ [t0 , tf ] to force the
BLOS angular rate to track the negative pitch rate (Eq.(36)), namely
rna

x2 → x2d (40)

where the reference signal x2d = −ωz .

3.2. Nonlinear tracking differentiator design


To obtain the rate of change of the reference signal x2d , the following non-
Jou

linear tracking differentiator is designed according to Ref. [40] as




ν̇1 = ν2
  (41)
 ν2 |ν2 |
ν̇2 = −r sgn ν1 − x2d +
2r
where under the acceleration limitation |ν̈1 | ≤ r, ν1 tracks x2d at the fastest
speed according to the time-optimal control theory, and ν2 is accordingly a
215 sufficiently accurate approximation of ẋ2d . r > 0 is the maximum allowable
acceleration of the nonlinear tracking differentiator (41).

15
Journal Pre-proof

of
3.3. Homogeneous observer design

Given that the total disturbance d and the BLOS angular rate η̇ (namely
x2 ) are both unknown, a homogeneous observer is designed to estimate that as:

pro

 β3 α
ż1 = z2 −
 |z1 − x1 | 3 sgn(z1 − x1 )

 ε

β2 α
ż2 = z3 − 2 |z1 − x1 | 2 sgn(z1 − x1 ) + f (x) + g(x)T (u) (42)

 ε



ż = − β1 |z − x |α1 sgn(z − x )
3 1 1 1 1
ε3
where β1 > 0, β3 > 0, β2 > β1 /β3 , 0 < α1 < 1, α2 = (2α1 + 1) /3, α3 =

220
re-
(α1 + 2) /3, 0 < ε < 1 are the observer parameters to be designed. z1 , z2 and z3
are the estimations of system states x1 , x2 and total disturbance d, respectively.
Let e1 = z1 − x1 , e2 = z2 − x2 , e3 = z3 − d be the estimation error, then the
estimation error equation, according to Eqs. (30) and (42), is given by

lP
 β3 α

 ė1 = e2 − |e1 | 3 sgn(e1 )

 ε

β2 α
ė2 = e3 − 2 |e1 | 2 sgn(e1 ) (43)

 ε



ė = − β1 |e |α1 sgn(e ) − d˙
3 1 1
ε3
rna

Let τ = t/ε and consider the following coordinate transformation



$i (τ ) = εi−1 ei (t), i = 1, 2, 3
(44)

ζ(τ ) = ε3 d˙

then the transformed estimation error equation is



 d$1 α
 = $2 − β3 |$1 | 3 sgn($1 )
Jou


 dτ


d$2 α
= $3 − β2 |$1 | 2 sgn($1 ) (45)

 dτ



 d$3 = −β |$ |α1 sgn($ ) − ζ
1 1 1

Ignoring the uncertain term ζ and letting r1 = 1, r2 = (α1 + 2) /3, r3 =

16
Journal Pre-proof

of
(2α1 + 1) /3, the vector field F$ of the system (45) satisfies
 
α
cr2 $2 − β3 cr1 α3 |$1 | 3 sgn($1 )
 
 
F$ (cr1 $1 , cr2 $2 , cr3 $3 ) = cr3 $3 − β2 cr1 α2 |$1 |α2 sgn($1 )

pro
 
α
−β1 cr1 α1 |$1 | 1 sgn($1 )
  
α
cr1 +m 0 0 $2 − β3 |$1 | 3 sgn($1 )
  
  α 
= 0 cr2 +m 0  $3 − β2 |$1 | 2 sgn($1 )
  
α
0 0 cr3 +m −β1 |$1 | 1 sgn($1 )
(46)
re-
where m = (α1 − 1) /3 < 0. It follows from Eq. (46) that F$ is homogeneous
of degree m < 0 with respect to the dilation (r1 , r2 , r3 ).
The boundedness of the estimation error of the homogeneous observer (42)
can be guaranteed from the following lemma.
lP
Lemma 1. Based on Assumption 4, for system (30) and homogeneous observer
(42), there exist L = µLσd > 0, 0 < γ < 1/2, and σ = (1 − γ) /γ, such that

|e1 | ≤ Lε3σ , |e2 | ≤ Lε3σ−1 , |e3 | ≤ Lε3σ−2 (47)


rna

225 can be fulfilled in finite time. The proof can be found in Ref. [39].

3.4. Dynamic surface controller design

Noting that the IGC model (30) has a strict feedback structure, the dynamic
surface control method is employed int this subsection to design a basic IGC law
without considering the FOV limit. And the FOV limit is handled by means of
Jou

230 the OIST technique in the subsequent subsection.


Step 1. To track the reference signal x2d , define the sliding surface s1 for
system state x2 as
s1 = z2 − ν1 (48)

Remark 5. The system state x2 is replaced by its estimation z2 to build the


sliding surface s1 , for the BLOS angular rate η̇ (namely x2 ) cannot be measured
by the strapdown seeker in practice.

17
Journal Pre-proof

of
Define a quadratic Lyapunov function for s1 as

1 2
V1 = s (49)
2 1

pro
Differentiating V1 yields

V̇1 = s1 ṡ1
  (50)
β2 α2
= s1 z3 − 2 |z1 − x1 | sgn(z1 − x1 ) + f (x) + g(x)T (u) − ν2
ε

To make V̇1 negative definite, T (u) is chosen as a virtual control input x3d ,

x3d = −
1
g(x)

β2
re-
and a desired feedback control is designed according to Eq. (50) as

α2
z3 − 2 |z1 − x1 | sgn(z1 − x1 ) + f (x) − ν2 + k1 s1 + l1 sgn(s1 )
ε


(51)
where design parameters k1 > 0, l1 > 0.
lP
To avoid the problem of “explosion of the complexity”, a first-order filter
is introduced. let x3d pass through the following first-order filter with time
constant τ
τ ẋ3c + x3c = x3d (52)
rna

