See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/27253830

Mathematical Models for a Missile Autopilot Design

Article · January 2002

Source: OAI

CITATIONS READS

15 5,346

2 authors, including:

Farhan A Faruqi
Australian Army Malaria Institute
40 PUBLICATIONS 352 CITATIONS

SEE PROFILE

All content following this page was uploaded by Farhan A Faruqi on 25 October 2017.

The user has requested enhancement of the downloaded file.

 Mathematical Models for a Missile Autopilot Design

Farhan A. Faruqi and Thanh Lan Vu

Weapons Systems Division

Systems Sciences Laboratory

DSTO-TN-0449

ABSTRACT
This report considers the derivation of the mathematical model for a missile autopilot in state
space form. The basic equations defining the airframe dynamics are non-linear, however,
since the non-linearities are “structured” (in the sense that the states are of quadratic form) a
novel approach of expressing this non-linear dynamics in state space form is given. This
should provide a useful way to implement the equations in a computer simulation program
and possibly for future application of non-linear analysis and synthesis techniques,
particularly for autopilot design of missiles executing high g-manoeuvres.

This report also considers a locally linearised state space model that lends itself to better
known linear techniques of the modern control theory. A coupled multi-input multi-output
(MIMO) model is derived suitable for both the application of the modern control techniques
as well as the classical time-domain and frequency domain techniques. This is validated by
comparing the model with the other published results, and through both open and closed-
loop systems simulations. The models developed are useful for further research on precision
optimum guidance and control. It is hoped that the model will provide more accurate
presentations of missile autopilot dynamics and will be used for adaptive and integrated
guidance & control of agile missiles.

RELEASE LIMITATION

Approved for public release

Published by

DSTO Systems Sciences Laboratory

PO Box 1500
Edinburgh South Australia 5111 Australia

Telephone: (08) 8259 5555

Fax: (08) 8259 6567

© Commonwealth of Australia 2002

AR-012-430
August 2002

APPROVED FOR PUBLIC RELEASE

 Mathematical Models for a Missile Autopilot
Design

Executive Summary
Requirements for next generation guided weapons, particularly with respect to their
capability to engage high speed, highly agile targets and achieve precision end-game
trajectory, has prompted a revision of the way in which the guidance and autopilot
design is undertaken. This report considers the derivation of the mathematical models
for a missile autopilot in state space form. The basic equations defining the airframe
dynamics are non-linear, however, since the non-linearities are “structured” (in the
sense that the states are of quadratic form) a novel approach of expressing this non-
linear dynamics in state space form is given. This should provide a useful way to
implement the equations in a computer simulation program and possibly for future
application of non-linear analysis and synthesis techniques.

This report also considers a locally linearised state space model that lends itself to
better known linear techniques of the modern control theory. A coupled multi-input
multi-output (MIMO) model is derived suitable for both the application of the modern
control techniques as well as the classical time-domain and frequency domain
techniques. This is validated by comparing the model with the other published results,
and through both open and closed-loop systems simulations. The models developed
are useful for further research on precision optimum guidance and control. It is hoped
that the model will provide more accurate presentations of missile auto-pilot dynamics
and will be used for adaptive and integrated guidance & control of agile missiles.
 Authors

Dr. Farhan A. Faruqi

Weapons Systems Division

Farhan A. Faruqi received B.Sc.(Hons) in Mechanical Engineering

from the University of Surrey (UK), 1968; M.Sc. in Automatic
Control from the University of Manchester Institute of Science and
Technology (UK), 1970 and Ph.D from the Imperial College,
London University (UK), 1973. He has over 20 years experience in
the Aerospace and Defence Industry in UK, Europe and the USA.
Prior to joining DSTO in January 1999 he was an Associate
Professor at QUT (Australia) 1993-98. His research interests
include: Missile Navigation, Guidance and Control, Target
Tracking and Precision Pointing Systems, Strategic Defence
Systems, Signal Processing, and Optoelectronics.
____________________ ________________________________________________

Thanh Lan Vu
Weapons Systems Division

Thanh Lan Vu joined the Guidance & Control group in the

Weapons Systems Dvision in August 1999. She received a Master
degree in Computing & Control Engineering from the Institute of
Norwegian Technology, Trondheim, Norway, in 1991. She also
received a PhD in Mechanical Engineering from the University of
Western Australia in 1998. Her thesis presents on-line system
identification method and active control of non-linear vibration.
Her research interests include: Guidance and Control, Signal
Processing, Analog and Digital Filters, Active Control of
Vibration, and System Identification.
____________________ ________________________________________________
 Contents

1. INTRODUCTION ............................................................................................................... 1

2. RIGID BODY DYNAMICS ............................................................................................... 2

2.1 Notation and Convention........................................................................................ 2
2.2 Euler’s Equations of Motion ................................................................................... 4
2.3 Linearised model for a two-axis symmetrical airframe ..................................... 9
2.4 Incorporation of accelerometer and gyro measurement model ..................... 12
2.5 Linearised model of the airframe including fin servos ................................... 14
2.6 Lateral Auto-pilot design ...................................................................................... 18

3. VERIFICATION OF THE DEVELOPED MODEL...................................................... 19

4. CONCLUSIONS ................................................................................................................ 23

5. REFERENCES..................................................................................................................... 23

APPENDIX A: STATE EQUATION................................................................................. 25

A.1. Quadratic State Vector .................................................................. 25
A.2. Linearisation of the Quadratic Vector........................................ 26
A.3. Calculation of Inverse Matrix ...................................................... 26

APPENDIX B: FROM STATE-SPACE FORM INTO TRANSFER-FUNCTION..... 28

APPENDIX C: MATLAB CODES..................................................................................... 29

C.1. Converting the state-space model into the
transfer- functions .................................................................................... 29
C.2. Open-loop and closed-loop simulation .................................... 31

APPENDIX D: VALUES OF THE NON-ZERO DERIVATIVES AND

PARAMETERS USED IN THE SIMULATION ................................................................ 36
 DSTO-TN-0449

1. Introduction
Requirements for next generation guided weapons, particularly with respect to their
capability to engage high speed, highly agile targets and achieve precision end-game
trajectory, has prompted a revision of the way in which the guidance and autopilot
design is undertaken. Integrating the guidance and control function is a synthesis
approach that is being pursued as it allows the optimisation of the overall system
performance. This approach requires a more complete representation of the airframe
dynamics and the guidance system. The use of state space model allows the application
of modern control techniques such as the optimal control and parameter estimation
techniques to be utilised. In this report we derive the autopilot model that will serve as
a basis for an adaptive autopilot design and allow further extension of this to
integrated guidance and control system design.

Over the years a number of authors [1-3, 6] have considered modelling, analysis and
design of autopilots for atmospheric flight vehicles including guided missiles. In the
majority of the published work on autopilot analysis and design, locally linearised
versions of the model with decoupled airframe dynamics has been considered. This
latter simplification arises out of the assumption that the airframe and its mass
distribution are symmetrical about the body axes, and that the yaw, pitch and roll
motion about the equilibrium state remain “small”. As a result, most of the autopilot
analysis and design techniques, considered in open literature, use classical control
approach, such as: single input/single output transfer-functions characterisation of the
system dynamics and Bode, Nyquist, root-locus and transient response analysis and
synthesis techniques [5,7]. These techniques are valid for a limited set of flight regimes
and their extension to cover a wider set of flight regimes and airframe configurations
requires autopilot gain and compensation network switching.

With the advent of fast processors it is now possible to take a more integrated
approach to autopilot design. Modern optimal control techniques allow the designer to
consider autopilots with high-order dynamics (large number of states) with multiple
inputs/outputs and to synthesise controllers such that the error between the
demanded and the achieved output is minimised. Moreover, with real-time
mechanisation any changes in the airframe aerodynamics can be identified (parameter
estimation) and compensated for by adaptively varying the optimum control gain
matrix. This approach should lead to missile systems that are able to execute high g-
manoeuvres (required by modern guided weapons), adaptively adjust control
parameters (to cater for widely varying flight profiles) as well as account for non-
symmetric airframe and mass distributions.

Typically, for a missile autopilot, the input is the demanded control surface deflection
and outputs are the achieved airframe (lateral) accelerations and body rates measured
about the body axes. Other input/output variables (such as: the flight path angle and
angle rate or the body angles) can be derived directly from lateral accelerations and
body rates.

1
DSTO-TN-0449

This report considers the derivation of the mathematical model for a missile autopilot
in state space form. The basic equations defining the airframe dynamics are non-linear,
however, since the non-linearities are “structured” (in the sense that the states are of
quadratic form) a novel approach of expressing this non-linear dynamics in state space
form is given. This should provide a useful way to implement the equations in a
computer simulation program and possibly for future application of non-linear
analysis and synthesis techniques. Detailed consideration of the quadratic/bilinear
type of dynamic systems is given in [4].

This report also considers a locally linearised state space model that lends itself to
better known linear techniques of the modern control theory. A coupled multi-input
multi-output (MIMO) model is derived suitable for both the application of the modern
control techniques as well as the classical time-domain and frequency domain
techniques. This is validated by comparing the model with others previously published
and through simulation of a decoupled single-input single-output (SISO) system.