Define the tracking error of the first-order filter (52) as

µ = x3c − x3d (53)

Step 2. To track the virtual control input x3c , define the sliding surface s2
for T (u) as
s2 = T (u) − x3c (54)
Jou

Define a quadratic Lyapunov function for s2 as

1 2
V2 = s (55)
2 2

Differentiating V2 yields

V̇2 = s2 ṡ2
  (56)
= s2 T̄ (u) (−cu + w) − ẋ3c

18
Journal Pre-proof

of
where  
∂T (u) u
T̄ (u) = = sech2 (57)
∂u δzm
To make V̇2 negative definite, a desired feedback control law for the auxiliary

pro
control signal w is designed as:

w̄ = −k2 s2 + cT̄ (u)u + ẋ3c
(58)

w = N (χ)w̄

where design parameter k2 > 0. N (χ) is a Nussbaum function given by


 π 
re-

N = χ2 cos

 χ̇ = γs2 w̄
2
χ
(59)

235 where design parameter γ > 0.


Then the basic control input command uc satisfies
lP
u̇c = −cu + w (60)

3.5. Closed-loop system stability analysis

Before investigations on the closed-loop system stability, the following lemma


rna

is introduced first.

Lemma 2. Let V (t) and χ(t) denote smooth functions that are defined on
[0, tf ), and V (t) ≥ 0 for ∀t ∈ [0, tf ). Then V (t) and χ(t) must be bounded
on [0, tf ) if they satisfy following inequality equation
Z t
T  e−Kt
V (t) ≤ V (0)e−Kt + 1 − e−Kt + (ψN (χ) − 1) χ̇eKτ dτ (61)
Jou

K γ 0

where K, T and γ are positive constants, and ψ is a positive variable. N (χ) is


240 a Nussbaum gain function with respect to χ(t).

Define a quadratic Lyapunov function for the tracking error µ of the first-
order filter (52) as
1 2
V3 = µ (62)
2

19
Journal Pre-proof

of
Differentiating V3 yields

µ2 µ2 µ2 ẋ2
V̇3 = µµ̇ = − − µẋ3d ≤ − + + 3d (63)
τ τ 2 2

pro
The derivative of virtual control input is always continuous and bounded[38].

Then, let 2Λ = supt∈[t0 ,tf ] |ẋ3d |, equation (63) satisfies

µ2 µ2
V̇3 ≤ − + + Λ2 (64)
τ 2

Combining Eqs. (53) and (54), V̇1 (50) is revised as


 
β2 α2
re-
V̇1 =s1 z3 − 2 |z1 − x1 | sgn(z1 − x1 ) + f (x) + g(x)T (u) − ν2
ε
= − k1 s21 − l1 |s1 | + g(x)(s1 + µ)s1 (65)
 
∆ 2
≤ − k1 − s1 + s22 + µ2
2
lP
Combining feedback control laws (58), (59) and substituting Eq. (56), V̇2 is
revised as
 
V̇2 =s2 T̄ (u) (−cu + w) − ẋ3c
  
=s2 −k2 s2 + T̄ (u)N (χ) − 1 w̄ (66)
rna

 π  
2 1 ∂T (u) 2
= − k2 s2 + χ cos χ − 1 χ̇
γ ∂u 2

Define a Lyapunov function for the closed-loop system as


3
X 1 2 1 2 1 2
V = Vi = s + s + µ (67)
i=1
2 1 2 2 2
Jou

From Eqs. (63), (65) and (66), the derivative of the closed-loop Lyapunov
function satisfies
     
∆ 1 3
V̇ ≤ − k1 − s21 − 2
k2 − 1 s2 − − µ2
2 τ 2
 π  
2 1 ∂T (u) 2
+Λ + χ cos χ − 1 χ̇ (68)
γ ∂u 2

Choose design parameters k1 , k2 and the time constant of the first-order

20
Journal Pre-proof

of
filter τ to satisfy following inequalities
  
 K ∆

 ≤ k1 −

 2 2

  
K

pro
≤ k2 − 1 (69)

 2

  

 K 1 3
 ≤ −
2 τ 2
where K is a positive constant.
Then the derivative of the closed-loop Lyapunov function (68) is further
revised as  
π 
re-
V̇ ≤ −KV + Λ +
1 ∂T (u) 2
2
γ ∂u
χ cos
2
χ − 1 χ̇

Direct integrating of the differential inequality (70) yields


(70)

Λ2   e−Kt Z t  ∂T (u) π  
−Kt −Kt 2
V ≤ V (t0 )e + 1−e + χ cos χ − 1 χ̇eKτ dτ
K γ t0 ∂u 2
(71)
lP
Hereto, the following Theorem is established to guarantee the stability of
the closed-loop system.

Theorem 1. Assuming that Assumption 1-5 are satisfied, the IGC model (30)
245 under the proposed dynamic surface controller holds the following properties:
rna

1. Both sliding surfaces s1 , s2 are uniform ultimately bounded and converge into
a compact set
p
|si (t)| ≤ 2Vub , ∀t > 0 (72)

where i = 1, 2, Vub is a positive constant denoting the upper bound of the


closed-loop Lyapunov function V .
Jou

2. All closed-loop signals are bounded.

Proof. 1. In accordance with Lemma 2, equation (71) illustrates that closed-


loop Lyapunov function V is bounded on [t0 , tf ) as

V ≤ Vub (73)

Combining Eqs. (67) and (73) yields


1 2 1 2 1 2
s + s + µ = V ≤ Vub (74)
2 1 2 2 2

21
Journal Pre-proof

of
then
p
|si | ≤ 2Vub , i = 1, 2 (75)

2. According to Lemma 1 and considering that the closed-loop Lyapunov func-

pro
250 tion V is bounded, all closed-loop signals, including the estimation errors ei
(i = 1, 2, 3), sliding surfaces si (i = 1, 2), and tracking error of the first-order
filter µ, are all bounded. 

3.6. Addressing the FOV limit

In this subsection, the basic control input command uc is modified to satisfy


re-
255 the FOV limit by using the OIST technique.
For the IGC model (30), define a system output as

y := h(x) = x1 (76)
lP
Considering the constraint condition (31), the constrained output y should
be confined within an admissible interval for ∀t ≥ 0, namely
 
y(t) ∈ Ωy (t) = y(t), y(t) (77)
rna

where y(t) = −kc , y(t) = kc .


Before addressing the FOV limit with the OIST technique, the following
definition and lemma are introduced first.

Definition 1. Let M = [m1 , · · · , mn ] a vector of positive constants and let


y 0 (t) = y(t), y 0 (t) = y(t), then for i = 1, · · · , n,
  z}|{
Jou

˙
y i (t) = mi y i−1 (t) − y (i−1) (t) + y i−1 (t)
  z}|{ (78)
˙
y i (t) = mi y i−1 (t) − y (i−1) (t) + y i−1 (t)
h i
are defined as propagated bounds, and let Ωiy (t) = y i (t), y i (t) for ∀t ≥ 0.

260 Lemma 3. Suppose that the constrained output y is of relative degree k with
respect to the control input u, and initial values of derivatives up to k − th of
the constrained output y are all confined within the propagated bounds y (i) (0) ∈

22
Journal Pre-proof

of
Ωiy (0), i = 0, 1, · · · , k. Then for ∀t ≥ 0, the constrained output y can be confined
within the admissible interval y(t) ∈ Ωy (t) if the derivatives up to k − th of the
265 constrained output y is confined within the propagated bounds y (i) (t) ∈ Ωiy (t).

pro
The proof can be founded in Ref.[33].

Replacing the equivalent input T (u) in the IGC model (30) with the control
input u, for the actuator saturation has been addressed in the last subsection.
Then, it can be verified that the relative degree of the system output y with
respect to the control input u is k = 2 through simple calculation, namely
re-ÿ = L2f h + Lg Lf h · u (79)

where Lg Lf h = g(x), L2f h = f (x) + z3 with z3 replacing d.


According to Definition 1, the propagated bounds are calculated as
lP
y 1 = m1 (y 0 − y) + ẏ 0 = m1 (kc − x1 )
 
y 1 = m1 y 0 − y + ẏ 0 = m1 (−kc − x1 )
(80)
y 2 = m2 (y 1 − ẏ) + ẏ 1 = − (m1 + m2 ) z2 − m1 m2 x1 + m1 m2 kc
 
y 2 = m2 y 1 − ẏ + ẏ 1 = − (m1 + m2 ) z2 − m1 m2 x1 − m1 m2 kc
rna

where the system state x2 is replaced by its estimation z2 .


Combining Eqs. (79) and (80) yields the final following time-varying lower
and upper bounds of the dynamic surface control input uc to guarantee the
constrained condition y ∈ Ωy as (if g(x) > 0)

y 2 − L2f h − (m1 + m2 ) z2 − m1 m2 x1 + m1 m2 kc − f (x) − z3


uc = =
Jou

Lg Lf h g(x)
(81)
y 2 (t) − L2f h − (m1 + m2 ) z2 − m1 m2 x1 − m1 m2 kc − f (x) − z3
uc = =
Lg Lf h g(x)

when g(x) < 0, the expressions of uc and uc are exchanged.


Finally, the IGC law based on OIST with considering the FOV limit is

23
Journal Pre-proof

of
summarized as


 ν̇1 = ν2



  

 ν2 |ν2 |

pro
 2
 ν̇ = −r sgn ν1 − x 2d +

 2r



 ż1 = z2 − β3 |z1 − x1 |α3 sgn(z1 − x1 )



 ε



 β


2 α
ż2 = z3 − 2 |z1 − x1 | 2 sgn(z1 − x1 ) + f (x) + g(x)T (u)

 ε



 β1

 α
ż3 = − 3 |z1 − x1 | 1 sgn(z1 − x1 )

 ε





 s1 = z2 − ν1














x 3d = −
1
g(x)

z 3 −
β2
ε2
re-
|z1 − x 1 |
α2
sgn(z1 − x 1 ) + f (x) − ν 2 + k s
1 1 + l 1 sgn(s1 )





 x3c = −τ ẋ3c + x3d

 s2 = T (u) − x3c

  

lP

 u

 T̄ = sech 2

 δzm





 w̄ = −k2 s2 + cT̄ (u)u + ẋ3c



  

 N = χ2 cos π χ



 2




rna


 χ̇ = γs2 w̄





 w = N (χ)w̄







 u̇c = −cu + w





 − (m1 + m2 ) z2 − m1 m2 x1 + m1 m2 kc − f (x) − z3

 uc = (g(x) > 0)

 g(x)



 − (m1 + m2 ) z2 − m1 m2 x1 − m1 m2 kc − f (x) − z3

 uc =
g(x)
Jou

(82)
270 The configuration of the proposed IGC law is depicted in Fig. 4.

Remark 6. By using the OIST technique, the FOV limit is transformed into a
time-varying saturation constraint on the control input for the first time. This
distinguishes the proposed IGC law from previous studies that always treats the
FOV limit as a state constraint and addresses the problem with the BLF.