2. Rigid Body Dynamics

2.1 Notation and Convention

Conventions and notations for vehicle body axes systems as well as the forces,
moments and other quantities are shown in Figure 2.1 and defined in Table 2.1.

Ixx, Lp

Iyy, Mq

Izz, Nr σ

Figure 2.1 Motion variable notations

2
 DSTO-TN-0449

The variables shown in Figure 2.1 are defined as:

m - mass of a vehicle.
α - incidence in the pitch plane.
β - incidence in the yaw plane.
λ - incidence plane angle.
σ - total incidence, such that: tan α = tan σ cos λ, and tan β = tan σ sin λ.
T – thrust.

Table 2.1: Motion variables

Roll Pitch Yaw
Vehicle Body Axes System axis axis axis
X Y Z
Angular rates p q r
Component of vehicle velocity along each axis u v w
Component of aerodynamic forces acting on vehicle along X Y Z
each axis
Moments acting on vehicle about each axis L M N
Moments of inertia about each axis Ixx Iyy Izz
Products of each inertia Iyz Izx Ixy
Longitudinal and lateral accelerations ax ay az
Euler angles φ θ ψ
Gravity along each axis gx gy gz
Vehicle thrust along the body axis T

ξ - aileron deflection.
η - elevator deflection.
ς - rudder deflection.

Figure 2.2 defines the control surface convention. Here the control surfaces are
numbered as shown and the deflections ( δ 1 , δ 2 , δ 3 , δ 4 ) are taken to be positive if
clockwise, looking outwards along the individual hinge axis. Thus:

1
Aileron deflection: ξ = ( δ 1 + δ 2 + δ 3 + δ 4 ) , if all four control surfaces are active; or
4
1 1
ξ = ( δ 1 + δ 3 ) , or ξ = ( δ 2 + δ 4 ) if only two surfaces are active. Positive control
2 2
defection (ξ) causes negative roll.

3
DSTO-TN-0449

1
Elevator deflection: η = ( δ 1 − δ 3 ) . Positive control deflection (η) causes negative
2
pitch.

1
Rudder deflection: ζ = ( δ 2 − δ 4 ) . Positive control deflection (ζ) causes negative yaw.
2

+δ4

+δ3 +δ1

+δ2

Figure 2.2 Control surfaces seen from the rear of a missile

2.2 Euler’s Equations of Motion

The six equations of motion for a body with six degrees of freedom may be written as
[1-3]:

m( u& + wq − vr ) = X + T + g x m (2. 1)
m( v& + ur − wp ) = Y + g y m (2. 2)
m( w& − uq + vp ) = Z + g x m (2. 3)

I xx p& − ( I yy − I zz )qr + I yz ( r 2 − q 2 ) − I zx ( pq + r& ) + I xy ( rp − q& ) = L (2.4)

2 2
I yy q& − ( I zz − I xx )rp + I zx ( p − r ) − I xy ( qr + p& ) + I yz ( pq − r& ) = M (2.5)
2 2
I zz r& − ( I xx − I yy ) pq + I xy ( q − p ) − I yz ( rp + q& ) + I zx ( qr − p& ) = N . (2.6)

d
Here (⋅) = - is the derivative operator.
dt

4
 DSTO-TN-0449

Equations (2.1 to 2.3) represent the force equations of a generalised rigid body and
describe the translational motion of its centre of gravity (c.g) since the origin of the
vehicle body axes is assumed to be co-located with the body c.g.

Equations (2.3 to 2.6) represent the moment equations of a generalised rigid body and
describe the rotational motion about the body axes through its c.g.

Separating the derivative terms and after some algebraic manipulation, Equations (2.1
to 2.3) may be written in a vector form as:

~ ~
 u  0 0 0 1 0 − 1  uq   X + T   g x 
d   
v = 0 − 1 0 0 1 0   ur  +  Y~  +  g 
dt       ~   y
 w 1 0 − 1 0 0 0   vp   Z   g z 
 vr  (2.7)
 
 wp 
 wq 
 

~ X ~ Y ~ Z ~ T
where: X = ; Y = ; Z= ; T = .
m m m m

Note that the states (u,v,w,p,q,r) appear as “quadratic terms (form)”.

Equations (2.4 to 2.6) can be written as:

 p  p2   L 
[A] d  q  = [B ] pq  + M 
dt  pr   N 
 r 
 2 (2.8)
q 
 qr 
 2
 r 

where: matrices [A] and [B ] are given by:

5
DSTO-TN-0449

 I xx − I xy − I zx 
[A] = − I xy I yy − I yz 
 − I zx − I yz I zz 
 

 0 I zx − I xy I yz (I yy− I zz ) − I yz 
[B ] = − I zx − I yz (I zz − I xx ) 0 I xy I zx 

 I xy
 (I xx − I yy ) I yz − I xy − I zx 0 

Here again, the states (p,q,r) appear in “quadratic form”.

Equation (2.8) may also be written as:

 p  p2  L
d  
dt  q  = [
[ A]−1
[B ] ]
 
 pq  + [A] [
−1
]
M 
 
(2.9)
 r   pr   N 
 2
q 
 qr 
 2
 r 

where the inverse [A]

−1
is given by (see Appendix A.3):

(
 I yy I zz − I yz 2 )(I zz I xy + I yz I zx ) (I yz I xy + I yy I zx )
[A] −1 1 
(
=  (I zz I xy + I yz I zx ) I xx I zz − I zx
∆
2
) (I xx I yz + I zx I xy )
(
(I yz I xy + I yy I zx ) (I xx I yz + I xy I zx ) I xx I yy − I xy 
2 
)
(2.9a)

( 2 2 2
and ∆ = I xx I yy I zz − I xx I yz − I yy I zx − I zz I xy − 2 I yz I zx I xy . )

The selection of the particular order of the terms in the “quadratic-state vectors”
[uq ur vp vr wp wq ]T of
Equation (2.7) and [p 2
pq pr q 2 qr r 2 ] of Equation (2.8) is
T

discussed in Appendix A.1.

Combining Equations (2.7) and (2.9), we obtain the full 6th order rigid body dynamics
state equations as:

6
 DSTO-TN-0449

 x [11]  [C ] [0 ]  x [12 ]  [I ] [0 ]  u [11]   g 

d           (2.10)
= +   +  
dt  [1]   
 x 2   [0 ] [ −1   ]
[A] [B ]   x 2   [0 ]
[2 ]  
[[A] ]−1  u [1]   0 
 2   

0 0 0 1 0 − 1

where [C] = 0 − 1 0 0 1 0

 
1 0 - 1 0 0 0 

x 1 = [u v w] ,
[1 ] T

x 2 = [p q r ] ,
[1 ] T

x 1 = [uq ur vp vr wp wq ] ,
[2 ] T

[2 ]
x2 = p2 [ pq pr q 2 qr r 2 , ]
T

[1] ~ ~ ~
u1 = X + T Y [ ]
~T
Z ,

u 2 = [L M N] ,
[1] T

g = gx [ gy gz ].
T

Equation (2.10) may be written in a compact form as:

d [1]
x = [F ]x + [G ]u + g [1]
[2 ] [1]
dt
(2.11)

where
[C ] [0 ]

[F ] =  
 : is the 6x12 (quadratic) state coefficient matrix.
 [0 ]
 [
[A] [B] 
−1
]
[I ] [0] 
[G ] =  
 : is the 6x6 coefficient matrix.
 [0 ]
 [[A] ] −1



7
DSTO-TN-0449

[1 ]
x = x1 [ [1 ]
x2
[1 ]
] = [u
T
v w p q r ] : is the 6x1 linear-state vector.
T

x
[2 ]
[
= x1
[2 ]
x2
[2 ]
] T
= [uq ur vp vr wp wq p2 pq pr q2 qr r2] :
T
is
the 12x1 quadratic-state vector.

[1 ]
u = u1 [ [1 ]
u1
[1 ]
] = [X~ + T~
T ~
Y
~
Z L M N ]T
: is 6x1 a vector function of control
inputs, forces and moments.

and g
[1 ]
[
= g 0 = [g x ]
T
gy gz 0 0 0 ] : is the 6x1 gravity (or disturbance)
T

vector.

Note that for a two-axis symmetrical airframe, Iyz = Izx = Ixy = 0. Hence, in this case, the
equation (2.9) can be reduced to:

~ ~
 p  0 0 I xx   pq   L 
d    ~   ~
q = 0 I yy 0   pr  +  M  (2.12)
dt    ~ ~
 r   I zz 0 0   qr   N 

~ I yy − I zz ~ I − I xx ~ I xx − I yy
where I xx = , I yy = zz , I zz = ,
I xx I yy I zz

~ L ~ M ~ N
L= , M = , N= .
I xx I yy I zz

As a result, the state equation for the system now becomes:

 [0]   x 1[2 ] 
[I ] [0] u 1[1] 
 x 1[1]  [C ]  g
d      +   
=  +   (2.13)
dt  [1]        
 x 2   [0 ] [H ]  x [22 ]   [0] [I ] u [21]   0 
   

~
0 0 I xx 
 ~ 
where: [H ] =  0 x 2 = [ pq qr ] , and
[2 ] T
I yy 0 , pr
~ 0 
 I zz 0

8
 DSTO-TN-0449

[1] ~
u2 = L[ ~
M
~T
N . ]
Remarks:

Equations (2.11) and (2.13) are complete non-linear description of the full 6-DOF
autopilot model. In fact, these equations contain quadratic terms in states and will be
classed as the quadratic dynamic model. This type of model is required when autopilot
design is undertaken for a missile executing high g- or high angle of attack
manoeuvres, and (u, v, w, p, q, r) are not small.