24
Journal Pre-proof

of
The IGC Law Based on OIST with Considering the FOV Limit
Nonlinear Tracking Differentiator
1   2
z 
  2 2 
2   r sgn  1  x2d  2r 
Longitudinal Missile Model   

pro
 X  V cos  1 , 2
 
 Y  V sin 
 T cos   D  mg sin  Dynamic Surface Controller
 V 
 m  s1  z2   1
  T sin   L  mg cos nx , n y , ,V , ,  z 
   x3d   1  z3   2 z1  x1  2 sgn  z1  x1   f  x    2  k 2 s2  l2 sgn  s2  
 mV  g  x   2 
 T sin   L  mg cos 
    z  x
 3c   
x 3c  x 3d
 mV  s2  T  u   x3c
 Mz 
 
 z I  2 u 
 z
 T  sech   
  zm 
 w   k 2 s2  cT  u  u  x3c
uc 
nx , n y , ,V , ,  z   
 N   cos  2  
2


Homogeneous Observer
3
 z1  z2   z1  x1 sgn  z1  x1 

2
3
re- z1 , z2 , z3

    s2 w
 w  N  w

 uc  cu  w

 u    m1  m2  z2  m1m2 x1  m1m2 kc  f  x   z3  g  x   0 
  c g  x
 z2  z3  2 z1  x1 sgn  z1  x1   f  x   g  x  T  u 
2

 
    m1  m2  z2  m1m2 x1  m1m2 kc  f  x   z3
  u 
z3   31 z1  x1 1 sgn  z1  x1 

 c g  x
  
lP
R, R,
Strapdown Seeker

Target Motion

Figure 4: Configuration of the IGC law based on OIST with considering the FOV limit
rna

275 4. Numerical simulations

In this section, the feasibility and efficacy of the proposed IGC law are
verified by numerical simulations.

4.1. Longitudinal missile model

A roll stabilized tail controlled axisymmetric missile mounted on a strapdown


Jou

seeker is considered in following simulations, whose nonlinear missile model is

25
Journal Pre-proof

of
given by[19] 

 Ẋ = V cos θ







 Ẏ = V sin θ

pro

 T cos α − D − mg sin θ


 V̇ =

m
T sin α + L − mg cos θ (83)

 θ̇ =

 mV



 T sin α + L − mg cos θ

 α̇ = ωz −

 mV



 Mz

ω̇z =
Iz
re-
where X, Y denote the missile coordinates with respect to the inertial coordinate
system. The missile thrust T is


500, t ≤ 0.5
T = (84)

0, t > 0.5
lP
The drag D, lift L and pitching moment Mz are modeled as



2
D = Cx0 qS + Cxα qSα2



L = Cyα qSα + Cyδz qSδz (85)




rna

 Mz = Mzα qSLα + Mzω̄z qSLω̄z + Mzδz qSLδz

where the maximum actuator deflection angle and angular rate are set as 20◦
280 and 40◦ /s with consideration of physical constraints.
The nominal structural parameters and aerodynamic coefficients are listed
in Table 1.
Jou

4.2. Comparative IGC law

To illustrate its superiority, the proposed IGC law is compared with a similar
design in the following simulations. Employing the extended state observer to
estimate unknown mismatched disturbances, this comparative IGC law based
on logarithmic-type BLF forces the BLOS angle to approach zero to intercept a
stationary target without violating the FOV limit. Its final control law described

26
Journal Pre-proof

of
Table 1: Nominal missile parameters

Structural parameters

pro
m Mass 4kg
Iy Moment of inertial 0.85kg · m2
S Reference area 0.013m2
L Reference length 1m
g Acceleration of gravity 9.8m/s2

Aerodynamic coefficients
re- 2
Cx0 = 0.18 Cxα = 7.65rad−2
Cyα = 15.63rad−1 Cyδz = 3.62rad−1
Mzα = −2.01rad−1 Mzω̄z = −0.05rad−1
Mzδz = −1.57rad−1
lP
in Ref. [34] is summarized as


 ei = zi0 − xi i = 1, 2, 3, 4







 ż10 = z11 − β10 e1 + x2


rna




 ż20 = z21 − β20 e2 + a22 x2 + a23 x3







 ż30 = z31 − β30 e3 + a33 x3 + x4





 ż40 = z41 − β40 e4 + a43 x3 + a44 x4 + b4 δz






 żi1 = −βi1 fal(ei , αi , σi )
 i = 1, 2, 3, 4





 s1 = x1


Jou






 x = −k1 s1 − l1 sgn(s1 ) − z11
 2d



 x2c = −τ2 ẋ2c + x2d
(86)



 s2 = x2 − x2c

 !



 1 |Ṙ| s1

 x3d = −a22 x2 − z21 − l2 sgn(s2 ) + ẋ2c − k2 s2 − 2

 a23 R kc − s21







 x3c = −τ3 ẋ3c + x3d






 s3
 = x3 − x3c





 x4d = −k3 s3 − a33 x3 − z31 − l3 sgn(s
27 3 ) + ẋ3c − a23 s2





 x4c

 = −τ4 ẋ4c + x4d





 s4 = x4 − x4c





 1 
 uc = −a43 x3 − a44 x4 − z41 − l4 sgn(s4 ) + ẋ4c − k4 s4 − s3
b4
Journal Pre-proof

of
where system states x1 = η, x2 = λ̇, x3 = α, x4 = ωz , control input u = δz ,
state coefficients

2Ṙ qSCyα
a22 = − a23 = −

pro
R mR
qSCyα qSLMzα
a33 =− a43 =
mV Iz
qSLMzω̄z L qSLMzδz
a44 = b4 =
Iz V Iz

and nonlinear function fal(e, α, σ)



re- 
|e|α sgn(e), |e| > σ
fal(e, α, σ) = e

 , |e| ≤ σ
σ 1−α
The configuration of the comparative IGC law based on BLF with consider-
ing the FOV limit is depicted in Fig. 5.
lP
The IGC Law Based on BLF with Considering the FOV Limit

Dynamic Surface Controller


 s1  x1
 x   k s  l sgn s  z
 2d 1 1 1  1  11
 x2 c   2 x 2 c  x2 d

 s2  x 2  x 2 c
 R  , z ,V
 s 
rna

 x  1   a x  z  l sgn  s   x  k s2  2 1 2 
 3d a23  22 2 21 2 2 2c 2
R kc  s1 
 , , R, R   
 x   x  x
 3c 3 3c 3d

 s3  x3  x3c
 x   k s  a x  z  l sgn  s   x  a s
 4d 3 3 33 3 31 3 3 3c 32 2
Longitudinal Missile Model
 x4 c   4 x 4 c  x4 d
 s x x  X  V cos
Strapdown Seeker

 4  
 Y  V sin 
4 4c

 1
 uc  b   a43 x3  a44 x4  z41  l4 sgn  s4   x 4 c  k 4 s4  s3  
Target Motion

T cos   D  mg sin 
 4  V 
 m
  T sin   L  mg cos
z11 , z21 , z31 , z41 uc  
 mV
Extended State Observer  T sin   L  mg cos
Jou

    z 
 mV
 ei  zi 0  xi i  1,2,3,4
 z  z   e  x  Mz
 10 11 10 1 2  z  I

 z20  z21   20 e2  a22 x2  a23 x3
z

 
 z30  z31   30 e3  a33 x3  x4
 , , R, R  z  z   e  a x  a x  b 
 40 41 40 4 43 3 44 4 4 z
  ei i sgn  ei  ei   i
 
 fal    ei  , z ,V
   1i ei   i
  i
 zi1    i1fal  ei , i , i  i  1,2,3,4

Figure 5: Configuration of the IGC law based on BLF with considering the FOV limit

28
Journal Pre-proof

of
285

To make the subsequent performance comparison as fair as possible, the


controller parameters of the comparative IGC law directly take the values rec-

pro
ommended by its proposer in [34]. The controller parameters of the proposed
IGC law are determined according to the following tuning procedure:
290 Step 1. Choose a positive constant K according to the expected convergence
rate of the closed-loop Lyapunov function V (Eq. (70)).
Step 2. According to Eq. (69), determine the lower bound of the design
parameters k1 , k2 , and the time constant of the first-order filter τ .

295
re-
Step 3. Further determine the upper and lower bounds of the controller
parameters. Specifically, the maximum allowable acceleration of the nonlinear
tracking differentiator (41) r > 0; the observer parameters of the homogeneous
observer (42) β1 > 0, β3 > 0, β2 > β1 /β3 , 0 < α1 < 1, 0 < ε < 1; the controller
parameter of the dynamic surface controller (51) l1 > 0; the parameters of the
lP
Nussbaum function (59) c > 0, γ > 0; and parameters to satisfy the FOV
300 constraint (80) m1 > 0, m2 > 0.
Step 4. Regarding the minimization of miss distance as an optimization
objective, determining the values of the above controller parameters is modeled
rna

as a optimal problem subject to state constraints, which is then addressed using


genetic algorithm.
305 Finally, the controller parameters of the proposed and comparative IGC laws
are listed in Table 2.

4.3. Case 1: attacking a stationary target


Jou

In this subsection, the performances of the proposed and comparative IGC


laws are compared against a stationary target. The strapdown homing missile
310 (83) is assumed to be launched from a height of 500m with 200m/s initial
velocity, zero attitude angle and angular rate, to attack a stationary ground
target 2000m away. Considering the FOV limit and the dead zone characteristics
of the strapdown seeker, the maximum BLOS angle kc is set as 20◦ and the
terminating distance of the IGC law Rf is set as 10m. When the relative missile-

29
Journal Pre-proof

of
Table 2: The controller parameters of the proposed and comparative IGC laws

The proposed IGC law

pro
β1 = 16.75 β2 = 10.49 β3 = 3.957 ε = 0.147 α1 = 0.307
k1 = 14.459 l1 = 80.517 k2 = 16.159 m1 = 3.971 m2 = 8.681
τ2 = 0.42 c = 0.711 γ = 0.829 r = 4.929

The comparative IGC law

β10 = 5 β20 = 5 β30 = 5 β40 = 5


β11 = 10
σ1 = 0.1
α1 = 0.6
re-β21 = 10
σ2 = 0.1
α2 = 0.6
β31 = 10
σ3 = 0.1
α3 = 0.6
β41 = 10
σ4 = 0.1
α4 = 0.6
k1 = 10 k2 = 15 k3 = 20 k4 = 25
l1 = 1 l2 = 1 l3 = 1 l4 = 1
lP
τ1 = 0.001 τ2 = 0.001 τ3 = 0.001

315 target distance R is less than 0.5m, the simulation stops. The simulation results
are shown in Fig. 6.
rna
Jou

(a) Velocity (b) The trajectories of missile

Figure 6: Simulation results of attacking a stationary target

30
Journal Pre-proof

of
pro
re-
(c) Sliding surfaces (d) The relative missile-target distance
lP
rna

(e) Angular rates under the propropsed IGC (f) Angles under the comparative IGC law
law
Jou

(g) Pitch rate (h) Inertial LOS angle and LOS angular rate

Figure 6: Simulation results of attacking a stationary target

31
Journal Pre-proof

of
pro
(i) BLOS anlge (j) Actuator deflection angle
re-
Figure 6: Simulation results of attacking a stationary target

Different from previous studies on IGC design, both the gravity and varying
missile velocity model are taken into account when designing the proposed IGC
lP
law. Figure 6(a) depicts the varying missile velocity due to the consideration of
320 thrust, and resultant missile trajectories are given in Fig. 6(b). From Fig. 6(c),
it is observed that all sliding surfaces of the proposed and comparative IGC laws
can converge to zero but note that, the final miss distance does not decrease to
0.5m in Fig. 6(d) when the latter works, even though its control objective has
rna

been achieved as shown in Fig. 6(f). In fact, although the BLOS angle under the
325 comparative IGC law has been forced to approach zero, the look angle, denoted
by the sum of the angle of attack and the BLOS angle, is almost overlapped with
the angle of attack and remains positive, which causes the missile velocity not
to point to the intended target, but to somewhere behind of the target when the
LOS angle is positive or to somewhere in front of the target when the LOS angle
Jou