A more detailed consideration of the algebraic structure of this type of dynamic

systems is given in [4].

2.3 Linearised model for a two-axis symmetrical airframe

~ ~ ~ ~ ~ ~
It is assumed that X , Y , Z , L , M and N are functions of u, v, w, p, q, r, ξ, η and ς.
Using first order linearisation about the nominal values u0, v0, w0, p0, q0, r0, ξ0, η0 and ς0,
and defining the aerodynamic derivatives as:

~ ~ ~ ~ ~ ~
~ ∂X ~ ∂X ~ ∂X ~ ∂X ~ ∂X ~ ∂X
Xu = , Xv = , Xw = , Xp = , Xq = , Xr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂X ~ ∂X ~ ∂X
Xξ = , Xη = , Xζ = ,
∂ξ ∂η ∂ζ

~ ~ ~ ~ ~ ~
~ ∂Y ~ ∂Y ~ ∂Y ~ ∂Y ~ ∂Y ~ ∂Y
Yu = , Yv = , Yw = , Yp = , Yq = , Yr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂Y ~ ∂Y ~ ∂Y
Yξ = , Yη = , Yζ = ,
∂ξ ∂η ∂ζ

~ ~ ~ ~ ~ ~
~ ∂Z ~ ∂Z ~ ∂Z ~ ∂Z ~ ∂Z ~ ∂Z
Zu = , Zv = , Zw = , Zp = , Zq = , Zr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂Z ~ ∂Z ~ ∂Z
Zξ = , Zη = , Zζ = ,
∂ξ ∂η ∂ζ

~ ~ ~ ~ ~ ~
~ ∂L ~ ∂L ~ ∂L ~ ∂L ~ ∂L ~ ∂L
Lu = , Lv = , Lw = , Lp = , Lq = , Lr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂L ~ ∂L ~ ∂L
Lξ = , Lη = , Lζ = ,
∂ξ ∂η ∂ζ

9
DSTO-TN-0449

~ ~ ~ ~ ~ ~
~ ∂M ~ ∂M ~ ∂M ~ ∂M ~ ∂M ~ ∂M
Mu = , Mv = , Mw = , Mp = , Mq = , Mr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂M ~ ∂M ~ ∂M
Mξ = , Mη = , Mζ = ,
∂ξ ∂η ∂ζ

~ ~ ~ ~ ~ ~
~ ∂N ~ ∂N ~ ∂N ~ ∂N ~ ∂N ~ ∂N
Nu = , Nv = , Nw = , Np = , Nq = , Nr = ,
∂u ∂v ∂w ∂p ∂q ∂r
~ ~ ~
~ ∂N ~ ∂N ~ ∂N
Nξ = , Nη = , Nζ = .
∂ξ ∂η ∂ζ

The six equations of motion of an airframe (using equation (2.13)) can thus be written
as:

∆ u& = r0 ∆v + v0 ∆r − q 0 ∆w − w0 ∆q
(
~ ~ ~ ~ ~ ~ ~ ~ ~
+ X u ∆u + X v ∆v + X w ∆w + X p ∆p + X q ∆q + X r ∆r + X ξ ∆ξ + X η ∆η + X ς ∆ς )
~
+ ∆T + ∆g x
(2.14a)
∆v& = p 0 ∆w + w0 ∆p − r0 ∆u − u 0 ∆r
~
( ~ ~ ~ ~ ~ ~ ~ ~
)
+ Yu ∆u + Yv ∆v + Yw ∆w + Y p ∆p + Yq ∆q + Yr ∆r + Yξ ∆ξ + Yη ∆η + Yς ∆ς + ∆g y
(2.14b)

∆w& = q 0 ∆u + u 0 ∆q − p0 ∆v − v0 ∆p
~
( ~ ~ ~ ~ ~ ~ ~ ~
)
+ Z u ∆u + Z v ∆v + Z w ∆w + Z p ∆p + Z q ∆q + Z r ∆r + Z ξ ∆ξ + Z η ∆η + Z ς ∆ς + ∆g z

(2.14c)
~
∆p& = I xx (q 0 ∆r + r0 ∆q )
(
~ ~ ~ ~ ~ ~ ~ ~ ~
+ Lu ∆u + Lv ∆v + Lw ∆w + L p ∆p + Lq ∆q + Lr ∆r + Lξ ∆ξ + Lη ∆η + Lς ∆ς )
(2.14d)
~
∆q& = I yy (r0 ∆p + p 0 ∆r )
( ~ ~ ~ ~ ~ ~ ~ ~ ~
+ M u ∆u + M v ∆v + M w ∆w + M p ∆p + M q ∆q + M r ∆r + M ξ ∆ξ + M η ∆η + M ς ∆ς )
(2.14e)
~
∆r& = I zz ( p 0 ∆q + q 0 ∆p )
(~ ~ ~ ~ ~ ~ ~ ~ ~
+ N u ∆u + N v ∆v + N w ∆w + N p ∆p + N q ∆q + N r ∆r + N ξ ∆ξ + N η ∆η + N ς ∆ς )
(2.14f)

10
 DSTO-TN-0449

Equations (2.14a to 2.14f) may be written in a matrix notation as:

~ ~ ~ ~ ~ ~
 ∆u&   Xu ( r0 + X v ) ( −q0 + X w ) Xp ( − w0 + X q ) ( v0 + X r )   ∆u 
 ∆v&  ( −r + Y~ ) ~
Yv
~
( p 0 + Yw )
~
( w0 + Y p )
~
Yq
~ 
( −u 0 + Yr )   ∆v 
   0 ~u ~ ~ ~ ~ ~
∆w&   ( q0 + Z u ) ( − p0 + Z v ) Zw ( −v 0 + Z p ) ( u0 + Z q ) Zr  ∆w
 = ~ ~ ~ ~ ~ ~ ~ ~   +
 ∆p&   Lu Lv Lw Lp ( I xx r0 + Lq ) ( I xx q0 + Lr )   ∆p 
 ∆q&   M ~ ~ ~ ~ ~ ~ ~ ~
Mv Mw ( I yy r0 + M p ) Mq ( I yy p0 + M r )  ∆q 
   ~
u
~ ~ ~ ~ ~ ~ ~  
 ∆r&   Nu Nv Nw ( I zz q0 + N p ) ( I zz p0 + N q ) Nr   ∆r 
~ ~ ~ ~
 Xξ Xη Xς  ∆T + ∆g x 
~ ~ ~   
 Yξ Yη Yς   ∆g y 
~   ∆ ξ
 Z~ξ ~
Zη Z ς     ∆g z 
~ ~ ~  ∆η +  
 Lξ Lη Lς     0 
M ~ ~ ~   ∆ς   
Mη Mς 0
~ ξ
~ ~    (2.15)
 N ξ N η N ς   0 

Note that Equation (2.15) is represented in a state-space form as:

d
∆x = [F1 ]∆x + [G1 ]∆u 1 + ∆w1 (2.16)
dt

~
 ∆u  ∆T + ∆g x 
 ∆v   
   ∆g y 
 ∆ξ 
∆w    ∆g z 
where ∆x =   , ∆u 1 = ∆η , ∆w1 =  ,
 
 ∆p   ∆ς   0 
 ∆q   0 
   
 ∆r   0 

11
DSTO-TN-0449

~ ~ ~ ~ ~ ~
 Xu (r0 + X v ) − q 0 + X w Xp − w0 + X q v0 + X r 
 ~ ~ ~ ~ ~ ~ 
− r0 + Yu Yv p 0 + Yw w0 + Y p Yq − u 0 + Yr 
 q 0 + Z~u − p 0 + Z~v ~
Zw − v0 + Z p
~ ~
u0 + Z q
~
Zr 
[F1 ] =  ~ ~ ~ ~ ~ ~ ~ ~ ,
 Lu Lv Lw Lp I xx r0 + Lq I xx q 0 + Lr 
 M ~ ~ ~ ~ ~ ~ ~ ~
Mv Mw I yy r0 + M p Mq I yy p 0 + M r 
 ~ u
~ ~ ~ ~ ~ ~ ~ 
 N u Nv Nw I zz q 0 + N p I zz p 0 + N q Nr 

~ ~ ~
Xξ Xη Xς 
~ ~ ~ 
 Yξ Yη Yς 
 Z~ξ Z~η Z~ς 
[G1 ] =  ~ ~ ~  ,
 Lξ Lη Lς 
M ~ ~ ~
Mη Mς
~ ξ
~ ~ 
 N ξ N η N ς 

~ ~ ~
T = T0 + ∆T , g x = g x 0 + ∆g x , g y = g y 0 + ∆g y , g z = g z 0 + ∆g z .