330 is negative. As a result, the LOS angle would gradually increase as the relative
missile-target distance decreases, as depicted in Fig. 6(h). However, constrained
by actual dynamic characteristics of the missile body, the orientation of missile
velocity changes significantly slower than the LOS angle. Thus, a larger miss
distance for the comparative IGC law is accordingly inevitable. In reality, the
335 control objective of the comparative IGC law is only valid for the constant
missile velocity model with no gravity, just as the simulation results for that as

32
Journal Pre-proof

of
shown in Fig. 7. By contrast, it can be observed from Fig. 6(e) and Fig. 6(g) that
after initial adjustment at first 4s, the angular pitch rate is stabilized around
zero and then the BLOS angular rate under the proposed IGC law successfully

pro
340 tracks the negative pitch rate, thereby nullifying the LOS angular rate and
allowing a precise attack. As for the abrupt-jumping of pitch rate at the end of
the guidance phase, it should be pointed out that this is due to the over-close
missile-target distance. Figure 6(i) indicates that the BLOS angle under the
proposed and comparative IGC laws can both be constrained within the FOV
limit, and corresponding actuator deflection angles are shown in Fig. 6(j).
re-
lP
rna

(a) Angles under the comparetive IGC law (b) The relative missile-target distance

Figure 7: Simulation results of attacking a stationary target under the comparative IGC law
using constant velocity model without consideration of the gravity

345

4.4. Case 2: attacking moving targets

This subsection shows the performances of the proposed IGC law against a
Jou

non-maneuvering receding target and a maneuvering approaching target. The


velocity of moving targets are both 30m/s, and the acceleration of the maneu-
350 vering one is 2m/s2 . Figure 8 shows the simulation results of this case.

33
Journal Pre-proof

of
pro
(a) Velocities of the missile and target (b) The trajectories of missile
re-
lP

(c) Angular rates against the receding target (d) Angular rates against the approaching tar-
rna

get

Figure 8: Simulation results against receding and approaching targets


Jou

(e) The relative missile-target distance (f) BLOS anlge

Figure 8: Simulation results against receding and approaching targets

34
Journal Pre-proof

of
Because of the movement of the target, the flying time of the missile is
extended to 14.243s when attacking the receding target or reduced to 8.772s
when attacking the approaching target, leading to a significant change in the

pro
final attacking velocity of the missile as shown in Fig. 8(a). The corresponding
355 missile trajectories are given in Fig. 8(b). Under the proposed IGC law, as
can be seen from Fig. 8(c)–(d), the BLOS angular rate is always able to track
the negative pitch rate, which guarantees that the inertial LOS angular rate
is stabilized around zero, thereby attacking the non-maneuvering receding and
maneuvering approaching targets accurately while satisfying the FOV limit, as
360 shown in Fig. 8(e)–(f).
re-
4.5. Case 3: Monte-Carlo analysis

To further testify the robustness of the proposed IGC law against parametric
uncertainties, a Monte-Carlo analysis on attacking stationary targets is imple-
lP
mented. The biased parameters, listed in Table 3, are all modeled as random
365 constants. For each Monte-Carlo trial, the uncertainties of these biased pa-
rameters are randomly sampled within given bias values. The results of 500
times Monte-Carlo simulations are given in Fig. 9, where each gray curve cor-
rna

responds to a Monte-Carlo trail and the red thicker line represents the result
using nominal values.
Jou

(a) The trajectories of missile (b) The relative missile-target distance

Figure 9: Monte Carlo simulation results

35
Journal Pre-proof

of
pro
(c) Velocity
re- (d) BLOS anlge
lP
rna

(e) Inertial LOS angular rate (f) Pitch rate


Jou

(g) Actuator deflection angle (h) Angle of attack

Figure 9: Monte Carlo simulation results

36
Journal Pre-proof

of
Table 3: The biased parameters

pro
Biased parameter Bias value

Atmospheric
Air density ρ ±5%
parameter

Mass m ±5%
Structural Moment of inertial Iy ±5%
parameter
re- Reference area S
Reference length L
±5%
±5%
2
Drag coefficient Cx0 , Cxα ±20%
Aerodynamic
Lift coefficient Cyα , Cyδz ±20%
coefficient
lP
Pitch moment coefficient Mzα , Mzω̄z , and Mzδz ±20%
rna

Median: 0.42157

Group:
Jou

Maximum: 0.49998
Minimum: 0.33495
Num Points: 500
Num Finite Outliers: 0
Num NaN's or Inf's: 0

(i) The distributions of miss distance

Figure 9: Monte Carlo simulation results

37
Journal Pre-proof

of
370 As can be seen from Fig. 9(e)–(f), even in the presence of severe parametric
uncertainties, the inertial LOS angular rate and pitch rate under the proposed
IGC law can both converge to zero, which ensures that the missile body is in

pro
a stable state and capable of attacking the target accurately. Owing to the
employment of the OIST technique, the BLOS angles in all Monte-Carlo trials
375 are confined within ±20◦ in Fig. 9(d), thus meeting the FOV limit. Figure 9(i)
gives the distribution of miss distance in a box plot, where the central mark
indicates the median of the miss distance data, and the bottom and top edges of
the box indicate the 25th and 75th percentiles, respectively. It can be observed

380
re-
that all miss distance data are less than 0.5m, which convincingly demonstrated
the strong robustness of the proposed IGC law against parametric uncertainties.