2.4 Incorporation of accelerometer and gyro measurement model

Generally, not all state variables in the state equation are accessible or measurable. The
common measurement variables, in most missiles or airplanes, are the angular rate
components (roll rate, p, pitch rate, q, and yaw rate, r) and the acceleration components
(ax, ay, az).

Assuming that the gyros provide ideal readings of the angular rates, we get:

pm = p (2.17a)
qm = q (2.17b)
rm = r . (2.17c)

where pm, qm and rm are the measured body rates. Normally, errors due to drifts and
noise are included. These appear as additional additive terms in equations (2.17a) to
(2.17c).
In contrast to the readings of the angular rate components, the readings of the
acceleration components are dependent on the location of the accelerometers, w.r.t. the
c.g. of the body.

12
 DSTO-TN-0449

The acceleration components measured at point O (where O is at a distance of dx, dy

and dz from the central point of gravity, c.g., along x-, y- and z-axis, respectively), may
be written as:
a x = u& + qw − rv − d x (q 2 + r 2 ) + d y ( pq − r&) + d z ( pr + q& ) (2.18a)
a y = v& + ru − pw + d x ( pq + r&) − d y ( p 2 + r 2 ) + d z (qr − p& ) (2.18b)
a z = w& + pv − qu + d x ( pr − q& ) + d y (qr + p& ) − d z ( p 2 + q 2 ) (2.18c)

If the accelerometers are mounted along the x-axis (ie. dy = dz = 0) which is usually the
case, then equations (2.18a-c) reduce to:

~ ~
a x = u& + qw − rv − d x ( q 2 + r 2 ) = X + T + g x − d x ( q 2 + r 2 ) (2.19a)
~
a y = v& + ru − pw + d x ( pq + r& ) = Y + g y + d x ( pq + r& ) (2.19b)
~
a z = w& + pv − qu + d x ( pr − q& ) = Z + g z + d x ( pr − q& ) (2.19c)

Note that the right hand side of Equations (2.19a) to (2.19c) come directly from
Equations (2.1) to (2.3).

Linearising Equations (2.19a) to (2.19c), and using the relationship (2.15) gives us:

∆y( t ) = [H 1 ]∆x( t ) + [J 1 ]∆u( t ) + ∆v( t ) (2.20)

∆y( t ) = [∆p ∆q ∆r ∆a x ∆a y ∆a z ] : is the output vector,

T
where:

~ ~ ~ ~ ~ T
0 0 0 X Y + N dx Z − M dx 
 ~u ~u ~ u ~u ~ u 
0 0 0 X Y + N dx Z − M dx 
 ~v ~v ~v ~ v ~v 
0 0 0 X Y + N dx Z − M dx 
 w w w w w  : is
~ ~ ~ ~  ~ ~ ~ 
[H 1 ] = 1 0 0 X Y
 p + ( q + I q + N )d Z + ( r − I r − M )d
 p  0 zz 0 p x   p 0 yy 0 p x  
 ~ ~ ~ ~  ~ ~ 
0 1 0 X − 2q d x Y
 q + ( p + I p + N )d  Z − M d 
 q 0  0 zz 0 q x
 q q x

 ~ ~ ~ ~ ~ ~ 
0 0 1 X − 2r d x Y + N dx  Z r + ( p0 − I yy p0 − M r )d x  
 r 0 r r  
the state output matrix,

13
DSTO-TN-0449

 0 0 0 
 0 0 0 
 
 0 0 0 
[J 1 ] =  ~ ~ ~  : is the matrix related to inputs in the
 Xξ Xη Xς 
 Y~ξ ~
+ Nξ d x
~
Yη
~
+ Nη d x
~
Yς
~
+ Nς d x 
~ ~ ~ ~ ~ ~ 
 Z ξ − Mξ dx Zη − Mηdx Zς − M ς d x 
measurement matrix,

∆u 1 ( t ) = [∆ξ ∆η ∆ς ] : is the input vector, and

T

T
 ∆T 
∆v1 ( t ) = 0 0 0 + ∆g x ∆g y ∆g z  : is the disturbance vector.
 m 

2.5 Linearised model of the airframe including fin servos

Assuming that the servo dynamics for the aileron, elevator and rudder can be
described adequately by a second order lag as:

∆ξ ks ξ
= 2 ,
∆ξ d s 2µ sξ s
2
+ +1
ω sξ ω sξ
(2.21a)

∆η ksη
= 2 , (2.21b)
∆η d s 2µ sη s
2
+ +1
ω sη ω sη

∆ζ ksζ
= 2 , (2.21c)
∆ζ d s 2µ sζ s
2
+ +1
ω sζ ω sζ

where ∆ξd, ∆ηd and ∆ζd are the demand aileron, elevator and rudder deflection,
respectively.

ksξ, ksη, and ksζ are the servo gain for the aileron, elevator and rudder,
respectively.

14
 DSTO-TN-0449

µsξ, µsη, and µsζ are the damping factor for the aileron, elevator and rudder,
respectively.

ωsξ, ωsη, and ωsζ are the natural frequency for the aileron, elevator and rudder,
respectively.

Equations (2.21a) to (2.21c) can be converted into differential equations as follows:

∆ξ&& = −ω sξ ∆ξ − 2 µ sξ ω sξ ∆ξ& + k sξ ω sξ ∆ξ d ,
2 2

(2.22a)

∆η&& = −ω sη ∆η − 2 µ sη ω sη ∆η& + k sη ω sη ∆η d ,
2 2
(2.22b)

∆ζ&& = −ω sζ ∆ζ − 2 µ sζ ω sζ ∆ζ& + k sζ ω sζ ∆ζ d .
2 2
(2.22c)

Hence, the state-space model for the autopilot of a missile including the servos and
airframe is:
∆x& 2 ( t ) = [A2 ]∆x 2 ( t ) + [B2 ]∆u 2 ( t ) + ∆w2 ( t ) , (2.23)
where

[
∆x 2 ( t ) = ∆u ∆v ∆w ∆p ∆q ∆r ∆ξ ∆η ∆ζ ∆ξ& ∆η& ∆ζ& , ]T

Note that the aileron, elevator and rudder deflection now become state variables.
Hence, the dimension of the state vector is increased to [12 × 1].
∆u 2 ( t ) = [∆ξ d ∆η d ∆ζ d ] ,
T

The inputs are now the demanded aileron, elevator and rudder deflection.

T
 ∆T 
∆w2 ( t ) =  + ∆g x ∆g y ∆g z 0 0 0 0 0 0 0 0 0 ,
m 

15
DSTO-TN-0449

 0 0 0 
 0 0 0 
 
 [F ]
1
[G ]1
0 0 0 
 
 0 0 0 
 0 0 0 
 
0 0 0
[A2 ] =  

0 0 0 0 0 0 0 0 0 1 0 0
 
0 0 0 0 0 0 0 0 0 0 1 0 
0 0 0 0 0 0 0 0 0 0 0 1 
 
0 0 − ω sξ − 2 µ sξ ω sξ 
2
0 0 0 0 0 0 0 0
 
− ω sη − 2 µ sη ω sη
2

0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 − ω sζ
2
0 0 − 2 µ sζ ω sζ 

(Here matrices F1 and G1 are the same as in Equation (2.16)).

 0 0 0 
 0 0 0 
 
 0 0 0 
 
 0 0 0 
 0 0 0 
 
0 0 0
[B2 ] =  
.
0 0 0
 
 0 0 0 
 0 0 0 
 
k sξ ω sξ 
2
0 0
 
k sη ω sη
2

 0 0 
 0 0 k sζ ω sζ 
2

The output vector (or the measurement equation) is given by:

∆y 2 ( t ) = H 2 ∆x 2 ( t ) + ∆v 2 ( t ) , (2.24)

where [
∆y 2 ( t ) = ∆p m ∆q m ∆rm ∆a x m ∆a y m ]T
∆a z m ,

16
 DSTO-TN-0449

(Note that gyro drift and noise and the accelerometer bias may be added to the right
hand side of Equation (2.24)).

 M M 0 0 0
 M M 0 0 0 

 M M 0 0 0
[H 2 ] =  ,
 [H ]
1
M [J ]
1
M 0 0 0
 M M 0 0 0
 
 M M 0 0 0 

(Here matrices H1 and J1 are the same as those in Equation (2.20)).

T
 ∆T 
∆v 2 ( t ) = 0 0 0 0 0 0 0 0 0 + ∆g x ∆g y ∆g z  .
 m 

Figure 2.3 shows the block diagram of an open-loop autopilot which contains the fin
servos and airframe.

Demand Fin deflections

input Outpu
Fin servos Airframe
∆u 2 ∆y 2

Demand
input ∆x 2 ∆x 2 Output
B2 H2 ∆y 2
∆u 2

A2

Figure 2.3 A block diagram of an open-loop autopilot system.