5. Conclusion
lP
Considering the adverse effect of body attitude motion on maintaining the
target locking, an integrated guidance and control scheme is proposed in this
paper for strapdown homing missiles. Being capable of fully exploiting the syn-
385 ergistic relationships between the guidance and control loops, this novel IGC law
rna

provides a more practical solution to the problem of FOV limit without having
to unreasonably assume that the angle of attack is small enough throughout
the whole homing phase. Owing to the detailed novel IGC model derived in
this paper, the proposed IGC law just relies on the BLOS angle, thus avoid-
390 ing designing sophisticated filters/estimators to distill or estimate the inertial
guidance information required to compute the guidance command in previous
Jou

studies. The results of performance comparison and Monte-Carlo simulation


fully demonstrate the excellent performance and strong robustness of the pro-
posed IGC law.

395 References

1. Vergez PL, McClendon JR. Optimal control and estimation for strap-
down seeker guidance of tactical missiles. J Guidance, Control, Dyn

38
Journal Pre-proof

of
1982;5(3):225–6. doi:10.2514/3.19767.

2. Lee CH, Kim TH, Tahk MJ. Biased PNG for target observability enhance-
ment against nonmaneuvering targets. IEEE Trans Aerosp Electron Syst

pro
400

2015;51(1):2–17. doi:10.1109/TAES.2014.120103.

3. Lee CH, Hyun C, Lee JG, Choi JY, Sung S. A hybrid guidance law for a
strapdown seeker to maintain lock-on conditions against high speed targets.
J Electr Eng Technol 2013;8(1):190–6. doi:10.5370/JEET.2013.8.1.190.

405 re-
4. Xin M, Balakrishnan SN, Ohlmeyer EJ. Guidance law design for missiles
with reduced seeker field-of-view. In: Proc. AIAA Guidance, Navig. Control
Conf. Keystone, Colorado, USA: AIAA; 2006:doi:10.2514/6.2006-6085.

5. Park BG, Kim TH, Tahk MJ. Optimal impact angle control guidance law
considering the seeker’s field-of-view limits. Proc Inst Mech Eng, Part G,
lP
410 J Aerosp Eng 2013;227(8):1347–64. doi:10.1177/0954410012452367.

6. Park BG, Kim TH, Tahk MJ. Range-to-go weighted optimal guidance with
impact angle constraint and seeker’s look angle limits. IEEE Trans Aerosp
Electron Syst 2016;52(3):1241–56. doi:10.1109/TAES.2016.150415.
rna

7. Lee CH, Kim TH, Tahk MJ, Whang IH. Polynomial guidance laws con-
415 sidering terminal impact angle and acceleration constraints. IEEE Trans
Aerosp Electron Syst 2013;49(1):74–92. doi:10.1109/TAES.2013.6404092.

8. Li R, Wen Q, Tan W, Zhang Y. Adaptive weighting impact angle opti-


Jou

mal guidance law considering seeker’s FOV angle constraints. J Syst Eng
Electron 2018;29(1):142–51. doi:10.21629/JSEE.2018.01.14.

420 9. Park BG, Jeon BJ, Kim TH, Tahk MJ, Kim YH. Composite guidance
law for impact angle control of tactical missiles with passive seekers. In:
Proc. Asia-Pac. Int. Symp. Aerospace Technol. Seoul, Korea: The Korean
Society for Aeronautical & Space Sciences; 2012:13–5.

39
Journal Pre-proof

of
10. Park BG, Kwon HH, Kim YH, Kim TH. Composite guidance scheme for
425 impact angle control against a nonmaneuvering moving target. J Guidance,
Control, Dyn 2016;39(5):1129–36. doi:10.2514/1.G001547.

pro
11. Tekin R, Erer KS. Switched-gain guidance for impact angle control under
physical constraints. J Guidance, Control, Dyn 2015;38(2):205–16. doi:10.
2514/1.G000766.

430 12. Ratnoo A. Analysis of two-stage proportional navigation with heading


constraints. J Guidance, Control, Dyn 2016;39(1):156–64. doi:10.2514/1.
G001262.
re-
13. Kim TH, Park BG, Tahk MJ. Bias-shaping method for biased propor-
tional navigation with terminal-angle constraint. J Guidance, Control, Dyn
435 2013;36(6):1810–5. doi:10.2514/1.59252.
lP
14. Park BG, Kim TH, Tahk MJ. Biased PNG with terminal-angle constraint
for intercepting nonmaneuvering targets under physical constraints. IEEE
Trans Aerosp Electron Syst 2017;53(3):1562–72. doi:10.1109/TAES.2017.
2667518.
rna

440 15. He S, Lin D. A robust impact angle constraint guidance law with seeker’s
field-of-view limit. Trans Inst Meas Control 2015;37(3):317–28. doi:10.
1177/0142331214538278.

16. Kim HG, Kim HJ. Impact time control guidance considering seeker’s field-
of-view limits. In: Proc. 55th IEEE Conf. Decis. Control. Las Vegas, NV,
Jou

445 USA: IEEE; 2016:4160–5. doi:10.1109/CDC.2016.7798900.

17. Qian S, Yang Q, Geng L, Zheng Z. SDRE based impact angle control
guidance law considering seeker’s field-of-view limit. In: Proc. 2016 IEEE
Chinese Guidance, Navig. Control Conf. Nanjing, China: IEEE; 2016:1939–
44. doi:10.1109/CGNCC.2016.7829086.

40
Journal Pre-proof

of
450 18. Wang X, Zhang Y, Wu H. Sliding mode control based impact angle con-
trol guidance considering the seeker’s field-of-view constraint. ISA Trans
2016;61:49–59. doi:10.1016/j.isatra.2015.12.018.

pro
19. Liu B, Hou M, Feng D. Nonlinear mapping based impact angle control guid-
ance with seeker’s field-of-view constraint. Aerosp Sci Technol 2019;86:724–
455 36. doi:10.1016/j.ast.2019.02.009.

20. Erer KS, Tekin R, Özgören MK. Look angle constrained impact angle
control based on proportional navigation. In: AIAA Guidance Control
re-
Conf. Kissimmee, Florida, USA: AIAA; 2015:doi:10.2514/6.2015-0091.