17
DSTO-TN-0449

2.6 Lateral Auto-pilot design

For the case of small perturbation, we may assume that (u0, v0, w0, p0, q0, r0) are
identically zero. In this case, the airframe model decouples into two lateral dynamics
(pitch and yaw) and one roll dynamics. We will consider the lateral autopilot dynamics
to validate the model derived in this report.

Figure 2.4 shows the block diagram of a closed-loop autopilot system.

Fin deflections
References Outpu

∆r - Fin servos Airframe ∆y 2

Gyros/

References∆u2 ∆x 2 ∆x 2 Output
B2 H2
∆r - ∆y 2

A2

Figure 2.4 A block diagram of a closed-loop auto-pilot system.

Ignoring the instrument (gyro, accelerometer) dynamics, the measured roll, pitch and
yaw angular rates (the gyro outputs) can be expressed as inputs to the gyros multiplied
the gyro gains, Kgr, Kgp and Kgy, respectively. Similarly, the measured longitudinal
acceleration, ax, and lateral accelerations, ay and az, are inputs to the accelerometers
multiplied accelerometer gains, Kax, Kay and Kaz, respectively. The accelerometer gains
affect the steady state response and may be set to 1 for transient tests. Rescaling
accelerometer gains, after selecting gyro gains, allows a unity gain autopilot to be
designed.

18
 DSTO-TN-0449

The reference signals, generally used for testing the transient time response of the
autopilot, are the desired accelerations in yaw direction, ayd, the pitch direction, azd, and
roll rate, pd. The reference roll rate is kept at zero to assess the missile dynamics in roll
against spurious disturbances. Hence, the reference vector, ∆r, can be written as:

∆r = [∆p d ∆a yd ] ,
T
∆a zd

For a case of lateral directional control, the control input signal for the fin servos can be
derived as follows:

∆η d = ∆a zd − K az ∆a z − K gq ∆q (2.25a)
∆ς d = ∆a yd − K ay ∆a y − K gr ∆r (2.25b)

For sake of simplicity, pd is set to zero since this case only considers the lateral
directional control. As a result, the control input vector can be written as:

∆u 2 ( t ) = ∆r( t ) − K∆y 2 ( t ) , (2.26)

where K is the feedback matrix as follows:

0 0 0 0 0 0 
 
K = 0 K gq 0 0 0 K az  .
0 0 K gr 0 K ay 0 

3. Verification of the Developed Model

In order to verify the developed model, the state-space model (Equations (2.23, 2.24
and 2.26)) was converted into transfer-function form using Matlab symbolic toolbox
(see Appendix B and Appendix C.1) for comparison with the results already published
[4].

Consider the following derivatives and variables to be non-zero:

~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Lu , L p , Lξ , M q , M w , M η , N r , N v , N ζ , X u , X p , X ξ , Yr , Yv , Yζ , Z w , Z η , Z q ,u 0 , d x ,

we obtain the transfer-function between the roll rate and the aileron deflection as:

~ ~ ~ ~ ~
p( s ) Lξ s + ( Lu + X ξ − X u Lξ )
= ~ ~ ~ ~ ~ ~ , (2.27)
ξ ( s ) s 2 − ( L p + X u )s + ( X u L p − X p Lu )

19
DSTO-TN-0449

The transfer-function between the pitch rate and the elevator deflection as:

~ ~ ~ ~ ~
q(s) M ηs − M ηZw + M wZη
= ~ ~ ~ ~ ~ ~ ~ , (2.28)
η(s) s 2 − (M q + Z w )s + M q Z w − M w Z q − M w u0

The transfer-function between the yaw rate and the rudder deflection as:

~ ~ ~ ~ ~
r( s ) N ζ s + N vYζ − N ζ Yv
= ~ ~ ~ ~ ~ ~ ~ , (2.29)
ζ ( s ) s 2 − ( N r + Yv )s + N v u0 + N r Yv − N vYr

The transfer-function between longitudinal acceleration and the aileron deflection as:

~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
a x ( s ) ( X u X ξ + X p X ξ Lξ )s − X u X ξ L p + X p X ξ Lu
= ~ ~ ~ ~ ~ ~ , (230)
ξ( s ) s 2 − ( X u + L p )s + X u L p − X p Lu

The transfer-function between lateral acceleration ay and the rudder deflection as:

~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
ay( s )( Yζ + N ζ dx )s 2 + ( N vYζ dx − N ζ Yv dx + N ζ Yr − N r Yζ )s + N vYζ u0 − N ζ Yv u0
= ~ ~ ~ ~ ~ ~ ~ ,
ζ(s) s 2 − ( N r + Yv )s + N v u0 + N r Yv − N vYr

(2.31)

And the transfer-function between lateral acceleration az and the elevator deflection as:

~ ~ 2 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
a z ( s ) ( Z η + M η dx )s + ( M w Z η dx − M η Z w dx + M η Z q − M q Z η )s + M η Z w u0 − M w Z η u0
= ~ ~ ~ ~ ~ ~ ~
η( s ) s 2 − ( M q + Z w )s − M w u0 − M w Z q + M q Z w

(2.32)

~ ~ ~ ~
Furthermore, if it is assumed that X u = X ξ = Yr = Lu = 0 , the transfer-function between
the roll rate and the aileron deflection may be simplified to:

~
p( s ) Lξ
= ~ , (2.33)
ξ( s ) s − Lp

The transfer-function between the yaw rate and the rudder deflection as:

20
 DSTO-TN-0449

~ ~ ~ ~ ~
r( s ) N ζ s + N v Yζ − N ζ Yv
= ~ ~ ~ ~ ~ , (2.34)
ζ ( s ) s 2 − ( N r + Yv )s + N v u0 + N r Yv

And the transfer-function between the lateral acceleration in yaw axis, ay, and the
rudder deflection measured at the c.g. can be rewritten as:

~ ~ ~ ~ ~ ~ ~
Yζ s 2 − N r Yζ s + N vYζ u0 − N ζ Yv u0
ay( s )
= 2 ~ ~ ~ ~ ~ . (2.35)
ζ(s) s − ( N r + Yv )s + N v u0 + N r Yv

Equations (2.33) to (2.35) are identical to those of the transfer-functions presented in

[P.Garnell and D.J.East [Equations (4.6-6), (4.6-8) and (4.6-7)].

The state-space model was used for simulation of open-loop and closed-loop responses
for a typical missile, using the same values as those used in [4] (see Appendices C and
D).

Figures 2.5 and 2.6 show the lateral accelerations of the missile due to a step input to
the rudder and elevator, respectively, for an open loop simulation. As can be seen from
these figures, there are large steady state errors. However, the steady state errors can
be reduced with a feedback loop as can been seen in Figures 2.7 and 2.8. These
simulation results are similar to the results presented in [4].

800

700

600
Lateral acceleration, ay [m/s 2]

500

400

300

200

100

0
0 0.5 1 1.5 2 2.5 3
Time [s]

Figure 2.5 Open loop simulation: Lateral autopilot response to a step demand acceleration

21
DSTO-TN-0449

800

700

600
Lateral acceleration, az [m/s 2]

500

400

300

200

100

0
0 0.5 1 1.5 2 2.5 3
Time [s]

Figure 2.6 Open loop simulation: Lateral autopilot response to a step demand acceleration.

60

50

40
Lateral acceleration, ay [m/s2]

30

20

10

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time [s]

Figure 2.7 Closed loop simulation: Lateral autopilot response to a step demand acceleration.

22
 DSTO-TN-0449

60

50

40
Lateral acceleration, az [m/s 2]

30

20

10

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time [s]

Figure 2.8 Closed loop simulation: Lateral autopilot response to a step demand acceleration.

4. Conclusions
Both the non-linear and linearised autopilot models have been derived in this report.
The state-space model of a missile autopilot was validated by comparing the model
with the other published results, and through both open and closed-loop systems
simulation. The non-linear dynamics model presented as structural quadratic algebraic
system is novel and will be used for developed non-linear control techniques suitable
for missile systems high g- manoeuvres and operating of a range of aerodynamics
conditions. The models developed in this report are useful for further research on
precision optimum guidance and control. It is hoped that the higher order model with
motion and inertial coupling will provide more accurate representation of missile
autopilot dynamics and should be used for adaptive and integrated guidance and
control of agile missiles.

5. References
1. Babister, A.W., Aircraft Dynamic Stability and Response, Pergamon, 1980.
2. Blakelock, J.H., Automatic Control of Aircraft and Missiles, John Wiley & Sons,
Inc., 1965.
3. Cook, M.V., Flight Dynamics Principles, Arnold, 1997.

23
DSTO-TN-0449

4. Faruqi, F.A., On the Algebraic Structure of a Class of Nonlinear Dynamical Systems,

DSTO report (to be sent for publication).
5. Garnell, P. and East, D.J., Guided Weapon Control Systems, Pergamon Press, 1977.
6. Kuo, B.C., Automatic Control Systems, 6th Edition, Prentice-Hall International
Editions, 1991.
7. Pallett, E.H.J., Automatic Flight Control, 3rd Edition, BSP Professional Books,
1987.

24
 DSTO-TN-0449

Appendix A: State Equation

A.1. Quadratic State Vector

Since the state equations (Equations (2.7) and (2.9)) include the quadratic terms, which
are comprised of all the combinations of state variables, we define two separate state
vectors, one linear state vector, and one quadratic state vector.

Let define the linear-state vector as:

x 1 = [u v w] ,
[1] T
(A1.1)

x 2 = [p q r ] ,
[1] T
(A1.2)

x
[1]
[ []
= x1
1 [1]
x2 ]
T
= [u v w p q r] ,
T
(A1.3)

[2 ] [1]
We shall consider the quadratic-state vector x corresponding to x . The quadratic-
state vector will be defined as a vector whose elements are components of a
homogeneous quadratic polynomial of these states taken in the same lexicographic
order. That is, the quadratic-state vector may be written as:

x
[2 ]
[
= u 2 uv uw up uq ur v 2 vw vp vq vr w 2 wp wq wr p 2 pq pr q 2 qr r 2 ]
T
(A1.4)

There are 21 terms in this quadratic state vector. Note that the dimension of the
n(n + 1)
quadratic state vector is when the dimension of the linear-state vector is n.
2
This type of representation has been used by other authors [3, 4] when describing high
order state combinations of dynamical systems. In the rigid body dynamic equations
(2.7) and (2.9), coefficients of a number of these terms are zero. For the sake of
simplicity (to avoid setting large number elements in the matrices to zero), only those
quadratic states that are associated with non-zero terms are retained. That is, the
[2 ]
quadratic-state vector x and its partitioned form may be written as:

25
DSTO-TN-0449

x
[2 ]
[ []
= x1
2
x2
[2 ]
] ∆ [uq ur vp vr wp wq p2 pq pr q2 ]
T
qr r 2 (A1.
5)

A.2. Linearisation of the Quadratic Vector

Given the quadratic-state vector (A1.4), the first-order locally linearised vector is given
by:

 (u0 ∆q + q0 ∆u )  q0 0 0 0 u0 0 
 (u ∆r + r ∆u )   0 0 0 u0 
 0 0   r0 0
 (v0 ∆p + p0 ∆v )   0 p0 0 v0 0 0 
   
 (v0 ∆r + r0 ∆v )   0 r0 0 0 0 v0   ∆u 
(w0 ∆p + p0 ∆w)  0 0 p0 w0 0 0   ∆v 
 (w ∆q + q ∆w)  0  
0  ∆w
[
∆x [2 ] = ∆x [12 ] ∆x [22 ] ] ∆  0

0
2 p0 ∆p
=
 0
0
0
q0
0
0
2 p0
w0
0

0   ∆p 
    
 ( p0 ∆q + q0 ∆p )   0 0 0 q0 p0 0   ∆q 
 ( p0 ∆r + r0 ∆p )   0 0 0 r0 0 p0   ∆r 
 2 q0 ∆q  0 0 0 0 2 q0 0 
   
 (q0 ∆r + r0 ∆q )   0 0 0 0 r0 q0 
  0 (A2.1)
 2 q0 ∆q   0 0 0 0 2 r0 

Note that (u0, v0, w0, p0, q0, r0) are local operating states.

We shall write this relationship in a compact form as:

∆x
[2 ]
[
= ∆x 1
[2 ]
∆x 2 = [X 0 ] ∆x
[2 ]
] [1] (A2.2)

The matrix [X 0 ] is defined via the equation (A2.1).

A.3. Calculation of Inverse Matrix

Given a matrix [G ] defined as:

 A − F − E
[G ] = − F B − D 
 − E − D C 

26
 DSTO-TN-0449

Its inverse [G ]
−1
is given by:

(
 BC − D 2 ) (CF + DE ) (DF + BE )
[G ]−1 1 
= (CF + DE )
∆
(AC − E 2 ) (AD + EF ) ,
(DF + BE ) (AD + EF ) (AB − F 2 )

(
where ∆ = ABC − AD 2 − BE 2 − CF 2 − 2 DEF . )

27
DSTO-TN-0449

Appendix B: From State-Space Form into Transfer-

function
Consider a system given as:

x& ( t ) = Ax( t ) + Bu ( t ) , (B.1)

and y( t ) = Cx ( t ) + Du ( t ) . (B.2)

Laplace transformation of Equations (B.1) and (B.2) yields:

sX(s) = AX(s) + BU(s) , (B.3)

and Y(s) = CX (s) + DU(s) . (B.4)

Equation (B.3) can be rearranged as:

X(s) = (sI − A) −1 BU(s) . (B.5)

Substituting (B.5) into (B.4), we obtain:

Y ( s ) = [ C( sI − A ) −1 B + D ]U ( s ) = H ( s )U ( s ) . (B.6)

Hence, the transfer-function of the system is:

Y(s)
H(s) = = C(sI − A) −1 B + D . (B.7)
U(s)

28
 DSTO-TN-0449

Appendix C: Matlab Codes

C.1. Converting the state-space model into the transfer- functions

The following m.file converts the state-space model into transfer-functions

clear all;

% Define the deravatives and parameters

syms Xu Xp Xxi f;
syms Yv Yr Yxi Yzeta f;
syms Zw Zq Zeta f;
syms Lu Lp Lxi f;
syms Mw Mq Meta f;
syms Nv Nr Nxi Neta Nzeta f;
syms u v w p q r f;
syms xi eta zeta xi_dot eta_dot zeta_dot f;
syms xi_d eta_d zeta_d f;
syms ks ws mus f;
syms ax ay az f;
syms dx u0 f;
syms s;

% Define the elements of the matrix A

a11=Xu; a12=0; a13=0; a14=Xp; a15=0; a16=0;
a21=0; a22=Yv; a23=0; a24=0; a25=0; a26=-u0+Yr;
a31=0; a32=0; a33=Zw; a34=0; a35=u0+Zq; a36=0;
a41=Lu; a42=0; a43=0; a44=Lp; a45=0; a46=0;
a51=0; a52=0; a53=Mw; a54=0; a55=Mq; a56=0;
a61=0; a62=Nv; a63=0; a64=0; a65=0; a66=Nr;

% Define the elements of the matrix B

b11=Xxi; b12=0; b13=0;
b21=Yxi; b22=0; b23=Yzeta;
b31=0; b32=Zeta; b33=0;
b41=Lxi; b42=0; b43=0;
b51=0; b52=Meta; b53=0;
b61=Nxi; b62=Neta; b63=Nzeta;

% Define the elements of the matrix C

c11=0; c12=0; c13=0; c14=1; c15=0; c16=0;
c21=0; c22=0; c23=0; c24=0; c25=1; c26=0;
c31=0; c32=0; c33=0; c34=0; c35=0; c36=1;
c41=Xu; c42=0; c43=0; c44=Xp; c45=0; c46=0;
c51=0; c52=Yv+Nv*dx; c53=0; c54=0; c55=0; c56=Yr+Nr*dx;
c61=0; c62=0; c63=Zw+Mw*dx; c64=0; c65=Zq-Mq*dx; c66=Yr;

29
DSTO-TN-0449

% Define the elements of the matrix D

d11=0; d12=0; d13=0;
d21=0; d22=0; d23=0;
d31=0; d32=0; d33=0;
d41=Xxi; d42=0; d43=0;
d51=Yxi+Nxi*dx; d52=0; d53=Yzeta+Nzeta*dx;
d61=0; d62=Zeta-Meta*dx; d63=0;

A=[a11,a12,a13,a14,a15,a16;
a21,a22,a23,a24,a25,a26;
a31,a32,a33,a34,a35,a36;
a41,a42,a43,a44,a45,a46;
a51,a52,a53,a54,a55,a56;
a61,a62,a63,a64,a65,a66];

B=[b11,b12,b13;
b21,b22,b23;
b31,b32,b33;
b41,b42,b43;
b51,b52,b53;
b61,b62,b63];

C=[c11,c12,c13,c14,c15,c16;
c21,c22,c23,c24,c25,c26;
c31,c32,c33,c34,c35,c36;
c41,c42,c43,c44,c45,c46;
c51,c52,c53,c54,c55,c56;
c61,c62,c63,c64,c65,c66];

D=[d11,d12,d13;
d21,d22,d23;
d31,d32,d33;
d41,d42,d43;
d51,d52,d53;
d61,d62,d63];

U=[xi;eta;zeta];

S=[s,0,0,0,0,0;
0,s,0,0,0,0;
0,0,s,0,0,0;
0,0,0,s,0,0;
0,0,0,0,s,0;
0,0,0,0,0,s];

% The transfer-function of the system is:

y(s)
% H (s) = = C(SI − A) −1 B + D
u (s)

30
 DSTO-TN-0449

% Calculate (SI − A ) −1
SIAinv= inv(S-A);

% Calculate the transfer-function

H=C*SIAinv*B +D;

% Display the transfer-function between the lateral

% acceleration, ay, and the rudder deflection,ζ.
H(5,3)