21. Zhao B, Xu S, Guo J, Jiang R, Zhou J. Integrated strapdown missile


460 guidance and control based on neural network disturbance observer. Aerosp
Sci Technol 2019;84:170–81. doi:10.1016/j.ast.2018.10.025.
lP
22. Paul Z. Tactical and Strategic Missile Guidance (6th Edition). Reston,
Virginia: AIAA, Inc.; 2012.

23. Jang SA, Ryoo CK, Choi K, Tahk MJ. Guidance algorithms for tactical
missiles with strapdown seeker. In: 2008 SICE Annual Conf. Tokyo, Japan:
rna

465

IEEE; 2008:2616–9. doi:10.1109/SICE.2008.4655108.

24. Maley JM. Line of sight rate estimation for guided projectiles with strap-
down seekers. In: Proc. AIAA Guidance, Navig. Control Conf. Kissimmee,
Florida,U.S.A.: AIAA; 2015:doi:10.2514/6.2015-0344.
Jou

470 25. Sun T, Chu H, Zhang B, Jia H, Guo L, Zhang Y, Zhang M. Line-of-
sight rate estimation based on UKF for strapdown seeker. Math Probl Eng
2015;2015. doi:10.1155/2015/185149.

26. Sun B, Fan J. A linear/nonlinear hybrid differentiator and application


for DSGC system. In: Proc. 32nd Chinese Control Conf. Nanjing, China:
475 IEEE; 2013:5172–6.

41
Journal Pre-proof

of
27. Wang P, Zhang K. Research on line-of-sight rate extraction of strap-
down seeker. In: Proc. 33rd Chinese Control Conf. Xi’an, China: IEEE;
2014:859–63. doi:10.1109/ChiCC.2014.6896740.

pro
28. Sun G, Zhu M, Yin S, Jia H. Optical axis stabilization of semi-strapdown
480 seeker based on disturbance observer. In: Proc. IEEE International Conf.
Information and Automation. Shenzhen, China: IEEE; 2011:550–3. doi:10.
1109/ICINFA.2011.5949054.

29. Lee S, Kim Y. Design of nonlinear observer for strap-down missile guidance

485
re-
law via sliding mode differentiator and extended state observer. In: Proc.
2016 International Conf. on Advanced Mechatronic Systems. Melbourne,
Australia: IEEE; 2016:143–7. doi:10.1109/ICAMechS.2016.7813436.

30. Wang T, Tang S, Guo J, Zhang H. Two-phase optimal guidance law con-
lP
sidering impact angle constraint with bearings-only measurements. Int J
Aerospace Eng 2017;2017. doi:10.1155/2017/1380531; article ID 1380531.

490 31. Kim TH, Lee CH, Tahk MJ. Time-to-go polynomial guidance with tra-
jectory modulation for observability enhancement. IEEE Trans Aerosp
rna

Electron Syst 2013;49(1):55–73. doi:10.1109/TAES.2013.6404091.

32. Sang DK, Tahk MJ. Guidance law switching logic considering the
seeker’s field-of-view limits. Proc Inst Mech Eng, Part G, J Aerosp Eng
495 2009;223(8):1049–58. doi:10.1243/09544100JAERO614.

33. Chambon E. Frequency- and time-domain constrained control of linear


Jou

systems application to a flexible launch vehicle. Ph.D. thesis; Institut


Supérieur de l’Aéronautique et de l’Espace (ISAE), Université de Toulouse;
2016.

500 34. Li X, Zhao B, Zhou J, Feng Z. Integrated guidance and control for missiles
with strap-down seeker. In: Proc. 36th Chinese Control Conf. Dalian,
China: IEEE; 2017:6208–12. doi:10.23919/ChiCC.2017.8028345.

42
Journal Pre-proof

of
35. Zhao B, Zhu C, Xu S, Jiang R, Zhang L, Zhou J. IGC design for missile with
strapdown seeker against maneuvering target. J Astron 2019;40(3):310–9.
505 doi:10.3873/j.issn.1000-1328.2019.03.008; in Chinese.

pro
36. Roger MC. Finite-time sliding mode controller for perturbed second-order
systems. ISA Transactions 2019;95:82–92. doi:10.1016/j.isatra.2019.
05.026.

37. Wen C, Zhou J, Liu Z, Su H. Robust adaptive control of uncertain nonlinear


510 systems in the presence of input saturation and external disturbance. IEEE
re-
Trans Autom Control 2011;56(7):1672–8. doi:10.1109/TAC.2011.2122730.

38. Wang W, Xiong S, Wang S, Song S, Lai C. Three dimensional impact


angle constrained integrated guidance and control for missiles with input
saturation and actuator failure. Aerosp Sci Technol 2016;53:169–87. doi:10.
lP
515 1016/j.ast.2016.03.015.

39. He S, Wang J, Wang W. A novel sliding mode guidance law without line-
of-sight angular rate information accounting for autopilot lag. Int J Syst
Sci 2017;48(16):3363–73. doi:10.1080/00207721.2017.1382607.
rna

40. Han J. Active Disturbance Rejection Control Technique—the technique for


520 estimating and compensating the uncertainties; chap. 2.3. Beijing: National
Defense Industry Press; 2013: 59. In Chinese.
Jou

43
Journal Pre-proof

The highlights of this paper are summarized as follows:


1. A novel IGC law with consideration of the FOV limit;
2. Based on the body LOS angel, a detailed IGC model is derived;

of
3. The FOV limit is handled by means of the OIST technique.

pro
re-
lP
rna
Jou

1
Journal Pre-proof

Conflict of interest statement

We declare that we have no financial and personal relationships with other


people or organizations that can inappropriately influence our work, there is no

of
professional or other personal interest of any nature or kind in any product,
service and/or company that could be construed as influencing the position

pro
presented in, or the review of, the manuscript entitled.

re-
lP
rna
Jou

You might also like