C.1. Open-loop and closed-loop simulation

%*********************************************************
%This main program simulates the response of the missile
%for a step input at the elevator and rudder.

clear all;

% Call the M-file called values.m

values;

% Call the M-file called init.m

init;

% Call the M-file called ssmodel.m

ssmodel;
for t=0:ts:5
REF=[theta_d;az_d;ay_d];
U=REF - K*Y;
X_dot_prev=X_dot;
X_dot=A*X+B*U;
X= X+(X_dot_prev + X_dot)/2*ts;
Y=C*X;
accy=[accy,Y(5,1)];
accz=[accz,Y(6,1)];
time = [time,t];
ay_d=50;
az_d= 50;
theta_d = 0;
end;
plot(time,acc1,time,acc3);
xlabel('Time [s]');
ylabel('Lateral acceleration, a_y [m/s^2]');
title('Closed loop response for a step input of 50 m/s^2 at the
rudder');
figure;
plot(time,acc2,time,acc3);
xlabel('Time [s]');

31
DSTO-TN-0449

ylabel('Lateral acceleration, a_z [m/s^2]');

title('Closed loop response for a step input of 50 m/s^2 at the
elevator');

%**************** Values.m *******************************

% Set values of the parameters
Xu=0.0; Xp=0.0; Xxi=0.0; Xeta=0.0; Xzeta=0.0;

Yv=-3; Yp=0; Yr=0; Yxi=0.0; Yeta=0; Yzeta=180;

Lu=0.0; Lp=0.0; Lxi=0.0;

Mu=0.0; Mv=0.0; Mw=-1.0; Mq=-3.0; Meta=-500.0; Mzeta=0.0;

Nv= 1; Np=0.0; Nw=0.0; Nr=-3; Nxi=0.0; Neta=0.0;

Nzeta=-500.0;

Zw=-3; Zq=Yr; Zeta =-Yzeta;

ks=0.0068; ksz=ks
mus=0.7; musz=mus;
ws=180;
dx =0.5;
u0=500;

% For the case of open-loop simulation, Kgr=Kay=Kgq=Kaz=0

Kgr= 30.75; Kay = 0.825;
Kgq=-Kgr; Kaz = Kay;

%*********************** init.m *************************

%Initialise the parameters
u=0; v=0; w=0; p=0; q=0; r=0;
xi=0; eta=0; zeta=0;
xi_dot=0; eta_dot=0; zeta_dot=0;
X=[u;v;w;p;q;r;xi;eta;zeta;xi_dot;eta_dot;zeta_dot];
X_dot=X;
Y=[0;0;0;0;0;0];
i = 1;
ay_d=0;
az_d=0;
theta_d=0;
xi_d(i)=0;
eta_d(i)=0;
zeta_d(i)=0;

ts=0.001;
accy=0;
accz=0;
time=0;

32
 DSTO-TN-0449

% ********************* ssmodel.m ***********************

% Define the state-space model
a1_1=Xu;a1_2=0;a1_3=0;a1_4=Xp;a1_5=0;a1_6=0;
a1_7=Xxi;a1_8=0;a1_9=0;a1_10=0;a1_11=0;a1_12=0;

a2_1=0;a2_2=Yv;a2_3=0;a2_4=0;a2_5=0;a2_6=-u0+Yr;
a2_7=Yxi;a2_8=0;a2_9=Yzeta;a2_10=0;a2_11=0;a2_12=0

a3_1=0;a3_2=0;a3_3=Zw;a3_4=0;a3_5=u0+Zq;a3_6=0;
a3_7=0;a3_8=Zeta;a3_9=0;a3_10=0;a3_11=0;a3_12=0;

a4_1=Lu;a4_2=0;a4_3=0;a4_4=Lp;a4_5=0;a4_6=0;
a4_7=Lxi;a4_8=0;a4_9=0;a4_10=0;a4_11=0;a4_12=0;

a5_1=0;a5_2=0;a5_3=Mw;a5_4=0;a5_5=Mq;a5_6=0;
a5_7=0;a5_8=Meta;a5_9=0;a5_10=0;a5_11=0;a5_12=0;

a6_1=0;a6_2=Nv;a6_3=0;a6_4=0;a6_5=0;a6_6=Nr;
a6_7=Nxi;a6_8=Neta;a6_9=Nzeta;a6_10=0;a6_11=0;a6_12=0;

a7_1=0;a7_2=0;a7_3=0;a7_4=0;a7_5=0;a7_6=0;
a7_7=0;a7_8=0;a7_9=0;a7_10=1;a7_11=0;a7_12=0;

a8_1=0;a8_2=0;a8_3=0;a8_4=0;a8_5=0;a8_6=0;
a8_7=0;a8_8=0;a8_9=0;a8_10=0;a8_11=1;a8_12=0;

a9_1=0;a9_2=0;a9_3=0;a9_4=0;a9_5=0;a9_6=0;
a9_7=0;a9_8=0;a9_9=0;a9_10=0;a9_11=0;a9_12=1;

a10_1=0;a10_2=0;a10_3=0;a10_4=0;a10_5=0;a10_6=0;
a10_7=-ws^2;a10_8=0;a10_9=0;a10_10=-2*mus*ws;a10_11=0;
a10_12=0;

a11_1=0;a11_2=0;a11_3=0;a11_4=0;a11_5=0;a11_6=0;
a11_7=0;a11_8=-ws^2;a11_9=0;a11_10=0;a11_11=-2*musz*ws;
a11_12=0;

a12_1=0;a12_2=0;a12_3=0;a12_4=0;a12_5=0;a12_6=0;
a12_7=0;a12_8=0;a12_9=-ws^2;a12_10=0;a12_11=0;
a12_12=-2*mus*ws;

b1_1=0;b1_2=0;b1_3=0;
b2_1=0;b2_2=0;b2_3=0;
b3_1=0;b3_2=0;b3_3=0;
b4_1=0;b4_2=0;b4_3=0;
b5_1=0;b5_2=0;b5_3=0;
b6_1=0;b6_2=0;b6_3=0;
b7_1=0;b7_2=0;b7_3=0;
b8_1=0;b8_2=0;b8_3=0;
b9_1=0;b9_2=0;b9_3=0;

33
DSTO-TN-0449

b10_1=-ks*ws^2;b10_2=0;b10_3=0;
b11_1=0;b11_2=-ksz*ws^2;b11_3=0;
b12_1=0;b12_2=0;b12_3=-ks*ws^2;

c1_1=0;c1_2=0;c1_3=0;c1_4=1;c1_5=0;c1_6=0;
c1_7=0;c1_8=0;c1_9=0;c1_10=0;c1_11=0;c1_12=0;

c2_1=0;c2_2=0;c2_3=0;c2_4=0;c2_5=1;c2_6=0;
c2_7=0;c2_8=0;c2_9=0;c2_10=0;c2_11=0;c2_12=0;

c3_1=0;c3_2=0;c3_3=0;c3_4=0;c3_5=0;c3_6=1;
c3_7=0;c3_8=0;c3_9=0;c3_10=0;c3_11=0;c3_12=0;

c4_1=Xu;c4_2=0;c4_3=0;c4_4=Xp;c4_5=0;c4_6=0;
c4_7=Xxi;c4_8=Xeta;c4_9=Xzeta;c4_10=0;c4_11=0;c4_12=0;

c5_1=0;c5_2=Yv+Nv*dx;c5_3=0;c5_4=0;c5_5=0;c5_6=Yr+Nr*dx;
c5_7=Yxi+Nxi*dx;c5_8=Yeta+Neta*dx;c5_9=Yzeta+Nzeta*dx;
c5_10=0;c5_11=0;c5_12=0;

c6_1=0;c6_2=0;c6_3=Zw-Mw*dx;c6_4=0;c6_5=Zq-Mq*dx;c6_6=0;
c6_7=0;c6_8=Zeta-Meta*dx;c6_9=0;c6_10=0;c6_11=0;c6_12=0;

A=[a1_1,a1_2,a1_3,a1_4,a1_5,a1_6,a1_7,a1_8,a1_9,a1_10,a1_11,a1_12;
a2_1,a2_2,a2_3,a2_4,a2_5,a2_6,a2_7,a2_8,a2_9,a2_10,a2_11,a2_12;
a3_1,a3_2,a3_3,a3_4,a3_5,a3_6,a3_7,a3_8,a3_9,a3_10,a3_11,a3_12;
a4_1,a4_2,a4_3,a4_4,a4_5,a4_6,a4_7,a4_8,a4_9,a4_10,a4_11,a4_12;
a5_1,a5_2,a5_3,a5_4,a5_5,a5_6,a5_7,a5_8,a5_9,a5_10,a5_11,a5_12;
a6_1,a6_2,a6_3,a6_4,a6_5,a6_6,a6_7,a6_8,a6_9,a6_10,a6_11,a6_12;
a7_1,a7_2,a7_3,a7_4,a7_5,a7_6,a7_7,a7_8,a7_9,a7_10,a7_11,a7_12;
a8_1,a8_2,a8_3,a8_4,a8_5,a8_6,a8_7,a8_8,a8_9,a8_10,a8_11,a8_12;
a9_1,a9_2,a9_3,a9_4,a9_5,a9_6,a9_7,a9_8,a9_9,a9_10,a9_11,a9_12;
a10_1,a10_2,a10_3,a10_4,a10_5,a10_6,a10_7,a10_8,a10_9,a10_10,a10_11,
a10_12; a11_1,a11_2,a11_3,a11_4,a11_5,a11_6,a11_7,a11_8,a11_9,a11_10,a11_11,
a11_12; a12_1,a12_2,a12_3,a12_4,a12_5,a12_6,a12_7,a12_8,a12_9,a12_10,a12_11,
a12_12];

B=[b1_1,b1_2,b1_3;
b2_1,b2_2,b2_3;
b3_1,b3_2,b3_3;
b4_1,b4_2,b4_3;
b5_1,b5_2,b5_3;
b6_1,b6_2,b6_3;
b7_1,b7_2,b7_3;
b8_1,b8_2,b8_3;
b9_1,b9_2,b9_3;
b10_1,b10_2,b10_3;
b11_1,b11_2,b11_3;
b12_1,b12_2,b12_3];

34
 DSTO-TN-0449

C=[c1_1,c1_2,c1_3,c1_4,c1_5,c1_6,c1_7,c1_8,c1_9,c1_10,c1_11,c1_12;
c2_1,c2_2,c2_3,c2_4,c2_5,c2_6,c2_7,c2_8,c2_9,c2_10,c2_11,c2_12;
c3_1,c3_2,c3_3,c3_4,c3_5,c3_6,c3_7,c3_8,c3_9,c3_10,c3_11,c3_12;
c4_1,c4_2,c4_3,c4_4,c4_5,c4_6,c4_7,c4_8,c4_9,c4_10,c4_11,c4_12;
c5_1,c5_2,c5_3,c5_4,c5_5,c5_6,c5_7,c5_8,c5_9,c5_10,c5_11,c5_12;
c6_1,c6_2,c6_3,c6_4,c6_5,c6_6,c6_7,c6_8,c6_9,c6_10,c6_11,c6_12];

D= [0,0,0;
0,0,0;
0,0,0;
0,0,0;
0,0,0;
0,0,0];

K = [0,0,0,0,0,0;
0,Kgq,0,0,0,Kaz;
0,0,Kgr,0,Kay,0];

35
DSTO-TN-0449

Appendix D: Values of the Non-Zero Derivatives and

Parameters used in the simulation

~ ~ ~ ~ ~
Yv = −3 , Yζ = 180 , N v = 1, N r = −3, N ζ = −500 ,
~ ~ ~ ~ ~
Z w = −3, Z η = −180 , M w = −1, M q = −3, , M η = −500 ,
k sξ = 0.0068 , k sη = 0.0068 , k sζ = 0.0068 ,
µ sξ = 0.7 , µ sη = 0.7 , µ sζ = 0.7 ,
ω sξ = 180 , ω sη = 180 , ω sζ = 180 ,
u0 = 500 , dx = 0.5

Feedback gains:

Kgr= 30.75; Kay = 0.825;

Kgq=-Kgr; Kaz = Kay;

36
 DSTO-TN-0449

DISTRIBUTION LIST

Mathematical Models for a Missile Autopilot Design

Farhan A. Faruqi and Thanh Lan Vu

AUSTRALIA

DEFENCE ORGANISATION

}
S&T Program
Chief Defence Scientist
FAS Science Policy shared copy
AS Science Corporate Management
Director General Science Policy Development
Counsellor Defence Science, London (Doc Data Sheet)
Counsellor Defence Science, Washington (Doc Data Sheet)
Scientific Adviser to MRDC Thailand (Doc Data Sheet )
Scientific Adviser Joint (Doc Data Sheet and distribution list only)
Navy Scientific Adviser (Doc Data Sheet and distribution list only)

Scientific Adviser - Army (Doc Data Sheet and distribution list only)

Air Force Scientific Adviser

Director Trials

Systems Sciences Laboratory

Chief of Weapon Systems Division
RLAWS
RLMWS
RLLWS
RLEWT
Guidance and Control Group, WSD (6 Copies)
Farhan A. Faruqi
Thanh Lan Vu

DSTO Library and Archives

Library Edinburgh 2 copies
Australian Archives

Capability Systems Staff

Director General Maritime Development (Doc Data Sheet only)
Director General Aerospace Development (Doc Data Sheet only)

Knowledge Staff
Director General Command, Control, Communications and Computers (DGC4)
(Doc Data Sheet only)
DSTO-TN-0449

Army
ABCA National Standardisation Officer, Land Warfare Development Sector,
Puckapunyal (4 copies)

Intelligence Program
DGSTA Defence Intelligence Organisation
Manager, Information Centre, Defence Intelligence Organisation

Defence Libraries
Library Manager, DLS-Canberra
Library Manager, DLS - Sydney West (Doc Data Sheet Only)

UNIVERSITIES AND COLLEGES

Australian Defence Force Academy
Library
Head of Aerospace and Mechanical Engineering
Hargrave Library, Monash University (Doc Data Sheet only)
Librarian, Flinders University

OTHER ORGANISATIONS
National Library of Australia
NASA (Canberra)
AusInfo

OUTSIDE AUSTRALIA

INTERNATIONAL DEFENCE INFORMATION CENTRES

US Defense Technical Information Center, 2 copies
UK Defence Research Information Centre, 2 copies
Canada Defence Scientific Information Service, 1 copy
NZ Defence Information Centre, 1 copy

ABSTRACTING AND INFORMATION ORGANISATIONS

Library, Chemical Abstracts Reference Service
Engineering Societies Library, US
Materials Information, Cambridge Scientific Abstracts, US
Documents Librarian, The Center for Research Libraries, US

INFORMATION EXCHANGE AGREEMENT PARTNERS

Acquisitions Unit, Science Reference and Information Service, UK
Library - Exchange Desk, National Institute of Standards and Technology, US
National Aerospace Laboratory, Japan
National Aerospace Laboratory, Netherlands

SPARES (5 copies)

Total number of copies: 51

 Page classification: UNCLASSIFIED

DEFENCE SCIENCE AND TECHNOLOGY ORGANISATION

DOCUMENT CONTROL DATA 1. PRIVACY MARKING/CAVEAT (OF DOCUMENT)

2. TITLE 3. SECURITY CLASSIFICATION (FOR UNCLASSIFIED REPORTS

THAT ARE LIMITED RELEASE USE (L) NEXT TO DOCUMENT
Mathematical Models for a Missile Autopilot Design CLASSIFICATION)

Document (U)
Title (U)
Abstract (U)

4. AUTHOR(S) 5. CORPORATE AUTHOR

Farhan A. Faruqi and Thanh Lan Vu Systems Sciences Laboratory

PO Box 1500
Edinburgh South Australia 5111 Australia

6a. DSTO NUMBER 6b. AR NUMBER 6c. TYPE OF REPORT 7. DOCUMENT DATE
DSTO-TN-0449 AR- 012-430 Technical Note August, 2002

8. FILE NUMBER 9. TASK NUMBER 10. TASK SPONSOR 11. NO. OF PAGES 12. NO. OF REFERENCES
J9505-21-207 01/219 DST 31 7

13. URL on the World Wide Web 14. RELEASE AUTHORITY

http://www.dsto.defence.gov.au/corporate/reports/DSTO-TN-0449.pdf Chief, Weapons Systems Division

15. SECONDARY RELEASE STATEMENT OF THIS DOCUMENT

Approved for public release

OVERSEAS ENQUIRIES OUTSIDE STATED LIMITATIONS SHOULD BE REFERRED THROUGH DOCUMENT EXCHANGE, PO BOX 1500, EDINBURGH, SA 5111
16. DELIBERATE ANNOUNCEMENT

No Limitations

17. CITATION IN OTHER DOCUMENTS Yes

18. DEFTEST DESCRIPTORS

Mathematical models, Automatic pilots, Missile guidance, Design criteria

19. ABSTRACT
This report considers the derivation of the mathematical model for a missile autopilot in state space form. The
basic equations defining the airframe dynamics are non-linear, however, since the non-linearities are
“structured” (in the sense that the states are of quadratic form) a novel approach of expressing this non-linear
dynamics in state space form is given. This should provide a useful way to implement the equations in a
computer simulation program and possibly for future application of non-linear analysis and synthesis
techniques, particularly for autopilot design of missiles executing high g-manoeuvres.

This report also considers a locally linearised state space model that lends itself to better known linear
techniques of the modern control theory. A coupled multi-input multi-output (MIMO) model is derived
suitable for both the application of the modern control techniques as well as the classical time-domain and
frequency domain techniques. This is validated by comparing the model with the other published results, and
through both open and closed-loop systems simulations. The models developed are useful for further research
on precision optimum guidance and control. It is hoped that the model will provide more accurate
presentations of missile autopilot dynamics and will be used for adaptive and integrated guidance & control of
agile missiles.

Page classification: UNCLASSIFIED

View publication stats